Adel el Hallak has a DGX Spark sitting in his house. It runs AI jobs overnight, and if he forgets to give it something to do before bed, he feels bad about it.

"At home I feel guilty if I don't kick off my job for it to work on all night," el Hallak, Nvidia's VP of Product, tells Tom's Guide. "It doesn't make noise. And I feel guilty if I don't do that for something on a personal front — because it's plugged in anyway."

It's a small, endearing detail from a wide-ranging interview — but it also points at where AI is headed, particularly local AI. Because when the person in charge of Nvidia's AI product strategy is already living with a personal AI computer in his home, treating it more like an appliance than a server, it could mean a lot of us will be doing the same thing in just a few years.

The case for ditching the cloud

Right now, most people interact with AI through a browser or app. That means, when you prompt ChatGPT or Perplexity or any other AI assistant, your data travels to a data center, and a response comes back. It works, but there's a cost — both financially and in terms of privacy. Every question you ask, every document you upload and every personal detail you share lives on someone else's infrastructure.

El Hallak thinks local AI changes that equation. "I could run a good enough model on my system at home," he said. Essentially, if the model runs on hardware you own, your data never leaves your house, and there's no monthly subscription.

The math is worth thinking about. Nvidia's DGX Spark currently retails for $4,699 — not cheap by any measure. But a ChatGPT Plus subscription runs $20 per month, or $240 per year. Over five years, that's $1,200 — and you still don't own anything. The Spark, meanwhile, runs unlimited queries with no per-token billing, no usage caps, and no internet required.

That's not an apples-to-apples comparison, of course. GPT-6 and Claude are significantly more capable than the open-source models you'd run locally on a Spark. But for a growing number of tasks — summarizing documents, drafting emails, searching your own files, running a coding assistant — a local model with 128GB of memory and access to Nvidia's full CUDA stack is more than enough.

It's not just for developers

The DGX Spark launched last October as what Nvidia called "the world's smallest AI supercomputer." At the time, it was squarely aimed at developers, researchers, and data scientists. It runs Linux. It weighs 2.6 pounds. It's the size of a Mac Mini.

But the ecosystem building around it is starting to look more consumer-friendly. As I wrote last month, Perplexity just launched Portable Computer, which packages a local AI model, agent tools, app connectors, and a sandboxed runtime into a single system that runs on DGX Spark hardware. Every task starts on your machine. If the local model hits something it can't handle, it asks your permission before sending that one step to the cloud. Your files stay put.

El Hallak sees this as a trend, not an outlier. He pointed to harness providers — the companies building the software layer between a model and a user — making setup dramatically easier. "You're not doing installs via a command line or a pip install," he said. "They're making it easier — like the old days of using Windows and setting up an application. They're bringing it back to a GUI."

If that sounds familiar, it should. It's the same arc that turned network-attached storage from a hobbyist pursuit into something Synology sells at Best Buy. The hardware existed for years before the software made it approachable.

The RTX Spark laptops change the picture

The DGX Spark desktop is the proof of concept. But the real consumer play may be what's coming next: RTX Spark laptops, which pack the same GB10 silicon into notebook form and are expected to ship this fall from Dell, HP, Lenovo, Asus, MSI, Acer, and others. Microsoft is building a Surface Laptop Ultra around the chip.

These aren't AI-only machines — they run Windows 11, play games, and handle conventional laptop workloads. But their 128GB of unified memory means they can also run large language models that would choke any current consumer GPU. The laptop version of what el Hallak has sitting on his desk at home, but without the Linux-only limitation.

"The fact that you will have the model running locally with a harness and a runtime that are made easier to install — I think it's going to really usher in local AI," el Hallak said.

The real argument isn't performance — it's ownership

El Hallak told a story from the enterprise world that translates perfectly to consumers: "There was a point where people were proud of the leaderboards and how many tokens they're using. Then somebody in their IT department got their bills, and all of a sudden I became a cost optimizer."

Anyone who's watched their Claude usage creep up — or hit the limits of a free tier — knows the feeling. And as AI agents get more capable and start touching more personal data (your calendar, your email, your finances), the question of where that processing happens isn't theoretical anymore.

"Local AI is just going to bring more tokens, and tokens equate to intelligence," el Hallak said. "When you pair intelligence with a harness, you're able to do things, and then the runtime secures it."

Or, as he put it more casually: "Data centers in the house."

More from Tom's Guide

• AI is turning ordinary photos of kids into sexual abuse imagery — here's how to protect your family
• OpenAI paid contractors to read ChatGPT conversations — here's how to protect yourself
• I stopped reading the fine print myself after discovering this overlooked ChatGPT trick

For eight grueling months leading up to the end of last year, engineers from Microsoft and Nvidia worked side by side from the silicon bootstrap level upward. And the goal? Making Windows 11 play nice with an unprecedented architectural shift: an Arm-based CPU sharing massive unified memory with a desktop-class Nvidia RTX GPU.

And then, the first RTX Spark machine finally booted. The red LED lit up, and as Microsoft's Principal Product Manager, Peter Dawoud, recalls it: “half of Santa Clara was jumping.” This was a huge undertaking, but now the real work begins: fine-tuning everything.

RTX Spark isn’t just another laptop sticker; it pushed Microsoft to fundamentally overhaul five foundational subsystems inside Windows 11 to bridge the gap between low-power Arm efficiency, top-tier AAA gaming, and local agentic AI.

I spoke to Dawoud about the software underpinning Team Green’s hardware because, as he said, “Windows kind of had to get all these subsystems right all at the same time, so that you can enjoy an RTX Spark device and take the most advantage out of it.”

A Prism Emulator overhaul

Take it from someone who has watched the many attempts to run Windows on ARM chips (and covered the Snapdragon X Elite launch): emulation has always had an Achilles’ heel when it comes to gaming: micro-stuttering and frame pacing.

“When you're playing video games, it's not just about your average FPS. 1% lows, stuttering is really critical, and your CPU actually plays a very big role in stuttering control,” Dawoud explained.

While average frame rates look acceptable, CPU timing can get erratic when you try to emulate a game and play it at the same time, which can make fast-paced gaming feel jerky. When somebody buys an Nvidia laptop in October, the chances are likely that they will want to play games on it — it’s an RTX GPU after all.

So work needed to be done, and Microsoft has overhauled the Prism emulation layer for RTX Spark’s pipelines. By shoring up how tasks are assigned to each of the CPU’s threads, frame rates are much more stable in my testing, and the improvement over past versions is night and day.

UI polish & the agentic future

If you’re going to make what I believe could be a MacBook Pro killer, you need to make sure the UI is ready for power users as well as casual users. And the mission in Dawoud’s mind to do this is clear: “make the OS kind of get out of the way, for you to do the work that you want to do.”

Microsoft is cleaning up Windows 11 for this kind of user: native taskbar customization, customizable notifications, killing unwanted widgets, and remembering displays you connect to—to preserve specific screen layout settings.

Then we turn our attention to the “reinvent the PC” part of the RTX Spark, which comes down to local agentic AI. I saw a demo of Hermes Agent connected to a user’s Google account that acted like his personal chief of staff: setting an agenda for the day, drafting email responses in Gmail, and updating a project tracker in Google Docs. And it all happened on-device without a single bit of cloud AI.

But without guardrails, this could get scary. “Having an agent that you can't control is probably the thing you don't want,” he stated. So the answer is Microsoft Extension Containers (MXC) — isolated sandboxes for your local models to act across your desktop securely without sensitive system control leaking.

Killing OEM bloatware

It’s always been a dream of mine to have a laptop that doesn’t come packed with OEM-specific power management software or an antivirus scanner. While this may not fix the latter, it goes a long way toward killing the need for the former.

Dawoud emphasized the need to eliminate the clunky tug-of-war between Windows and external software. Currently, there are several ways to tweak power management across different apps, but RTX Spark uses the open-source Modern Power and Thermal Framework—letting Windows natively control all thermal and power tuning.

“Users, when they want to set the best performance, they can do it in Windows, and they'll know that the OEM's tuning for best performance has now been set,” he said. “They don't need to go to another app and be like, 'Okay, is that app saying good performance and Windows is saying good performance at the same time?'”

This also helps these laptops maintain the same performance on and off the charger.

The UMA secret

On paper, a Unified Memory Architecture (UMA) sounds great — a big pool for both the CPU and monstrous GPU to dip into at leisure, and take however much (or little) they need. But in reality, CPUs and GPUs digest memory completely differently.

CPUs crave small memory chunks (4-kilobyte pages) to distribute tasks across their many cores. They take a pizza, pull all the ingredients off of it, and share them around. GPUs, however, take that entire pizza, fold it in half, and eat it taco-style. They need massive blocks (up to 2-megabyte pages) to plow through parallel 3D rendering or AI tasks.

“RTX Spark has a very powerful GPU, so making sure that we give it as much memory as possible when it needs it, but also give it the right page sizes, is really critical,” Dawoud commented.

So Microsoft's fix is to add variable page sizing directly to the memory manager. This is a bit complex, but it means Windows can dynamically read the pages of data flowing through and send them to the right places—the 2MB to the GPU and 4KB to the CPU simultaneously.

“Optimizing that is the way you can weaponize the entire stack to get the most performance out of it,” He emphasized.

Taming Nvidia’s custom CPU cores

Unlike Qualcomm’s standardized Snapdragon X chips, Nvidia’s 18- and 20-core Grace CPUs use a proprietary cluster arrangement of performance and efficiency cores to maximize speed while delivering all-day battery life.

Dawoud explained that Nvidia's core configuration is entirely unique, and “previous Arm cores that have run on Windows are not like that.”

Now, historically, supporting an exotic CPU layout required rewriting the entire kernel underpinning Windows — a risky move for sure. But Windows 11 has a new profiling stack that can tune itself dynamically on the fly, meaning the scheduler that throws tasks everywhere can adapt to Nvidia’s unique chip here without breaking legacy compatibility.

The PC has finally grown up

The 14-year partnership between Microsoft and Nvidia finally feels like it is paying off. Dawoud pointed out that this wasn't a sudden alliance — the journey started over a decade ago with the Tegra-powered Surface RT.

But this specific launch represents a fundamental shift in how the two giants collaborate. By letting Microsoft engineers rebuild the OS from the silicon bootstrap level upward, RTX Spark gave Windows the foundation it needed to finally make Arm-based PC gaming a reality.

Microsoft and Nvidia haven’t just bolted a GPU onto an Arm processor; they’ve rewritten the rulebook beneath the surface. This is a huge moment for both companies, and if Windows on Arm succeeds where previous attempts faltered, it will all be traced back to that one moment Microsoft spent eight agonizing months building the foundation to make it feel effortless.

More from Tom's Guide

• I held the Asus ProArt P14 next to my MacBook Pro 14 — and Apple should be genuinely terrified of Nvidia’s RTX Spark
• I tested Acer's wild Project DualPlay Mini — and it's a gaming handheld and laptop in one gadget
• I wanted to hate Nvidia’s DLSS 5 ‘AI filter’ — but the real DLSS 5 is photorealistic black magic

So while NBA 2K27 is officially the first game to launch with Nvidia’s newly launched DLSS 5, you’ve probably already seen it pop up in hundreds of more games than that. And I’ll say it again just like I said it before — in photorealistic games, it’s fantastic. But If you venture out of that into anything more stylized, it’s going to look weird.

That’s why it makes the world of sense in 2K27, but during my hands-on testing with my Asus ROG Zephyrus G14 (armed with an RTX 5070 Ti), I noticed something bizarre. The moment I entered a menu or triggered a replay, it suddenly felt like it was trudging through molasses.

I dug into the raw telemetry data to figure out what was going on, and what I found completely changed my understanding of Nvidia’s “3D-guided neural rendering.” That intensive new graphics pipeline of DLSS 5 may indeed be a graphical revolution — but on my humble laptop (and lower end desktops), I get the sense Frame Generation is a requirement to make it playable.

The laptop I’m playing on

I’m seeing a lot of people test it on massive desktop rigs. I’ll of course get round to doing this when I’m not on my honeymoon (my wife will kill me if I started benchmarking DLSS 5 in Cancun), but first I wanted to show you some numbers on something that isn’t really being covered here — my Asus ROG Zephyrus G14 gaming laptop.

Nvidia’s been talking about how the model running this mode of neural rendering has been made super efficient over the past few months. In fact, when it first debuted, it ran on a separate RTX 5090 desktop GPU; whereas now, it’s able to run on a single card with the game.

So let’s give it a go in a notebook with a lower power level running through those graphics.

By the numbers

To figure out why the game was suddenly stuttering, I ran the telemetry on three different game states with DLSS 5 turned on and off. And as you can see, while the gameplay numbers are very close matching, elsewhere, the difference is pretty significant.

In fact, this kind of lines up with PC Gamer’s report of a 50-60% drop in performance. So what’s going on here? Let’s talk about it.

The heavy cost of the transformer

So why does turning off the new tech triple frame rate in menus and replays? Because DLSS 5 is not just an upscale filter. It is a highly complex transformer model that interacts with the on-screen geometry and injects neural rendering.

Speaking to Nvidia, it’s clear that this AI pipeline is a heavy one, and the answer to why you don’t see that same impact in gameplay is a simple one: Frame Generation.

During live gameplay, I engaged the 6x Frame Generation multiplier to take that slower neurally rendered base and fill the gaps. However, developers typically disable frame gen during static menus, UI-heavy pause screens or camera-cut replays.

You can see how much this first iteration of DLSS 5 is sweating the system by looking at a couple of factors in the gameplay. Yes, the framerate is barely touched, but the 1% Lows takes a dip and the latency is upped quite drastically, as well as the GPU utilization spiking.

In a controller-centric game like NBA 2K27, that’s not necessarily a noticeable hit, but I can imagine that when this in its current form is moved to games that require faster inputs with a keyboard and mouse, it could become more apparent.

Take away the prop of frame generation in menus, and you see the DLSS 5 penalty more realistically on an RTX 5070 Ti laptop GPU.

But it looks incredible

Like I said in my first impressions, when faced with photorealistic graphics, DLSS 5 is undeniably impressive. There is one very specific takeaway when I looked at these “before and after” pairs of screen captures: the core game look is preserved impressively.

DLSS 5 does a flawless job of preserving the game’s visual integrity. The shots of Jalen Brunson and Victor Wembanyama show off the micro-details on the skin and the way the sweat catches the harsh stadium lighting. The upscaler isn’t adding any aggressive smoothing or plastic-looking artifacts — it’s remarkably lifelike.

Then there’s the fabric and material textures. Brunson’s Knicks jersey and the fabric of the couch in the cutscene get an upgrade to show off the stitching more with high-resolution clarity, and without any noticeable AI-induced blurring.

The lighting reacts to on-screen characters more realistically, there isn’t any noticeable AI “fizzing” or ghosting around finer textures like hair — something upscalers have had problems with in the past.

Is this a frame-generated future?

DLSS 5 integration proved to me that the uncanny valley is dead, but it also proves that native frame rates are on their way out with it.

Nvidia themselves keep saying that Moore’s law is dead, and getting crafty with AI is what will bridge that generational gap. That threshold has been crossed — photorealism is getting too heavy for high-end silicon to render natively.

And DLSS 5 proves that future of photorealism isn’t just about rendering better pixels; it’s about accepting that a good chunk of the frames you see on your screen aren’t technically going to be “real” at all.

We’ve been coming to this moment in the PC gaming space for a while now, and it’s also happening on consoles too.

And personally, I have no problem with it if it looks this good — provided it’s used the right way (such as in photorealistic games).

More from Tom's Guide

• Best of IFA 2026: We found the 26 top new gadgets
• I held the Asus ProArt P14 next to my MacBook Pro 14 — and Apple should be genuinely terrified of Nvidia’s RTX Spark
• I tested Acer's wild Project DualPlay Mini — and it's a gaming handheld and laptop in one gadget

As I keep saying, there’s a big two-device compromise problem. I love my 14-inch MacBook Pro for its rigidity, battery life, silent thermals, great ergonomics, and raw performance for my prosumer workload. But macOS locks me out of native PC gaming (mostly), so I need another machine to do that.

So when Nvidia’s RTX Spark was announced, I saw the possibility of a future daily driver pull into view — the everything machine I’d always wanted. And then when I held the Asus ProArt P14 in my right hand and the 14-inch M5 MacBook Pro in my left, it became clear I’d found my top pick of Team Green’s roster.

That’s because Asus didn’t just build a competitor; they pretty obviously reverse-engineered the MacBook Pro’s footprint and low-profile industrial aesthetic. It was going to take a lot to get me to switch from my Apple lifestyle, but coming as close to the Apple feel as possible is a helpful start, to say the least!

So for the first time, a Windows machine has the silicon architecture to challenge Apple Silicon, and Asus is packing it into a machine that rolls tanks onto the MacBook Pro’s turf.

Design and build: spotting the physical differences

Stacking them directly, you can see how much Asus took inspiration from Apple, as the ProArt P14 matches the MacBook Pro 14 almost millimeter for millimeter. But there are subtle differences.

First is the finish. Apple’s anodized aluminum has a smoother, more metallic texture, whereas Asus’ stealthy matte charcoal coating has a much more interesting texture under your fingers.

And second is the weight balance. Holding them both one-handed, the Pro definitely feels denser, but the distribution feels better balanced. Whereas in the ProArt P14, it’s lighter but more heavily weighted toward thermal management at the back of the system (more on that later).

Ports-wise, both are effectively set up for creators with full-size HDMI, high-speed USB-C, a headphone jack, and a full-size SD card slot. But the ProArt flex here comes with the legacy USB-A port on the right flank — a minor detail that saves you from carrying dongles.

Screen showdown: 2.8K OLED vs Liquid Retina XDR

So when it comes to picking between these screens, I know the default is to say “OLED is better,” but there are levels to this.

Because yes, the infinite contrast, instant pixel response times for gaming, zero blooming, and no camera notch give the ProArt P14 an advantage. It’s a gorgeous display — this is Asus’ bread and butter after all, as colors melt off the screen.

But the MacBook Pro (I’m assuming, as we won’t know until full testing — this is just based on what’s always happened before) will likely have a higher sustained peak brightness for outdoor work.

Really, it comes down to what you’re going to do with it. For gaming and color-accurate video grading in indoor/studio settings, the OLED wins on pure punch. However, for working out and about in bright coffee shops, Apple’s Mini-LED will probably retain the top spot.

The tactile friction point: Key feel and trackpad

The MacBook Pro has always been one of my favorites ergonomically speaking —packing shallow but snappy scissor keyboard switches set in a well. It feels familiar, reliable, and fatigue-free, but it does lack the travel for deep gaming presses.

Meanwhile, the ProArt P14’s keys have a deeper dish, a little more key travel, and a softer landing underneath. The added depth creates a more satisfying impact, and the WASD stability for gaming feels a little better. But the slight added mushiness may come back to haunt me if I type all day.

Then there’s the trackpads. Now don’t get me wrong, the ProArt’s physical trackpad is massive and smooth, with a nice tactility to the physical click. But Apple’s Force Touch haptics are basically witchcraft, with an identical click anywhere on the glass and zero chance of any mechanical dead zones.

Under the hood

Currently, I’m on an M5 MacBook Pro, but you can spec it up to M5 Pro or M5 Max with unified memory shared between CPU and GPU cores. The end result is lightning-fast performance across all your tasks that has always been one step ahead of the competition from Qualcomm, Intel, and AMD.

And now, the ProArt P14 is taking on Apple with the RTX Spark N1X platform — marrying an Arm-based Nvidia Grace CPU with a dedicated Blackwell RTX GPU, backed up by up to 128GB of unified memory.

In terms of how much you can spec it, ProArt P14 configurations start from that baseline N1X with fewer cores and 24GB of memory, so the versatility of configs means you can run the full gamut against the MacBook Pro.

Apple’s machine falls down in modern AAA PC gaming backlogs — some of it works through Metal compatibility, but most will be left to crippling emulation or cloud streaming.

But with N1X’s CUDA cores, the ProArt P14 can run 1440p games at 100 FPS, complete with DLSS support and anti-cheat compatibility. Beyond that, I expect they’ll be pretty evenly matched in creator performance and AI. Differences there will come down to the software optimization for the silicon.

The potential dealbreakers

I’m just going to summarize some quick things that may become the thorn in the ProArt P14’s side once we get one in the labs to fully test.

• The fan noise: When I tested the ProArt P14 for gaming, it was the loudest RTX Spark laptop in the room. The MacBook Pro stays near-silent during everyday tasks, and hopefully the P14 is the same — but it’s clear that an insane amount of Nvidia silicon requires aggressive airflow.
• Battery endurance: Nvidia is claiming “all-day” battery life and the same performance both on and off the charger. During my testing, I can confirm the latter, but we don’t know about the former yet! If it can achieve that 12-14+ hours of mixed real-world productivity, Apple loses its biggest single advantage.
• The price: The MacBook Pro starts at $1,999. To get the equivalent amount of memory (24GB), that price goes up to $2,199. This is the big question mark for N1X laptops, and hopefully they’ll be competitive on cost.

Early verdict

This is a way-too-early verdict, as I’m piecing together the few times I’ve had to test the ProArt P14 and its potential against months of MacBook Pro use. But I’ll give it a stab.

More from Tom's Guide

• I tested Acer’s wild flip-and-swivel handheld — why Project DualPlay Mini is the ultimate netbook love letter
• I wanted to hate Nvidia’s DLSS 5 ‘AI filter’ — but the real DLSS 5 is photorealistic black magic
• As a PC snob, I wanted to catch GTA 6 making console compromises. Instead, I’m in awe of the trickery

For years, creatives and gamers have had to deal with the two-device compromise: you buy an M-series MacBook Pro for the seamless productivity and all-day battery life, and keep the power-hungry Windows gaming rig at home.

I’ve tested RTX Spark a few times now, and I believe it’s the end of that compromise. And at IFA 2026, Team Green has confirmed that the RTX Spark N1X is coming this October.

Nvidia is aggressively targeting Apple’s throat. By shifting to the Arm-based Grace CPU paired with Blackwell graphics, as you’d see in its RTX 50-series GPUs, and massive pools of memory, Nvidia is finally bringing the Apple Silicon paradigm to Windows — without sacrificing PC gaming.

Impressive power in a 14mm chassis

This isn’t the first foray with Windows on Arm that we’ve seen. Qualcomm’s Snapdragon X2 Elite continues to deliver strong performance and power efficiency. But RTX Spark goes at things a different way by chucking a massive GPU at the silicon.

In my time testing, that leads to super-powered prosumer productivity performance (especially with the software integrations across Blender and Adobe’s creative suite), stellar gaming performance and enough fuel for Nvidia’s mission to “reinvent the PC” with local agentic AI. And for that power efficiency side of things, it manages to keep the same performance both on and off the charger!

Up until this point, we’d only known about the top-of-the-line N1X chip. Now we know about the baseline option (also called N1X…this could get a little confusing).

Obviously it’s worth noting that with the lower power consumption of a laptop, even though it does have a massive amount of cores, you won’t get the same level of performance as a desktop-class GPU. Just fun for comparison's sake, you know?

The form factors

We already know about eight laptops launching with RTX Spark N1X. You can see where I rank them all in a tiered list I did a couple of months ago. Basically, to match the premium specs of that chip, you’re getting premium shells that tick all the boxes needed to take on a MacBook Pro — metallic bodies, gorgeous displays (OLED or mini LED), solid ergonomics and plenty of I/O.

There are also a couple of mini PCs from Asus and MSI launching with the chip inside, and here at IFA, we’re getting word on two new systems coming to the party in October too:

• Lenovo Yoga 9n 2-in-1: Another convertible RTX Spark system has arrived to take on MSI’s Prestige Flip N16. The chassis is just 14mm thin.
• Acer SFF RTX Spark: A small form factor desktop that would look pretty perfect under my living room TV as well as my desk.

So that’s the next advantage over the MacBook — variety of designs.

The friction point: Predicting the price

While Nvidia was eager to show off the specs, it was noticeably silent on the price. Well, I’m gonna do some bac-of-the-napkin math here to fill this gap and make a prediction. Based on components and the direct competition with Apple, it’s obvious that these machines are going to carry a heavy premium.

• The baseline prediction: The lower-tier configuration (18-core CPU, 5,120 CUDA cores, 24GB RAM) will likely start around the $1,999 to $2,499 range — pricing it directly against the base MacBook Pro models.
• The agentic premium: If you want that maxed out configuration with 128GB of unified memory to run massive local AI models, expect a jaw-dropping price tag. Given that a maxed-out MacBook Pro easily pushes past the $4,500 mark, an equivalent N1X will likely land in the $4,000 to $5,000+ range.

Bookmark this and remind me in the comments if I get this wrong!

The early verdict

Way back at Computex 2026, Nvidia CEO Jensen Huang made it clear that he wants to reinvent the PC with RTX Spark. This is all fair and good, but I believe the real showcase of N1X here is not the local AI computation. That part is impressive, but it’s still up for debate on whether people will actually use their computers in this way.

The real game-changer here (to me at least) is the fundamental architectural shift that bridges the gap between serious productivity and high-end gaming — complete with all-day battery life (providing it does match up to Nvidia’s claims of course).

It’s the MacBook Pro that actually has games to play on, and if everything plays out the way I’m anticipating (and there’s no nasty surprises on the performance front), then Apple is finally going to have genuine, uncompromised competition.

More from Tom's Guide

• I wanted to hate Nvidia’s DLSS 5 ‘AI filter’ — but the real DLSS 5 is photorealistic black magic
• I tested 6 RTX Spark laptops for gaming — and the winner completely shocked me
• Tom’s Guide Computing Awards 2026: 22 game-changing laptops, desktops and peripherals

So there I was — Nvidia giving me a full breakdown of what DLSS 5 is and how it works, alongside a chance to play NBA 2K27 on multiple systems to show how it’s evolved from the controversial AI slop filter that divided opinions in the office to something much more refined.

I thought things were heading in the right direction. But then, modders extracted the leaked DLSS 5 files from 2K27 and stuffed it into unsupported games. The online reaction was immediate and fracturing: half praising the added detail while the other half declared it the AI-filtered destroyer of frame rates.

Now that the embargo has lifted and I can (finally) talk about DLSS 5 (coming September 3 to NBA 2K27 first), as someone who spent time with the official integration, these mods are missing the point. Brute forcing a neural rendering file into a game without engine-level hooks is like judging a luxury car after hot-wiring it with missing spark plugs.

The big question: is DLSS 5 just an AI filter?

So the answer, technically, is no. But there’s a lot to it. When you think about AI filters, they slap a 2D post-processing effect over a completed image. As I quickly learnt when talking to the team behind DLSS 5, this is so much more.

This is a 3D-guided neural rendering pipeline — ingesting all that vital 3D geometry data, such as surface normals, semantics of lighting, motion vectors and scene depth, to work within the physical rules of the game engine rather than hallucinating.

And with all that information, it does something Nvidia calls real-time determinism. Think of it like the neural equivalent of when a bar is full — the rendering queue has a one-in/one-out policy.

Unlike offline video gen models that process chunks of video and take a guess at what it’ll look like (probabilistically), DLSS 5 processes every single frame in real-time to maintain stability between each one (deterministically).

This means that instead of seeing somebody’s face change with every scene change in AI-generated video, the neural rendering on offer here doesn’t just imagine something new out of nowhere.

What’s new since we saw DLSS 5?

When the next generation DLSS was announced earlier this year, the reaction across Tom’s Guide’s gaming Slack channel was swift and severe. The AI beautification of characters felt weird, and Digital Foundry’s test was combed over to see how much of the lighting and atmosphere had changed for the worse.

Of course, this was an extremely early build, and a lot has changed since then. Nvidia was quick to tell me they should’ve communicated DLSS 5 better at GTC, and Team Green is laying everything out about this new upscaler to us.

Put simply, Nvidia’s providing some very strong tools here, but ensuring artistic intent remains at the heart of how you work with them — rather than having them replace the artistry.

Performance reality check

If you were to look at the DLSS 5-injected modding over the weekend, it’d be fair to believe that frame rate takes a huge hit when doing this frame-by-frame neural rendering.

However, that’s not the experience I had across multiple tiers of systems playing the official DLSS 5 integration on NBA 2K27. Now, it’s worth noting that these demos I tried all had multi-frame generation turned on and ramped up too.

Frame counts were not turned on, so I can’t give you anything more numbers-wise than what Nvidia has given me here. But what I can confirm is gameplay felt silky smooth with no noticeable latency.

The RTX 5090 desktop rig was of course the star of the show, with frame rates so high that it completely saturated the refresh rate of the gaming monitor. And when played using GeForce Now, it looked like a flawless 120 FPS frame rate.

But the real star of the show to me is that RTX 5060 laptop. This goes to show the huge work Nvidia’s done to shrink DLSS 5 down and optimize it to run across the entire breadth of its GPUs, and the smoothness felt like it was around the 90-100 FPS mark in play.

How does it look?

Visuals-wise, in a game like this that is actively pursuing photorealistic quality in its graphics, DLSS 5 really sings. Whether its the massive improvement in facial textures and definition in hair, the deeper contrast in and more realistic lighting, and the upgrades made to the crowd watching too, everything comes together nicely here.

If this were to placed on a game that’s more artistically stylized, it would look weird (and you’re seeing that on some of the modded work). Nvidia themselves are clear in saying that developers can just choose not to use it if that is what developers are aiming for.

But in the case of NBA 2K27, the updates made to DLSS 5 since we last saw the super filtered looks from previous are really evident. The changes are more nuanced, and with the granular levels of control, they make a big positive difference to my eyes.

A specialized weapon

So let’s get down to brass tacks. DLSS 5 is built specifically to bridge the uncanny valley in photorealistic titles. It won’t make sense for cel-shared or heavily stylized indie games. But for sports sims, cinematic shooters and immersive open worlds, to my eyes and early impressions, this feels like a generational leap.

Of course, I want to test it more to be definitive about it, which is what I’ll absolutely do when it officially drops on September 3rd with NBA 2K27 across RTX 50-series systems and GeForce Now.

One thing is clear: this isn’t an AI filter — it’s the start of real-time neural rendering. I'm looking forward to seeing how it develops, and how the people will respond to it.

More from Tom's Guide

• I tested 6 RTX Spark laptops for gaming — and the winner completely shocked me
• As a PC snob, I wanted to catch GTA 6 making console compromises. Instead, I’m in awe of the trickery
• I tested NVIDIA’s new DLSS 4.5 Ray Reconstruction — it finally fixes path tracing’s biggest flaw, but here’s the reality check

Anthropic has reportedly signed a $35 billion cloud computing deal with Nvidia-backed Lambda for access to infrastructure at a massive new data center in Texas, according to Reuters.

The project, which is being developed by Hut 8 in Nueces County, Texas, will provide Anthropic with access to roughly 350 megawatts of capacity. Nvidia GPUs will provide the computing muscle, with the additional capacity intended to help meet growing demand for Claude.

It's an eye-watering amount of money. But what's even more remarkable is that the $35 billion agreement comes less than a week after Anthropic reportedly committed another $45 billion to rent computing power from Nscale.

In other words, Anthropic has reportedly committed around $80 billion to just these two computing deals in the span of a week.

So why does the company behind a chatbot need that much infrastructure? Let's dig into it.

Claude needs a lot more compute

If you open Claude on your phone or laptop, it's easy to forget what happens after you hit send.

Your request ultimately has to be processed by specialized hardware running inside a data center. Multiply that by millions of prompts, API calls and complicated AI agent tasks, and the infrastructure requirements quickly become enormous.

And Anthropic isn't preparing for Claude's current level of demand. It's building for where it thinks that demand is headed.

Anthropic said in May that its annualized revenue had surpassed $47 billion, up from approximately $9 billion at the end of 2025. The company specifically said some of the money from its latest funding round would be used to "expand compute to meet growing demand for Claude."

Claude Code appears to be particularly important to growth

Anthropic said earlier this year that Claude Code's annualized revenue had climbed above $2.5 billion and that weekly active users had doubled since the beginning of 2026. Business subscriptions to the coding agent had quadrupled over the same period.

That's important because tools such as Claude Code can require considerably more interaction with an AI model than asking a chatbot a single question.

Instead of producing one response and stopping, an AI coding agent can inspect files, write code, run commands, analyze the results, fix mistakes and continue working through a task. Anthropic's own research into roughly 400,000 Claude Code sessions found that usage has been shifting toward more end-to-end agentic tasks, including running software, analyzing data and producing documents.

The more work we hand over to AI agents, the more inference compute companies such as Anthropic need behind them.

$35 billion isn't even Anthropic's biggest recent compute deal

Just days ago, Reuters reported that Anthropic had agreed to spend $45 billion over six years for computing capacity from Nscale's data center campus in West Virginia.

That agreement covers about 460 megawatts of capacity and is expected to use Nvidia's next-generation Vera Rubin hardware.

Anthropic has been assembling compute capacity from multiple directions. It announced a $50 billion U.S. data center investment with Fluidstack last year, and in May said a partnership with SpaceX would substantially increase the amount of compute available for Claude.

That SpaceX capacity had an immediate impact users could actually see: Anthropic doubled Claude Code's five-hour rate limits for Pro, Max, Team and some Enterprise users, removed a peak-hours restriction and raised API limits for Claude Opus models.

It's a useful example of why these enormous infrastructure deals ultimately matter to people using Claude. More compute doesn't automatically mean a smarter AI model, but it can determine how many people can use it, how frequently they can use it and how much work increasingly autonomous AI agents can perform.

Nvidia is everywhere in this deal

There's another interesting wrinkle to Anthropic's latest agreement: Nvidia seems to be everywhere.

Lambda is backed by Nvidia. The data center will contain Nvidia hardware. And according to The Wall Street Journal, Nvidia itself will hold the lease on the Texas data center.

Nvidia is also an Anthropic investor. The companies announced a broader strategic partnership last November that included plans for Anthropic to use Nvidia hardware and for the two companies to work together on optimizing future Nvidia architectures for Anthropic's AI workloads.

Even Anthropic's $45 billion Nscale agreement reportedly involves Nvidia's upcoming Vera Rubin chips.

That shows just how central Nvidia has become to the infrastructure sitting behind the current AI boom. Companies such as Anthropic may be competing with OpenAI, Google and others to build the best AI models, but much of that race still depends on securing enormous quantities of computing hardware and electricity to run them.

What this means for Claude users

Don't expect your Claude subscription to suddenly look different because Anthropic signed another data center deal. This is a much longer-term bet.

Anthropic is effectively wagering tens of billions of dollars that Claude usage — particularly through products such as Claude Code and future AI agents — is going to continue growing dramatically.

We've already seen what additional computing capacity can do. Anthropic used its SpaceX deal to raise Claude usage limits earlier this year. More infrastructure could similarly help the company accommodate more users, longer-running agents and increasingly compute-intensive models without constantly running into capacity constraints.

And with another $35 billion reportedly going toward Claude's computing needs, Anthropic clearly doesn't expect our appetite for AI to slow down anytime soon.

More from Tom's Guide

• Perplexity’s new local-first AI runs on your PC — and asks before using the cloud
• ChatGPT can learn your repetitive tasks by watching you — here are 7 prompts to try
• OpenAI is leaving Cursor in November — here are your 3 options

For years, my ultimate computing friction point has been the two-device compromise: I buy a thin and light MacBook for productivity, and keep a massive gaming rig at home for play. But after going hands-on with six Nvidia RTX Spark laptops launching later this year, I can confidently say that era is over.

Put simply, this is shaping up to be a game-changer. Packing 6,144 CUDA cores into thin and light notebooks like these, it’s bridging the gap between serious productivity and high-end gaming. And crucially, game developers are already jumping on board to support it — ensuring vital anti-cheat software compatibility is ready to go.

But packing a powerful GPU into a sleek chassis doesn’t automatically make it a great gaming machine. Other factors are just as vital like how the keyboard feels, how loud the fans are and display quality.

I spent extensive time playing on six different models, and here’s my definitive RTX Spark gaming tier list, ranked from worst to best.

A dream for gamers

Before I get there, though, this isn’t my first rodeo with Spark. I did get to try the chip out at Computex this year (on my Birthday of all days). And since that was a much earlier version of what I saw here at Gamescom, here are some general updates:

• More laptops, more fun: At Computex, only the Surface Laptop Ultra was turned on and usable. However, this time around they were all active and usable.
• Backing up their claims: Nvidia is touting 1440p gaming at over 100 FPS. Of course, with no visible FPS counter, I can’t confirm that specifically, but to the eye, all the games were running significantly over 60 FPS — settings at a range of medium to high (depending on the title) and DLSS turned on.
• So many developers on board: Turns out if you’re Nvidia, you have a fair bit of influence to get publishers on board. Windows on Arm is getting a shot in the arm thanks to anti-cheat support and a massive number of studios jumping on board to work on Arm-native versions of their games.
• Emulation looks much better: The fact Pragmata was even running through the Windows Prism emulation layer to make it play nice on an Arm chip was quite the achievement at Computex. But there definitely was a little bit of fuzziness to the gameplay looking back. Now, with four games including 007 First Light all emulated, they’re all crystal clear and playing great with no discernible latency.

Put simply, things have definitely moved on since June, and anything with that RTX Spark sticker on the body is going to be a mighty tasty machine to get, come this fall. That being said, I’m here to talk about how some of these specific laptops fare for gaming on a build level — not a chip level.

RTX Spark laptop (for gaming) tier list: Specs

C-tier: A laptop wearing a suit

Now don’t get me wrong, the Dell XPS 16 is an undeniably good-looking laptop that will easily elicit an “oh that’s a cool thing” reaction when you pull it out in a meeting. The utilitarian body is sleek and that Tandem OLED display is super vivid.

But is this a safe space? Because that keyboard is problematic for gaming to put it mildly. It’s a flat chiclet board with zero gaps between the keys. Doom: The Dark Ages exposed the gaming flaws of a keyboard like this, as the lack of key definition had me panickedly looking down mid-battle to ensure I was on the WASD.

Like I said earlier this year, with a little time you can get used to this for productivity. But for gaming, it's not quite there.

B-Tier: Productivity first, gaming second

These two are quite a bit better for gaming on an ergonomic level, but aren’t all the way there compared to some superior tactility further up this list.

First up is the Microsoft Surface Laptop Ultra: packing a delicious 15-inch mini-LED PixelSense Ultra touchscreen display with 3:2 aspect ratio. That taller screen may be more productivity-focused, but it’s gonna be great for RTS games too.

But again, the keyboard falters here slightly. Not that it’s bad, it’s a great-feeling tactile set of keys for productivity for sure. But for gaming, the dish of each key is far too subtle and the sharpness of those edges do cut into your wrists. So it firmly sits in the “it’s okay” category for gameplay.

Slightly better for play is my original personal favorite from Computex — the Asus ProArt P14. Basically, Asus understood the assignment and gave me a Windows equivalent of a MacBook Pro in its design and ergonomics. The keyboard’s better here for demanding gameplay too, and that 3K OLED display is mesmerizing.

But what do you get when you cram this much power into a small body? Loud fans to get all the heat out. To my ears at least, this was the loudest laptop in the testing room. It’s a great hybrid machine, but make sure you get a good gaming headset to drown out that fan noise.

A-Tier: Great all-rounders

Now we’re moving closer to the elite if gaming is a priority for your RTX Spark machine. The Lenovo Yoga Pro 9n boasts a great premium design with another vivid OLED display.

For gameplay, there are very few genuine downsides here — capped off by Lenovo’s trademark excellent keyboard ergonomics. The only reason it’s here in the A-tier is because one company has pulled a blinder with its board (more on that in a moment).

Then there’s the MSI Prestige N16 Flip AI+. Yes, the keyboard is a little too heavy on the RGB lighting for my personal taste, but it feels good to tap away on, the OLED screen is a lovely flash flood of color, and it’s a 2-in-1 form factor that allows for tent mode gaming with a controller. That’s something none of these other options can do.

S-Tier: The unexpected champion

Which leaves one: the HP OmniBook Ultra 16. When I first saw this laptop at Computex, it looked like a decidedly “just fine” device. I expected it to just be another laptop more suited to work over play. I couldn’t have been more wrong.

There’s of course a crispy and gorgeous OLED panel here that moves fluidly, but that’s not the real champion feature here. That belongs to the keyboard, which elevated this machine to the top of the list. It’s a chiclet board, but the well defined keys and travel make it feel almost mechanical-like. It feels fantastic during gameplay, and with slick stylings to the chassis, it’s the ultimate stealth gaming laptop.

Plenty to look forward to

The RTX Spark lineup proves that the days of compromising between a sleek daily driver and a capable gaming machine are finally coming to a close. These six distinct options will hit the market soon, and while I can scrutinize them from a gaming perspective, they’re meant for so much more.

As CEO Jensen Huang said himself when I got the chance to talk to him: “we’re focused on reinventing the PC.” There’s going to be something for everyone in this range. So I have to ask — if you’ve got the budget for one of these notebooks, which one would you spend your money on?

More from Tom's Guide

• Tom’s Guide Computing Awards 2026: 22 game-changing laptops, desktops and peripherals
• As a PC snob, I wanted to catch GTA 6 making console compromises. Instead, I’m in awe of the trickery
• I tested NVIDIA’s new DLSS 4.5 Ray Reconstruction — it finally fixes path tracing’s biggest flaw, but here’s the reality check

If you want to understand the future of gaming, you have to survive Gamescom first. With over 350,000 attendees flooding into Cologne, Germany, the city’s convention center is a chaotic, RGB-lit maze of upcoming titles, booth tours and next-generation hardware.

The Tom’s Guide team was on the ground to cut through the noise, as we spent the week walking the halls, getting exclusive behind-the-scenes access, playing a ton of titles early and getting our hands on the tech that will redefine how we play over the next few years.

From Nvidia’s astonishing new MacBook-killing superchips and Alienware’s 5-star OLED monitor, to the return of the Alien franchise and gaming furniture we actually want in our living rooms, we found the gear that truly stood out.

After covering every square inch of the show floor, here are our official Tom’s Guide Best of Gamescom 2026 awards.

Best of show: Asus ROG Xbox Ally X20

I’ll put my hands up and admit that maybe I was a little hard on the ROG Xbox Ally X20 back at Computex 2026. In the midst of Intel Arc G3 Extreme eating AMD’s lunch in handheld gaming, I was quick to count it out. But as the $1,299 price emerged for this impressively upgraded Ally, my mind changed on it quite drastically.

That OLED panel is glorious, the ergonomic tweaks with TMR joysticks, reshaped bumpers and face buttons and rubberized grips make the most comfortable PC handheld even more comfortable, and that transparent design is a nerd’s joy to behold.

Yes, I can only hope that FSR 4 comes to the AMD Ryzen Z2 AI Extreme chipset here for boosted performance in the future. But with 60 FPS being easy to achieve if you tweak the settings in AAA titles, this is a really well-rounded system for gaming enthusiasts on the go.

Jason England - Managing Editor of Core Tech

Best game: Alien Isolation 2

When it came to getting hands-on with upcoming games, I came into Gamescom 2026 with one mission: Play Alien: Isolation 2. The larger Alien franchise is a personal favorite of mine, and Alien: Isolation was a dream game come true. I still can’t quite believe we’re getting a sequel. But we are; I can confirm it’s real, and I got to play 30 minutes of it at Gamescom.

My demo was short, but it doesn’t appear to be shaking up the first game’s formula all that much. Once again, it’s a tense survival-horror experience, in which players are powerless against a stalking Xenomorph. But when you hit a home run, why change anything big?

The scene-setting prologue boasts a phenomenal sense of atmosphere and highlights the game’s biggest change: the setting has been swapped from a claustrophobic spaceship to a more expansive planet surface. Offering the same extraordinary attention to detail and movie-accurate authenticity, Alien: Isolation 2 is shaping up to be my dream sequel!

Rory Mellon - Senior Entertainment Editor

Best gaming tech: Nvidia RTX Spark

Nvidia had a very busy Gamescom with the likes of DLSS 4.5 Ray Reconstruction and key updates to GeForce Now. But hidden away in the depths of the convention center, I got the first real chance to play games on their upcoming RTX Spark machines. And put simply, I was blown away.

The promise of 1440p at over 100 FPS is absolutely on show here with the likes of 007 First Light and Arc Raiders pulling it off with near-maxed out settings and DLSS turned on. And some of these games were being emulated in real-time to be compatible with this Arm chip (as well as the anti-cheat software that runs beneath it).

With these upcoming machines, Team Green has done the one thing I’ve always wanted — a Windows MacBook Pro competitor. No longer do I need a sleek, long-lasting laptop for work and a bulky gaming machine for play. They are now combined into one notebook, and the options look awesome.

Jason England - Managing Editor of Core Tech

Best gaming monitor: Alienware 39 5K2K OLED (AW3926QW)

Alienware really took things to the next level with its fantastic Alienware 39 5K2K OLED (AW3926QW). The fact that this is the first device I’ve given a full 5-star rating to says it all. If you’re looking for the finest gaming experience possible, then you need this gaming monitor in your setup.

What makes the AW3926QW stand out is its four-stack, three-subpixel RGB stripe tandem OLED panel, which Alienware says is the first of its kind to hit the market. What all of those fancy terms mean for real-world use is higher OLED brightness, more vibrant colors, sharper text, and exceptional contrast. You can also swap between 5K at 165Hz for cinematic games and WFHD (1080p) at 330Hz for competitive gaming.

Top that with a sleek design and an immersive 39-inch curved panel, and you have one of the best gaming monitors ever. The fact that it’s a relatively affordable $1,099 for what you get almost makes this a steal. If you have room for it on your desk, then the Alienware 39 5K2K OLED (AW3926QW) comes with my highest recommendation. Trust me, you won’t be disappointed.

Tony Polanco - Senior Computing Writer

Best gaming peripheral: Genki Manta Controller

Speaking to the team behind the Manta Controller, it felt like they’d thought of everything. On the surface, this controller looks similar to an Xbox pad with a built-in 2.7-inch color display, but it’s designed to be customizable. And I mean, fully customizable. Via the display, you can tweak just about every setting, letting you adjust each element to fit your play preference.

Take, for example, its face buttons; these aren’t your standard buttons that register input at a set point (usually when fully pressed down); they are analog, letting you determine the point at which the input is registered. The triggers are the same. And that’s just the start of the options on the Manta Controller. You could spend hours tweaking everything to your liking.

For players who want an almost obscene level of control over their controller, the Manta Controller could be a vital gaming peripheral. Plus, all these adjustments can be made on the controller display rather than utilizing a separate desktop or mobile app, which lets you dive into the myriad of adjustment options without even having to put the pad down.

Rory Mellon - Senior Entertainment Editor

Best gaming accessories: Ikea x Xbox YXSTABY

Normally, gaming furniture is the kind of stuff you’re personally proud of, but I wouldn’t dare seriously consider it in my future home with my wife. It’s all a bit of an angular, RGB-laden cringe fest — that was until I got to see/sit on all the YXSTABY options created as part of the Ikea x Xbox tie-up.

The aesthetics are refined, but with cheeky nods to gamers in the aesthetics and ergonomics. For example, the joystick stool that wobbles around is both a cheeky nod to the Xbox controller, while also being optimal for leaning forward and locking in.

The same can be said for that lounge chair too, paired with some nice storage options including that rolling coffee table. This is a full suite, and it all comes together really nicely.

Jason England - Managing Editor of Core Tech

Best sim racing gear: Initium Asetek ecosystem expansion

A while ago, I called the original Asetek Initium sim racing rig one of the best investments for driving enthusiasts. And now, the company is proving that by vastly expanding the ecosystem for those who want to start bridging the gap between getting started and becoming an enthusiast.

From expanding into different disciplines like rally driving with the shifter and handbrake, improving pedal feel with the hydraulic brake expansion kit, to the much more intense-feeling 10 newton-meter DD10 hub, there’s something for everyone here to start building up from that base level.

Jason England - Managing Editor of Core Tech

Best retro gaming tech: Evercade Nexus

Evercade has been making dedicated retro consoles and handhelds since 2020. However, its latest offering, the Evercade Nexus, looks set to take on rivals like the Nintendo Switch 2 with a slick portable device that proves sometimes you just can’t beat the classics.

Compatible with Evercade’s entire library of cartridge-based software, including various arcade compilations (Atari, Interplay, Nacom, etc.) and collections dedicated to beloved franchises like Doom, Tomb Raider and Duke Nukem, the Nexus has the potential to be a retro game fan's dream handheld console. And it’ll launch with a Banjo-Kazooie double pack, which completely transported me back to being a young kid playing my first game.

It’s the first Evercade handheld with dual analog sticks, sports a 5.89-inch IPS display, and the returning “Tate Mode” lets you enjoy vertical-screen arcade titles as they were designed. It’s also not lost on me that in an era when owning your game is under threat, Evercade is designed exclusively to use physical cartridges. There’s no need for digital downloads or “Game-Key Cards” that hold nothing more than a license to install software on your console.

Rory Mellon - Senior Entertainment Editor

More from Tom's Guide

• As a PC snob, I wanted to catch GTA 6 making console compromises. Instead, I’m in awe of the trickery
• I tested the first 5K G-Sync Pulsar gaming monitor — 144Hz looks like 500Hz of clarity, and it ruined my home display
• I just went eyes-on with the world’s fastest OLED gaming monitor — and motion blur is officially dead

Super high-resolution monitors usually force you to compromise on motion clarity. Compared to lower-resolution esports panels, it’s simply next-to-impossible to push the same insane 500Hz+ refresh rates. That was until Nvidia found a way around that.

Enter the 5K AOC Agon Pro AGP327KG — the first of its kind armed with Team Green’s G-Sync Pulsar. I saw this technology at CES 2026 in some competitive-tier gaming monitors, but now it's in something with staggeringly high pixel density; to me, this is a game changer.

G-Sync Pulsar explained

What is G-Sync Pulsar? Good question! You’ve already got variable refresh rate and extreme low motion blur on deck, so what else does it do? Well, a simple way to explain it would be that Nvidia’s regulating the pulse of the monitor itself — specifically the backlight.

To make an IPS monitor like this light up, you need a backlight strobing to illuminate the pixels. In normal screens, the whole backlight will strobe independently of what’s being shown on screen. That means that as the game moves, monitor backlight strobes may reveal pixels in mid-transition, which leads to the human eye perceiving blur.

G-Sync Pulsar brings variable-frequency strobing tied to the refresh rate, so the backlight strobe pulses in lock step with every game frame (also known as a scanout). That ensures a stable image with every pop of light, which gives you a crispier picture and the perception of higher motion clarity.

Previously, Pulsar was limited to smaller 27-inch 1440p panels, and in my time testing it, I was blown away by the tack-sharp definition of everything on screen. Slapping it on a 5K 144Hz monitor shows exactly what the tech can do when given the space to stretch its legs, and I’m floored by it.

The end result

Just like the smaller screens, there was a level of clarity that meant my eyes didn’t need to focus on specific subjects on-screen. It all jumped into focus in a significant way while playing Delta Force.

Whether it’s text clarity of the signs on the wall, the targets at a distance or the on-screen elements, the traditional blur I’d become used to just became all the more noticeable. So fair warning — this tech will turn you into a display snob hunting down the absolute peak of crispiness.

Some final extra nice bits

As for the Agon Pro itself, there are a few user-friendly things I wanted to shout out. First off, this is packed with I/O, including 2x HDMI, 1x DisplayPort, 1x USB Type-C, a USB-B upstream and 3x USB-A. This has all the connectivity needed to be your central hub for your setup.

Second, there’s ambient adaptive tech such as a built-in light sensor that can automatically adjust both brightness and color temperature based on your room’s ambient lighting — nice touch for eye care and accuracy (especially with the deep contrast ratio of that glossy screen)!

Finally, there’s customizing the screen itself. There’s a joystick to get around the OSD menu, which can feel clunky. But thanks to that Nvidia tie-in, you can change some of the key settings in the Nvidia app without needing to fall out with a clunky menu.

We’re waiting on confirmed pricing and an official launch date, and given the high resolution and Pulsar inside, we should brace our wallets. But lowkey, this is looking set to be a belter of a monitor for your high-fidelity gaming needs.

More from Tom's Guide

• I sat on all the IKEA x Xbox furniture — it’s the ultimate stealth gaming upgrade
• I tested the $1,299 ROG Xbox Ally X20 — with that OLED display, the $300 price hike is surprisingly palatable
• I just went eyes-on with the world’s fastest OLED gaming monitor — and motion blur is officially dead

When you enable full path tracing in a modern game, the lighting looks stunning — but the illusion is often broken by visual artifacts like smudged reflections, ghosting details, light trails that fade weirdly and dark corners that “boil” with digital noise.

Nvidia is attempting to eradicate those issues for good with DLSS 4.5 Ray Reconstruction, and I’ve spent a few days with this tech to see if the company’s new gaming AI model can finally clean up path tracing’s messy edges.

The truth of it is the same as I’ve felt about ray tracing and path tracing at large. If you stop to pixel-peep, the visual upgrade is undeniable. But if you’re expecting a massive paradigm shift in the middle of a frantic firefight, you need to temper your expectations.

Sorry for the reality check, and credit to Team Green for managing to make trash bags shinier and more reflective in Alan Wake 2, but this is definitely an iterative upgrade.

How to activate DLSS 4.5 Ray Reconstruction

Just like a lot of Nvidia’s gaming features, you’ve got to go into the overrides. Pop into the specific game that supports DLSS 4.5 Ray Reconstruction, open the DLSS overrides window, click the dropdown on Ray Reconstruction and select “Model F.”

At that point, you’re off to the races. The only game I noticed any weird glitching on with the override turned on is Portal RTX, which I’m sure will be ironed over with an update. But everything else worked without a hitch.

What is Ray Reconstruction?

Well, the demands of path tracing are starting to outstrip the raw hardware capabilities of consumer GPUs. Calculating the real physical light paths for millions of pixels every second is just too much to do.

Instead of painting every single pixel, the game flings a handful of dots of paint across the canvas — just enough to capture rough reflections, colors and shadows. The end result is a noisy, static-filled mess of dots. Something is needed to join the dots.

The old fix was a technique graphics cards used called “denoisers.” I can go into the specifics about these hand-tuned technologies, but think of it like taking a wet sponge and smudging those paint dots together. It does the job in filling the gaps, but fine details get blurred out and subtle lighting can flicker and “boil.”

So instead, Nvidia’s DLSS Ray Reconstruction hands that unfinished canvas to an AI network trained on supercomputers with millions of flawless images. Think of it like an expert art restorer with a photographic memory — looking at the scattered dots of light, recognizing the 3D geometry of the scene, and predicting how the light naturally behaves.

The end result (as you’ll see) is crispier puddle reflections, detailed textures without blurring or ghosting and shinier, less smudged light bounces. It takes this taxing work off the GPU, so it can focus on rendering the game, while AI pattern recognition does the heavy lifting of calculating the true light data.

The faster, smarter engine in DLSS 4.5

To understand what DLSS 4.5 Ray Reconstruction is actually doing, you have to look at the math. Nvidia has replaced its own on-board denoiser with a new 2nd generation transformer model. This AI brain processes 20% more parameters and delivers a massive 35% boost in compute capability.

That means this hypothetical art restorer is now able to more deeply analyze the space, and intelligently use game engine pixel sampling and motion data to improve the accuracy and stability of lighting — doing so millions of times a second.

And the best part? Nvidia claims this massive increase in AI compute comes with a negligible performance impact.

Testing that claim, my RTX 5080 setup saw no frame rate penalty whatsoever, and for lower-end hardware in Nvidia’s 50-series family (like the RTX 5060 Ti (16GB) and the RTX 5070 Ti-armed Asus ROG Zephyrus G14 that I’m testing), the frame drops are negligible (average drop of 2 FPS).

Of course, that should extend to the RTX 40 and 50 series GPUs. But if you are running an older RTX 20 or 30 series card with the older architecture, limitations mean you may see a more noticeable performance penalty.

The pixel-peeping test

So since this is one for those that just want to scrutinize the specifics of a frame, let’s dip into this with a few games on my RTX 5080 monster rig — putting DLSS 4 and DLSS 4.5’s Ray Reconstruction head-to-head. And I’m not going to lie, when you stop to take a beat and look around, the differences in still frames are striking.

In Cyberpunk 2077, the bright red glow from the “Livepunch Booze” neon sign washes over the floor in a softer, blurrier spread with DLSS 4. Enabling 4.5 shows off the tighter light bounce and higher-contrast, which casts sharper, concentrated red and cyan reflections on those polished floor panels.

Throw in much sharper, more cohesive specular highlights and a deeper contrast in darker areas without that fuzzy noise that plagued the older denoiser, and the tweaks are small, but significant.

Moving over to Pragmata, the moving neon makes DLSS 4.5 Ray Reconstruction much more noticeable. There’s a moving laser jumping puzzle that makes it so much more obvious — seeing the lasers just fizzle away on the older model vs immediately snapping away.

Finally, in Alan Wake 2, Nvidia showed me a demo of shinier trash bags. But if anything, they’re doing the tech a bit of a disservice. This has always been a showcase game for Team Green, and something as simple as walking around a safe room shows off small detail improvements that come together to make quite a difference.

Saga’s hair, coat and surrounding surfaces are much better defined with smoother, natural lighting bouncing off them. The grain of the wood paneling goes from slightly blurred to no blur whatsoever, lighting sources bounce more tightly off surfaces, and mirror reflections are a lot sharper.

My verdict

Full credit to Nvidia here — improving image fidelity by running an AI model with 20% more parameters without tanking frame rates on modern hardware is a genuine engineering triumph.

Of course, I say that specifically about the RTX 50-series gear I tested this on. Nvidia told me that it’s not expected to meaningfully impact performance on older GPUs. I can’t test that, but we’ll find out soon enough as some of my friends in the industry dish their thoughts.

But DLSS 4.5 Ray Reconstruction is an iterative polish. It cleans up the edge cases and makes your photo mode captures look absolutely spectacular. But during actual, fast-paced gameplay, the difference between DLSS 4 and DLSS 4.5 effectively blends into the background.

That being said, for the average gamer in motion, you’ll be hard pressed to notice the differences in the midst of gameplay. For the players, this is just a nice, quiet cleanup job. A small upgrade that will become more significant as we head into a world of path traced games.

More from Tom's Guide

• 5 beginner PC building tips from the mad genius who built a $34k, 353-pound desktop with not one, but two RTX 5090s
• Best graphics cards in 2026: I've tested every GPU to find the best bang for your buck
• Exclusive: I asked Lenovo’s GM to predict the future of monitors, and it's the end of our chaotic, cable-cluttered desks and monitors that fry your retinas

Most AI assistants require you to send your questions, files and other data to somebody else’s servers. Perplexity’s new Portable Computer turns that arrangement around by running the entire AI system on hardware you own.

Announced Tuesday, Portable Computer brings Perplexity’s agentic Computer platform to Nvidia DGX Spark and other Linux machines with compatible Nvidia graphics cards.

The company says every task begins on your machine. If the local model reaches a step that requires more advanced reasoning, Portable Computer asks for permission before sending that individual step to a frontier model in the cloud. The answer then returns to the local workflow, while files and other sensitive information can remain on the device.

That could make it appealing to anyone who wants the convenience of an AI agent without automatically handing everything it touches to a cloud provider.

There is, however, a fairly expensive catch.

More than a locally installed chatbot

You can already download open models and run them on your own computer. The difficult part is turning one of those models into an AI agent that can work with files, use tools and complete a task involving several steps.

Doing that usually means finding and configuring an inference engine, agent harness, sandbox and any connectors needed to reach outside apps. You’re also responsible for getting all those pieces to work together.

Portable Computer packages that stack into one system. It includes the local model, inference engine, agent tools, app connectors and an isolated sandbox where it can execute code and use tools.

The AI can work with files stored on your computer, read PDFs, run shell commands, take dictation and search the web. Users can also connect services including Google Drive, Gmail, Slack and GitHub.

Code and tool use happen inside the local sandbox, according to Perplexity. That provides some separation between the AI agent and the rest of the machine, although I’d still want more detail about exactly what it can access and how those permissions are controlled.

It knows when to ask for help

The hybrid routing is probably the most interesting part of Portable Computer.

A smaller local model may be enough to locate a file, extract information from a PDF or carry out a straightforward series of steps. It is less likely to match the reasoning capabilities of the largest models from OpenAI, Anthropic or Google.

Perplexity’s answer is to let the task remain local until the model encounters something it cannot reliably handle. Portable Computer can then ask whether it may send that step to a cloud model.

Users don’t have to select a privacy mode or create their own routing rules. The system makes the decision while it is working, but the user still has to approve the handoff. A supposedly local AI system becomes much less private if it quietly falls back to the cloud whenever a task gets difficult.

Perplexity also says locally completed work won’t consume billing credits. Only the portion sent to the cloud will count toward cloud usage.

At launch, users can choose a 27-billion-parameter Qwen model or Perplexity’s own post-trained version, called PPLX 27B. Nvidia’s Nemotron 3.5 Lightning is expected to follow. Users will also be able to bring their own model and inference server.

The hardware is the catch

Portable Computer may run on your PC, but not just any PC.

The software launches first on Linux and requires Nvidia hardware with at least 24GB of VRAM. It supports the desktop-sized Nvidia DGX Spark as well as compatible computers running DGX OS or Ubuntu on ARM or x64 hardware.

Windows support is scheduled to arrive in September. Perplexity hasn’t announced support for macOS.

Portable Computer will be available to Perplexity Pro, Max, Enterprise Pro and Enterprise Max subscribers. Even with a compatible subscription, most people won’t have the necessary hardware sitting under their desks.

That makes this initial release more relevant to AI enthusiasts, developers and companies that already own high-end Nvidia systems than someone deciding between ChatGPT and Perplexity on an ordinary laptop.

Still, the idea behind it could reach a much wider audience. AI assistants are gaining access to increasingly personal information, including email, work documents and files stored on our computers. Running more of that work locally gives users another way to decide what leaves their machine and when.

Portable Computer isn’t ready to replace a cloud chatbot for most people. It does, however, offer a glimpse of what a more private AI assistant could look like once the hardware needed to run one becomes easier to afford.

More from Tom's Guide

• The best way to use ChatGPT for writing isn't writing — 7 prompts I use to make it my creative partner
• 5 features that give Claude the edge over ChatGPT
• I don’t use AI to write my emails — here’s how I use it to manage my inbox instead

Lenovo’s Legion laptops traditionally sit comfortably in the chunky, high-performance desktop replacement zone. They’re thick, with heavy power supplies, and abysmal battery life. However, this year’s Legion 5i changes that formula.

The Lenovo Legion 5i Gen 11 offers a more streamlined chassis and better battery life compared to previous models, but still offers the same robust gaming performance, a crisp OLED display, along with Lenovo’s industry-leading keyboard for an all-around powerhouse of a laptop. If not for a sub-par onboard audio experience and expensive upgrades, it would be perfect.

The real question is: are those downsides enough to keep the Legion 5i Gen 11 from becoming one of our best gaming laptops?

Lenovo Legion 5i Gen 11: Cheat sheet

• What is it?: The Legion 5i Gen 11 is a more toned-down version of Lenovo’s Legion Pro, offering a fantastic balance between powerful gaming performance and crisp visuals with more affordable pricing compared to the Pro model.
• Who is it for?: Gamers and STEM students who want a powerful RTX 50-series laptop with Lenovo’s sturdy build quality, while sticking to a more affordable price tag.
• What does it cost?: The starting price for the Legion 5i Gen 11 is $1,899 while our review spec costs a bit more, running the tab up to $2,825. However, Lenovo is running a special on an RTX 5060 configuration that brings the price down to just $1,899.
• What do we like?: The new, sleeker chassis makes the Legion 5i a bit more portable, and the improved battery life is fantastic for students or working professionals. The Legion 5i also sports one of Lenovo’s fantastic keyboards, a crisp OLED display, and a powerful Nvidia RTX 50-series GPU.
• What don’t we like?: Upgrading your RAM and storage, or switching to the RTX 5070 model come at steep premiums which make the affordable Legion 5i Gen 11 a much less attractive bargain.

Lenovo Legion 5i Gen 11: Specs

Lenovo Legion 5i Gen 11: The ups

The Legion 5i is Lenovo’s mid-tier gaming laptop, and it’s got plenty to recommend it as your next gaming laptop. Between the new sleek chassis, crisp OLED display, fantastic keyboard, useful battery life, and powerful NVIDIA RTX 50-series graphics the Legion 5i has some serious selling points.

Sleek redesign

Lenovo has been thinning out the Legion lineup a bit over the years, and the 11th Gen Legion ditches the clunky hub on the back for a more beveled edge across the entire laptop. While it doesn’t shave a ton off the Z-height of the laptop, the loss of the hub and the bevelling make the Legion 5i feel much slimmer despite being nearly equal in thickness.

The design overhaul also makes the Legion 5i a bit easier to carry, even with a slightly heavier weight than the 10th Gen Legion 5i which was 0.8 inches thick and weighed 4.1 pounds. The 11th Gen is 0.79 inches thick and weighs 4.25 pounds.

Overall, the new design still keeps some recognizable Legion touches from the smudge-magnet Eclipse Black colorway to the Legion logo on the top cover panel, and Lenovo’s camera and microphone hub at the top edge.

While this doesn’t drastically change the dimensions of the Legion, it does make it feel more like a modern gaming laptop than previous iterations (particularly the 9th Gen and earlier) which were starting to feel dated.

Industry-leading keyboard

Lenovo makes good keyboards. That’s just kind of a universal fact at this point. And the Legion 5i Gen 11 is no exception to that rule.

The typing experience is the perfect combination of crisp with deep keytravel to avoid bottoming out your fingers while frantically typing instructions to your party in the in-game text window. It’s both satisfying and comfortable, which is a hard sweet-spot to hit for a laptop keyboard.

The off-set arrow keys also make navigating games a little easier and the included num-pad is great for binding macros and additional skills for input heavy RPGs and MMOs.

The Legion’s keyboard features 1.6mm key travel with a 0.3mm dish on each key, with swappable keycaps and Lenovo’s Spectrum RGB software for controlling the 1-zone or 24-zone RGB depending on your keyboard upgrades.

And even with just 1-zone RGB, you still get the same 1.6mm key travel and 0.3mm dish, which means if you don’t go for the 24-zone RGB upgrade, you still get the same great typing experience.

Crisp display

OLED displays offer some serious gaming benefits, and they’re even better when you’re using your gaming laptop to binge a TV series over a weekend. Getting an OLED display on a mid-range gaming laptop isn’t always a guarantee, so it's incredible to see that as a standard panel option on the Legion 5i.

From crisper blacks and more vibrant colors, the Legion 5i’s display looks absolutely stunning when watching nature documentaries like The Loneliest Whale or when playing visually intense games like Avowed. Toggle on HDR, and you’ve got a gaming laptop that can actually display the final season of Game of Thrones.

While my eyes absolutely enjoyed my time with the Legion 5i, that experience is also backed up by our lab testing. The Legion 5i Gen 11 covered 137.5% of the DCI-P3 color gamut with an average brightness of 490 nits and a peak HDR brightness of 1,049 nits.

This is pretty great for an OLED, and absolutely fantastic when compared to other mid-range gaming laptops like the MSI Stealth 16 AI+ or HyperX Omen 16 Valorant Edition which both lack HDR support and manage just 85% and 81% coverage of the DCI-P3 spectrum respectively.

Solid RTX 5070 gaming chops

While the Legion 5i isn’t Lenovo’s flagship gaming laptop, it should still have solid gaming performance. Particularly our review model which features the powerful Intel Core Ultra 7 251HX and Nvidia GeForce RTX 5070 combination.

My current go-to games aren’t exactly strong benchmarks so I expected the Legion 5i to have no problems whatsoever running Baldur’s Gate III and Final Fantasy XIV on high graphics presets at the Legion’s max 1600p resolution. In Baldur’s Gate III, I had the Legion running on High graphics presets at 1600p and was easily getting over 90fps.

In Final Fantasy XIV, I had the Legion running High (Laptop) graphics presets at 1600p and could get frame-rate peaks of over 100fps though the game tended to settle around 90fps in most zones. Any frame drops I noticed were more due to pop-in rate and high player traffic than anything hardware wise with the Legion.

To make the Legion 5i work a bit harder, I also booted up Avowed which has a tendency to run at low frame-rates because of its intense graphics and poor optimization. The Legion did stumble here a bit, but then I’ve also seen RTX 5090 laptops struggle with Avowed.

However, while keeping my resolution set to 1600p (because Avowed sometimes gets awkward on lower resolutions) , opting for Medium graphics settings got me a perfectly playable game which held at about 59 to 61 FPS, and I didn’t lose too much graphic detail either.

In our lab tests, the Legion also performed well. Our review unit for the Legion 5i Gen 11 features the upgraded Nvidia GeForce RTX 5070 (12GB) model, which outperformed the HyperX Omen 16 Valorant Edition which featured the 8GB version of the RTX 5070. The Legion also held the lead over the MSI Stealth 16 AI+ which featured the RTX 5060.

This meant the Legion 5i averaged 9 fps more than the HyperX and 21 FPS more than the Stealth on Black Myth Wukong at 1080p on Medium settings. The Legion’s lead died down a bit on Cinematic settings and 1600p resolution. The Legion also held a 9 FPS lead on the HyperX on Cyberpunk 2077 at 1080p and a 6 FPS lead at 1600p. The Stealth lagged behind the Legion by 12 FPS at 1080p and 8 FPS at 1600p on Cyberpunk.

While the Legion can still struggle to surpass 60fps on intense titles at max resolution, it can easily clear the 60 FPS mark on High or Ultra settings at 1080p. Which is pretty impressive for an RTX 5070 laptop. It does help that Lenovo opted for the 12GB VRAM model which gives the Legion’s RTX 5070 just enough extra horsepower to clear the competition.

Respectable battery life

Legion laptops rarely win awards for having long battery life, and usually that’s expected. Legion laptops tend to sit in the desktop replacement sector, where you don’t really expect much in terms of unplugged performance for laptops that feature power-hungry CPUs and GPUs.

However, the Legion 5i Gen 11 does pack a pretty respectable punch when it comes to web surfing and gaming battery life. The Legion 5i lasted 7 hours and 3 minutes on our web surfing battery test and it survived an impressive 1 hour 27 minutes on the PCMark 10 Gaming battery life benchmark.

For context, the HyperX Omen 16 lasted just 2:41 on our web surfing battery test, while the MSI Stealth was one of our longer-lasting gaming laptops with a web surfing battery life of 10:49. For a Legion laptop to get over 7 hours of web surfing battery life? That’s a pretty damn impressive turnaround. The Legion 5i Gen 9 barely made it three and a half hours.

As for the PCMark 10 Gaming battery benchmark, the HyperX Omen 16 lasted 1:30 and the Stealth lasted 1:55. It’s pretty rare to see a gaming laptop hit 2 hours on the PCMark 10 test since any discrete GPU is going to put a major drain on the laptop’s internal battery. However, getting 1:27 is pretty solid. It’s not a great amount of game-time but it is enough that you don’t need to immediately worry if your cat unplugs your laptop mid-raid.

Lenovo Legion 5i Gen 11: The downs

The Legion’s downsides aren’t particularly plentiful, and there are ways to avoid these pitfalls. You can avoid the pricey upgrades by sticking with the pre-configured $1,899 model and you can use a quality headset or set of speakers to get better audio, they are problematic enough to be considered downsides.

RAMaggeddon Strikes Again

Gaming hardware has always been expensive, but budget gaming hardware is now almost non-existent.

Lenovo is currently offering a pre-configured version of the Legion 5i Gen 11 at $1,899. That model features the same Intel Core Ultra 7 251HX CPU, 16GB RAM, 1TB SSD, and 15-inch OLED display as our review configuration. However, it has the Nvidia RTX 5060 onboard instead. This is actually cheaper than the $2,355 starting configuration with just 512GB of SSD storage.

And really, that $1,899 price tag wouldn’t be considered budget just a year ago, but the RAM shortage has made any gaming laptop under $2,000 feel like a steal.

Unfortunately, the base configuration is more expensive, so if Lenovo ever stops running that sale, the Legion 5i goes from somewhat affordable to expensive. Especially if you need to upgrade the RAM, SSD storage, CPU, or GPU as those will come at steep premiums.

Upgrading to the Intel Core Ultra 9 290HX processor costs an additional $110 while upgrading to the Nvidia RTX 5070 graphics card adds $300 to the total price. Upgrading your storage to 1TB will set you back $190, upgrading to 2TB will set you back an additional $430 on top of that, while increasing your memory from 16GB to 32GB will come with a $370 price-tag.

Which is how you get our RTX 5070 configuration at $2,825. And that’s the kind of prices we saw on RTX 5090 laptops just a year ago. So if you want the Legion 5i Gen 11, you better grab the pre-configured model while Lenovo still has that special running. Because otherwise its hard to justify the price.

Don’t rely on the built-in audio

Considering the Legion 5i’s status as Lenovo’s mid-range Intel gaming laptop, a certain level of quality is expected on all fronts. And audio is no exception.

While laptop audio setups often have to make sacrifices due to the form factor, some laptops offer more robust built-in sound systems than others. And while the Legion 5i’s audio is perfectly usable, it’s hardly enjoyable.

Lenovo packed a quad (2x2W) Harman speaker system with Nahimic Audio drivers into the Legion 5i chassis. Unfortunately, those speakers are bottom-firing which leads to some issues based on your gaming surface.

Downward-facing laptop speakers rely on audio engineering to bounce the audio to the user, so they can be less consistent in audio quality than top or side-firing audio systems.

While the Legion 5i’s speakers are perfectly fine at lower volume levels (40% and below), the bottom-firing speakers get muddied, and everything sounds overly-sharp with no bass when played at higher volumes.

So it’s not exactly ideal for a movie or TV show binge. And while you might be able to play a casual game with the Legion’s speaker system, I would opt for a quality studio headset for anything more intense than Runescape.

Lenovo Legion 5i Gen 11: Verdict

Traditionally, a Lenovo Legion laptop is a chunky, high-performance desktop replacement laptop that you’d rarely dream of carrying to school or to work.

Between the previous generation and this current gen, that’s starting to change. The Legion 5i offers a sleeker, more modern chassis that’s much easier to carry, and it’s got battery life that’s actually useful for getting some work done.

Combined with one of Lenovo’s industry leading keyboards, a crisp OLED display, and the same robust gaming performance you’d expect from a Legion laptop and the Legion 5i is a potent powerhouse.

It could be the perfect laptop, if only its onboard audio system was a bit more robust and its base price and upgrades weren’t so expensive.

If you are going to nab the Legion 5i Gen 11, I would absolutely recommend sticking with the discounted $1,899 preconfigured model as it's actually affordable and comes with nearly all the same features as our review model.

The gaming performance will be a bit weaker, but it’s not worth the nearly $1,000 increase in price just to upgrade to the RTX 5070 model.

When I built my first PC last year, the sheer terror of plugging everything in correctly almost broke me. Mistakes were made and I’m definitely not going to have a personal expert making sure I don’t do something wrong like I did back then!

That’s something I learned when I built my second PC over the summer. There’s a whole bunch of conflicting advice on how to make building a breeze, so who can I turn to?

I didn't go to a standard YouTube tutorial. I went to Robbin Snijders (known by his artistic alter ego ‘warbb’) — the mad genius who just built 'Yuga,’ a 353-pound, $34,000 cube of pure technological terror powered by dual RTX 5090s.

Meet Yuga

This has been nearly a year in the making, and as you can see from the size of it, this is a beast. So big in fact, you need to remove doors and push it into places in a crate like it’s King Kong.

It is a 32 x 32-inch cube that weighs 353 pounds — each element 3D printed to the most precise millimeter to fit together and tell this visual story of technology colliding with art. Honestly, I’d spend most of my time just looking at it, but let’s open the hood!

Inside, you’ve got two RTX 5090s, a high-end Asus ProArt motherboard to handle all that horsepower, and a 5,500 watt power supply to keep it all alive…yep, you read that right. At full load, you’ll be using the power equivalent of seven microwave ovens or 55 standard 55-inch LED TVs running simultaneously.

Keeping it cool is the largest Noctua fan I’ve ever seen at 200mm — something that I’m sure could keep me cool while I’m sleeping. In total, if you were to make this yourself with the components, the custom case and the design work, this would cost you roughly $34,000.

But how does it perform? We ran Maxon Cinema 4D, one of the most advanced 3D animation apps on the market, without a single hitch. Snijders showed me the file used to virtually build this machine, which contained over 300 layers and nothing slowed down.

Moving over to Cinebench, all that power came together to deliver a benchmark score that is roughly 3.5x faster than the fastest system we’ve ever tested at Tom’s Guide: the M3 Ultra Mac Studio.

What advice can we learn from this genius?

So before I start dreaming up elaborate heist plans to steal this system for myself, let’s be honest. The vast majority of you reading this will not get (or even need) a PC of this magnitude. But the artistry here does show off some key disciplines that you need if you want to build a tower of your own.

Because if one thing is becoming clear — in this challenging time of memory prices, going the DIY route is the best way to go for your wallet. So I spoke to Snijders about what advice he’d give to first-time builders that he used in crafting this masterpiece.

More from Tom's Guide

• Nvidia RTX Spark Geekbench scores just leaked — should Apple be worried about the fastest Windows laptops on the planet?
• Best graphics cards in 2026: I've tested every GPU to find the best bang for your buck
• SK Hynix warns 2027 will be 'worst year' for RAM prices — right as a damning report reveals something that will add fuel to the price-fixing lawsuit

Nvidia's ARM-based gaming laptops powered by the RTX Spark is around the corner and appear to be ready to launch thanks to new Geekbench listings.

The new ARM-based CPUs were first revealed at Computex in May by CEO Jensen Huang. "Our focus is to reinvent the PC. Windows is of course improving, but the basic architecture of a PC has largely been the same now for about 40 years, and we want to reinvent it," Huang told Tom's Guide.

How that looks is starting with an Arm-based design over the conventional x86 architecture, but Nvidia does have experience in that realm. The company already makes Arm gaming chips that power the Nintendo Switch 2.

The RTX Spark platform is starting out on a range of Windows gaming laptops and mini PCs with a focus on gaming and agentic AIs. It should have fairly low power consumption.

Geekbench scores revealed

Two different Geekbench listings reveal results for the RTX Spark N1X processor in 20-core and 18-core configurations.

If we focus on the 20-core configuration, it returned surprising scores with a single-core of 2,593 and a multi-core score of 23,424. The multicore is fairly impressive, comparing favorably with Intel's i9-14900KS, which managed a score of 23,785 in the same test.

View the full results here, though I will note that the final production scores might differ.

If you're unfamiliar, Geekbench measures how well the processor handles everyday tasks by running a bunch of simulated workloads from processing numbers to crunching numbers.

Single-core is all about how fast the CPU handles one task at a time, like editing a photo. Multi-core kicks in for multitasking like running several apps or rendering by using all the chip's "cores."

Here's how those numbers compare to Apple's M5, the newer Qualcomm Snapdragon X2, and Intel's Panther Lake chips.

Surprisingly, the RTX Spark performs well against the M5 and Intel X9 in multicore. Moving up to the M5 Pro and Max gets you higher scores. The same goes for the higher-end Qualcomm Snapdragon X2 Extreme, which has recorded over 25,000.

Still, the RTX seems like it will compare favorably with most every chip on the market.

Where it seems to fail is in the single-core, where scores are significantly lower. It seems it might feel sluggish in those single tasks, but running multiple apps is where the RTX Spark might shine.

This is a pretty good indication of how the RTX Spark platform will handle CPU-heavy tasks. It doesn't tell us the whole picture, especially when it comes to gaming, but we do know the GPU will support DLSS 4.5.

RTX Spark laptops are supposed to arrive later this fall with Windows 11 out of the box. It should be compatible with most Windows games, and Nvidia is reportedly ensuring the chip will work with common anti-cheat and DRM software.

Nvidia has already confirmed that it will launch with laptops from Asus, MSI, Microsoft and more.

More from Tom's Guide

• Nvidia RTX Spark is here, and it's going to 'reinvent the PC' — here's everything you need to know about this all-new laptop chip
• Intel Arc G3 is here to shake up handheld gaming, and AMD should be terrified — Acer Predator Atlas 8, a new MSI Claw and more launching soon
• I tested the Intel Core Ultra 7 270K Plus and 5 250K Plus against AMD’s best gaming CPUs — this should be a mismatch, but the results shocked me

To say that people weren’t impressed when Nvidia DLSS 5 was announced would be an understatement. Gamers and critics alike expressed concerns over heavy-handed “beauty filters” smoothing out character faces and AI overwriting developer art direction. The initial requirement of a second GPU for rendering also didn’t go over well.

At SIGGRAPH, Nvidia held a follow-up keynote that was practically built to address the concerns folks have with DLSS 5. Instead of focusing on how photo-realistic DLSS 5 could make graphics appear, the keynote centered around the amount of control artists will have over the technology.

Here’s a breakdown of what Nvidia discussed at SIGGRAPH and how DLSS 5 might not be as bad as we feared.

Granular control

As you might recall, the biggest complaint with DLSS 5 is how the graphics appeared to have an Instagram/TikTok filter slapped over them. A less charitable description would be an “AI slop” filter. Nvidia addressed this concern directly, saying that game artists would have full control over a game’s graphics.

Developers won’t be forced into all-or-nothing with AI. They’ll be able to mask out individual characters, faces, props, or background elements so DLSS 5 either ignores them or processes them with tailored settings. Structural and Tone Intensity sliders will give devs fine control over how things appear. There will also be three distinct AI models for developers to assign.

No AI overwrites

Another concern is that the AI would hallucinate on low-poly geometry or rewrite hand-crafted visuals. DLSS 5 isn’t generating a game scene from scratch but relies strictly on internal game engine buffers to preserve the artist’s original intent.

Nvidia emphasized that traditional rasterization and ray tracing still build the entire frame first, defining geometry, lighting anchors, motion vectors, and surface normals. Nvidia framed DLSS 5 as a third category of AI rendering alongside upscaling reconstruction and frame generation. The engine creates the world while the model enriches complex lighting elements.

Latency

Generative video models usually process frames in batches, which can introduce input lag. That’s obviously a non-starter for fast-paced games like CSGO 2.

Nvidia says that DLSS 5 processes the current frame using the game’s engine’s native motion vectors without looking ahead to future frames. By avoiding multi-frame batch processing, Nvidia’s goal is to eliminate both generative latency and temporal shimmering artifacts.

One GPU needed, not two

When Nvidia first demoed DLSS 5, the computational overhead was so heavy that it reportedly required dual high-end GPUs just to run smoothly. Naturally, this isn’t feasible for most players, but Nvidia says you’ll be able to run DLSS 5 on a single GPU.

To that end, Nvidia revealed that they have distilled the massive generative model into a much smaller, single-step diffusion model. The updated implementation has been engineered to run on a single RTX GPU with efficient VRAM allocation instead of requiring multi-GPU rigs that only hardcore users own.

Outlook

I’ve championed Nvidia DLSS for years, but I was one of the folks who wasn’t enamored with that initial DLSS 5 reveal. Now, after SIGGRAPH, I'm less worried about this fall's upcoming update.

We’ll have to wait to see how the final version functions, but I’m glad that Nvidia has addressed all the major concerns. If we can get richer game graphics without the infamous AI look, then we’ll all be winners.

More from Tom's Guide

• What is Nvidia DLSS? Explaining the AI tech transforming PC gaming — and why it’s getting controversial
• Nvidia's 'GPT moment' for graphics rejected by gamers; CEO Jensen Huang rushes to defend DLSS 5
• DLSS 4.5 is now live — I tested Nvidia’s upscaler to see which model you should actually use

The Asus ROG Zephyrus G16 is often one of our favorite gaming laptops, though the 2025 model lost us with its expensive price tag, sub-par keyboard, and abysmal battery life.

While the 2026 model does have better battery life to go along with its powerful CPU and GPU performance, lightweight design, and vibrant display, it still suffers from a hefty price tag and disappointing legacy features like the same chiclet keyboard and heavy proprietary power brick.But can the new Zephyrus G16’s premium features make up for its high price tag to become one of our best gaming laptops?

Asus ROG Zephyrus G16 (2026): Cheat sheet

• What is it?: The Asus ROG Zephyrus G16 is a premium, light and thin gaming laptop.
• Who is it for?: Gamers on the go, students, and professionals who need a laptop with a powerful GPU while still being able to commute to and from work or school.
• What does it cost?: Pricing for the Asus ROG Zephyrus G16 starts at $3,699
• What do we like?: The Zephyrus G16 is a portable powerhouse with strong CPU and GPU performance, impressive battery life, a sleek design, and vibrant display.
• What don’t we like?: The Zephyrus G16, like all gaming hardware in 2026, has been hit with the same memory shortage price increase which makes it far less affordable for students and gamers. The power brick is also a bit heavy for a commuter laptop.

Asus ROG Zephyrus G16 (2026): Specs

Asus ROG Zephyrus G16 (2026): The ups

The Zephyrus laptop line has always been premium with the build quality and features to match. The 2026 model is no different, with a sleek design, robust audio, vivid display, strong CPU and GPU performance, and good battery life.

A sleek gaming machine

Asus hasn’t changed much on the Zephyrus G16’s chassis this generation. So it’s still a slim, easily portable aluminum chassis with a shiny chrome stripe that runs on an angle across the top panel lid. The Asus Republic of Gamers logo is subtly embossed on the lower left-hand corner of the panel, leaving the rest of the design on the more understated side.

Combined with slim display bezels, a smooth keyboard deck, and an edge-to-edge touchpad that takes the entire vertical space between the keyboard and the edge of the palm rest, the inside deck of the laptop is as sleek as the outside panels, maintaining the Zephyrus’ modern aesthetic.

The top-firing quad-speaker audio system features Dolby Atmos tuning and achieves impressive volume with crisp clarity, so you can go without a pair of headphones. The built-in audio is loud enough to be heard over the Zephyrus’ cooling fans, so you don’t need to worry about losing audio entirely while gaming.

The 16-inch OLED display is also on the vibrant side, boasting Pantone validation for color accuracy. While our own testing validated an 80% accuracy for the display, we do expect some variation on OLED panels based on our testing methods compared to the methods used to measure display accuracy used by manufacturers like Asus.

The Zephyrus’s display is also bright, hitting an average SDR brightness of 469 nits, with a peak HDR brightness of 1,052 nits. Whether you’re using the Zephyrus G16 for a Netflix binge or for a visually immersive gaming session, it’s going to be an enjoyable experience.

Stronger CPU and GPU performance

The Asus ROG Zephyrus G16 (2025) wasn’t exactly a slouch when it came to performance, it just wasn’t quite powerful enough to justify its price tag.Compared to last year’s Zephyrus G16, the 2026 Zephyrus packs a more powerful punch in CPU multicore performance on Geekbench 6 and Handbrake, which is helpful since the 2025 Zephyrus was a bit underwhelming in terms of raw power.

The Zephyrus also gets stronger multicore performance than the MSI Stealth 16 AI+, despite having the same Intel chipset. Even when put against our best premium gaming laptop, the Razer Blade 16 (2026), the Zephyrus comes out with a slight edge on multithreaded performance.

As for the Zephyrus’ GPU performance, things get a bit more complicated. The 2026 Zephyrus scraped together a few more FPS on Cyberpunk 2077 at 1600p compared to the 2025 Zephyrus (which was running at 1440p resolution). The 2026 Zephyrus G16 also benchmarked well when compared to other RTX 5080 gaming laptops like the Razer and Lenovo Legion Pro 7i Gen 10.

The Zephyrus G16 trailed only a few FPS behind the to the Lenovo Legion Pro 7i on games like Assassin’s Creed Shadows, Black Myth: Wukong and Cyberpunk at Ultra 1080p settings, though the Razer Blade 16 outperformed both the Legion and Zephyrus G16.

I was easily able to achieve framerates of 200fps or higher on games like Doom: The Dark Ages on Ultra settings at 1600p resolution with DLSS enabled, while native performance without DLSS was around 60fps.

Highly optimized titles like Baldur’s Gate III and Final Fantasy XIV: Dawntrail also performed well above 100fps even without super sampling. So you are getting proper gaming performance even if the Razer Blade 16 might be just a bit more powerful.

All-day battery life

Gaming laptops are rarely known for having great battery life, but there have been a few notable exceptions over the years. The Asus ROG Zephyrus G16 (2026) is one of those rare gaming laptops that gets over 10 hours of web surfing battery life.

At 150-nits of brightness, the Zephyrus G16 lasted over 13 hours and 45 minutes. That’s an hour longer than the Razer Blade 16 (2026), about 3 hours more battery life than the MSI Stealth 16 AI+, and more than 3x the battery life of the Lenovo Legion Pro 7i.

Things do take a bit of a downturn when it comes to gaming battery life, however. The Zephyrus has just 55 minutes of gaming battery life on the PCMark 10 Gaming battery test, while the MSI Stealth and Razer Blade 16 both have nearly two hours of gaming battery time.

However, you’re unlikely to game on a laptop when running on battery power, so that’s less of a detriment than it could be otherwise. Just make sure to pack your charger, especially if you’re going to be doing a lot of GPU-heavy work.

Asus ROG Zephyrus G16 (2026): The downs

While the ROG Zephyrus G16 (2026) has plenty to recommend it, all those positives do come with a few downsides. The biggest one is the price-tag. But the new Zephyrus shares a chassis with the 2025 model, which means it has the same lackluster keyboard and hefty proprietary power brick.

Beyond expensive

All tech hardware has become nigh unaffordable, but some laptops have gotten hit by the supply chain price hike worse than others. The ROG Zephyrus was always a pricey, premium gaming laptop that fit into the working professional to STEM student venn diagram. But with a starting price of $3,699 for the RTX 5070 Ti version, it’s hard to see how the 2026 Zephyrus G16 could ever function as a student laptop.

Last year’s Zephyrus G16 at least had a starting configuration that was priced at $2,799. That’s still steep for your average college student but it’s still more affordable than this year’s model which is at least $900 more expensive and comes with less storage capacity than last year’s entry level configuration.

While some of this price hike is due to supply chain issues, the Zephyrus G16 has been steadily increasing in price over the past few years which is beyond the realm of the AI Data Center memory shortage. And that makes the hefty price tag of the 2026 Zephyrus even harder to justify.

Portability stops at the power brick

I’ve always appreciated the svelte design of the Zephyrus G16 and G14 laptops. They’re slim, they’re lightweight, and yet they pack a powerful punch that’s more than enough to do some high-power gaming even when on the road. Thin, portable gaming laptops are among my personal favorites.

But there’s one aspect of the Zephyrus G16 that feels stuck in the past. Between the proprietary AC-port and the heavy 250W charge brick, the Zephyrus G16 is a bit less portable than it appears. While the laptop itself weighs just 4.3 pounds, the power supply itself weighs an additional 1.3 pounds. And that brings the total carry weight of the Zephyrus G16 to 5.6 pounds.

Sure, it gets 13 hours of web-surfing battery life. But if I’m doing anything that will turn on the discrete Nvidia GPU, that battery life expectancy drops through the floor. The gaming battery life of the G16 is only 55 minutes, so the power brick isn’t an optional accessory.

It’d be one thing if Asus used a Thunderbolt 4 port for the G16. But the RTX 5080 is too power hungry for that. And thus we’re still stuck in the dark ages with a hefty external power supply that makes this ultraportable laptop just a little less comfortable to carry on a long commute.

Asus ROG Zephyrus G16 (2026): Verdict

The Asus ROG Zephyrus G16 is often one of our favorite gaming laptops, however we were rather disappointed by the 2025 model thanks to its high price tag, sub-par keyboard, and grim battery life.

The 2026 model does fix the battery life issue, but it still suffers from the same lackluster keyboard and heavy proprietary power brick as last year’s model. And it’s even more expensive.

But if you’re looking for a highly commutable, powerful gaming laptop or STEM machine, the Zephyrus G16 is still a solid choice. As long as you’ve got the cash.

Whenever I get the chance to test out the best gaming laptops, my first port of call is to actually game as a form of benchmarking. For years, Control has long been one of my go-to titles, but that will quickly change once Control Resonant arrives on PC.

Horrifying, bleak and messy — not exactly "positive" words to use for developer Remedy Entertainment's upcoming action-heavy sequel, but I mean these in the most endearing way possible from what I saw when playing Control Resonant on an RTX 5090 gaming PC during a three-hour hands-on session.

Fighting off hordes of The Hiss (the main threat in the game) is thrillingly horrifying with their grotesquely detailed forms, the bleak atmosphere of a warped Manhattan is enthralling and the mind-bending, twisted environments you get to explore are satisfyingly messy. If you thought The Oldest House in Control was a trip, just wait for what Control Resonant has to offer.

With Nvidia's RTX 50-series GPUs lending their power, though, with DLSS turned on and ray tracing set to high, Remedy's follow-up had me immersed. From the expert use of lighting in a chaotic world to the smooth, fast-paced combat, I already see Control Resonant being an all-new PC gaming benchmark.

You can check our Control Resonant preview for how the game plays and our interview with Remedy Entertainment's Creative Director for more on what is now one of my most highly anticipated games of the year. But for how it handles on PC, let's dive in.

Control Resonant PC requirements

While I played Control Resonant on a monster of a gaming PC (the RTX 5090 says it all), I was assured by the team at Remedy Entertainment that PCs that don't reach those heights, like many of the more affordable best gaming PCs, will be able to play the game without trouble.

Take a look at the system requirements for Control Resonant below, along with the PC specs I played on. Spoiler alert: you will need a lot of storage.

Brimming with power

First, a few preliminary points. As with most previews, this isn't the final version of Control Resonant. Not only is lighting and framerate still being worked on, but also audio mixing, hit impacts and more. This is still very much in beta, but honestly, from the time I had playing through the opening act and a mid-game area called The Sinkhole (it's a masterpiece), it's already looking to be in good shape.

I can imagine a lot of that has to do with the RTX 5090 working behind the scenes. With DLSS upscaling (no frame generation) and DLSS Ray Reconstruction (for sharper reflections and accurate lighting), zipping and levitating through a disfigured Manhattan as the all-powerful Dylan Faden (our up-close-and-personal protagonist) was a treat for the eyes.

And that was with graphics and ray tracing presets set to high, only made better by the 4K (3840 x 2160) resolution I was playing on. From the many twisted forms of The Hiss shown in terrifying detail to the deep contrast of red hues streaking across the screen, along with the subtle yet clear reflections in windows of buildings turned upside down, it's all beautifully crafted — and I often found myself taking it all in.

After leaving The Oldest House after the opening act, I was impressed by the visual distortion effect beaming off cars, with The Hiss floating in midair endlessly in the background. That, along with blaring, dangerous red lighting reflecting off the pavement. For those stationary moments, it all hits right.

In combat, with Remedy changing pace from a third-person shooter to pure melee action, things can get chaotic, but unsurprisingly, this rig could handle it all. As in, the copious amounts of rubble Dylan hurls at enemies while quickly dashing to slash through them using his Aberrant weapon (I chose the sword-like Flurry form as my weapon of choice), while the almost faded rainbow oil mist blasts off foes as he hits them.

It's smooth, clean action, which is what you'd want in a fast-moving action game. It's the little nuances, like the aftereffect of dashing being so crystal clear, like a sleek afterimage of static video, that make it work.

Now, I wasn't able to see frame rates or tinker with different settings, as it appeared the game was locked at a rock-solid 60 frames per second (FPS) for beta purposes. But even with the mass of Hiss shooting red beams from every direction during The Sinkhole section, along with the gravity-bending environment Dylan can shift through, it's clear Control Resonant will be taking full advantage of what Nvidia's GPUs have to offer.

And I'm here for it.

Reaching a new PC benchmark

While Remedy is still polishing off Control Resonant, especially the PC version, we can expect plenty of RTX features to be fully on show (just like path tracing in Resident Evil Requiem). That includes DLSS 4.5 and its dynamic multi-frame gen tech (expect frame rates to hit over 200 FPS), DLSS Super Resolution and path tracing for photorealistic lighting, no less.

That's exciting, considering just how well it already looks, and there's still plenty of time for the team at Remedy to cook. During my session, I had a chat with Lead Gameplay Designer Sergey Mohov, and he explained how there's now more to play around with.

"We weren't limited by hardware in terms of gameplay. It's very helpful that computers and consoles these days can actually run games like this, and we're able to put all of these enemies on screen, all of these effects and make it look nice while feeling very satisfying to kill them all.

"The hardest thing was actually making the world, making it bigger and increasing draw distances. But we've invested in improving our in-house engine (Northlight Engine) to handle these things. But, you know, we've done it."

I believe Mohov is spot on, taking the improvements we saw in Alan Wake 2 and pushing it even further. Of course, just like how Control is now available on iOS, there's the aim for Resonant to be flexible enough to be played on, well, anything that isn't just an RTX 5090 while still delivering great visuals and performance.

"We have a lot of people who are really excited about putting all the latest tech in. Nvidia is helping out, and we'll have all the latest Nvidia stuff in [Control Resonant]. So that's going to be pretty cool. But if somebody can run Control on, like, a potato or a fridge, I'll be pretty happy."

Control Resonant is set to launch on September 24, available on PS5, Xbox Series X/S, macOS and cloud streaming via Nvidia GeForce Now. But just from my brief experience, it will be a PC purchase for me.

More from Tom's Guide

• The 6 best video games we've played in 2026 (so far) on PS5, Switch 2, PC and Xbox
• I just finished Star Fox on Switch 2 — and it’s better than the game I imagined I was playing as a kid
• 5 recently announced video games I’m most excited about, including a sequel I’ve been waiting 13 long years for

OpenAI has unveiled Jalapeño, its first custom-built AI processor, developed with Broadcom specifically to run the large language models that power ChatGPT. Unlike the graphics processors (GPUs) that dominate today's AI infrastructure, Jalapeño was designed from the ground up for one job: answering user prompts as quickly and efficiently as possible.

This move is huge for the future of ChatGPT as the AI assistant could become faster, cheaper to operate and more reliable over the next few years.

Why OpenAI built its own chip

Nvidia remains the undisputed leader in AI hardware, supplying the GPUs that train and run many of the world's most advanced AI systems. But those chips were designed to tackle a wide variety of computing workloads, not exclusively the task of serving billions of chatbot responses every day.

Now Jalapeño takes a different approach. Instead of being a general-purpose processor, it's an inference chip, meaning its hardware is optimized specifically for generating answers after an AI model has already been trained. That's the part of the AI process you interact with every time you ask ChatGPT a question.

According to OpenAI and Broadcom as reported by Reuters, the new chip was architected around the memory, networking and computing patterns used by modern large language models, allowing it to perform those workloads more efficiently than conventional hardware. Early internal testing suggests it delivers better performance per watt than today's leading AI accelerators, although independent benchmarks have not yet been released.

Why this matters for ChatGPT users

If OpenAI can answer more questions while using less electricity and fewer expensive chips, the economics of running ChatGPT begin to change. That doesn't necessarily mean your ChatGPT Plus subscription will suddenly become cheaper, but lower infrastructure costs could allow OpenAI to respond to prompts faster, support more users during peak demand, reduce outages caused by hardware shortages and roll out more capable AI models without dramatically increasing operating costs.

In other words, beyond building a faster chip, OpenAI is making ChatGPT itself more scalable.

OpenAI is becoming more vertically integrated

Until now, OpenAI has largely relied on partners to provide the hardware powering its AI systems. By designing its own processors while continuing to build its own models and products, the company is gradually taking control of more of the entire AI stack.

That's a strategy we've seen succeed elsewhere in technology. Companies that design both their hardware and software can optimize them to work together, improving performance while reducing long-term costs.

OpenAI has described Jalapeño as the first step in a multi-generation compute platform, suggesting this won't be its last custom processor. To me, this looks more like an effort to reduce dependence on a single supplier while lowering one of OpenAI's biggest ongoing expenses.

Nvidia still dominates AI training, and OpenAI continues to rely on Nvidia hardware across much of its infrastructure. Jalapeño is designed for inference, not to replace every GPU in OpenAI's data centers.

The takeaway

Interestingly enough, AI actually helped design the chip. The company says Jalapeño went from concept to production-ready design in just nine months, which is an unusually fast timeline for advanced semiconductor development.

It's another reminder that AI isn't just writing code anymore, but increasingly, it's helping engineers build the hardware that future AI systems will run on.

Although most users will never see Jalapeño, (they'll never need to), but they will . notice the results every time ChatGPT responds a little faster, serves a few more people or unlocks capabilities that were previously too expensive to run.

More from Tom's Guide

• Anthropic will pay workers $85,000 to learn AI — and it reveals the next big AI job trend
• Anthropic's Fable 5 ban exposed AI's next big problem — but Sakana's Fugu may have the answer
• Is your job 'AI-resilient'? Find your risk score with our career calculator

Training and running AI models requires enormous amounts of computing power, and keeping those powerful chips cool often means consuming large amounts of water. It's become a growing source of concern as tech companies build massive AI data centers across the United States.

Now NVIDIA says it has a solution. The company recently unveiled a new liquid-cooling system designed for its next-generation AI infrastructure that it claims could reduce cooling water consumption to nearly zero in certain environments.

The announcement arrives as public scrutiny of AI's environmental footprint continues to grow, with communities questioning whether new data centers are placing too much strain on local resources.

So what exactly is NVIDIA proposing, and does it really solve AI's water problem?

Why AI data centers use so much water

AI systems run on specialized chips called GPUs, many of which are designed by NVIDIA. These chips generate enormous amounts of heat when processing AI workloads.

Historically, many data centers have relied on cooling towers that use water evaporation to remove heat. While effective, the process can consume millions of gallons of water annually.

As AI models become larger and more powerful, cooling demands have increased alongside them, turning water consumption into one of the most controversial aspects of the AI boom.

The issue has become especially sensitive in regions already facing water shortages or drought conditions.

NVIDIA's new approach

According to NVIDIA, its latest cooling system circulates warm liquid directly around AI hardware at temperatures reaching approximately 45°C (113°F).

Rather than relying on traditional cooling towers, the warmer liquid can be paired with dry coolers, which function more like large radiators and do not require significant water consumption.

NVIDIA says this could reduce facility cooling water use from roughly 2.6 million gallons per megawatt annually to nearly zero in favorable climates.

The company also argues that operating at higher liquid temperatures improves overall energy efficiency while creating opportunities to reuse waste heat elsewhere.

In other words, NVIDIA's goal is o redesign the entire thermal management process around water conservation.

The important caveat

The company's claims primarily address one specific category of water use: cooling. And while that matters, cooling isn't the only way AI consumes water.

Water is also used throughout the broader AI supply chain, including electricity generation, semiconductor manufacturing and data center construction.

Even if cooling water usage falls dramatically, AI systems still carry an environmental footprint beyond the walls of the data center itself.

Of course, there's also the question of scale. AI demand continues to grow at an extraordinary pace. Critics argue that efficiency gains can sometimes be offset by rapid expansion, meaning total resource consumption may continue rising even as individual facilities become more efficient.

Why this matters

The industry's largest companies are now under pressure to demonstrate that future AI growth can occur without placing unsustainable demands on water and energy resources. NVIDIA is not alone here — Microsoft has made similar zero-water cooling claims for its newest data centers.

NVIDIA isn't claiming that AI no longer has an environmental footprint. Instead, it's targeting one of the most visible concerns surrounding modern data centers: water-intensive cooling system

Whether NVIDIA's approach becomes the new standard remains to be seen, but it offers a glimpse of how the industry may attempt to address one of its most persistent criticisms.

More from Tom's Guide

• Anthropic will pay workers $85,000 to learn AI — and it reveals the next big AI job trend
• Anthropic's Fable 5 ban exposed AI's next big problem — but Sakana's Fugu may have the answer
• Is your job 'AI-resilient'? Find your risk score with our career calculator

Apple’s Mac lineup has gone from strength to strength since the rollout of its own Apple Silicon chips beginning with the M1, and while other manufacturers are catching up, it may have its most impressive rival yet.

Nvidia’s new RTX Spark leans on the company’s strengths in AI, gaming, and creative applications, and while it’s not out yet, it could supercharge the Windows laptop in a similar fashion to Apple’s own revolution in 2020.

Here’s all we know about each, and which could come out on top.

Release

The RTX Spark was revealed at Computex 2026, and we've already had some early hands-on time with it. The chip will roll out via manufacturers including Asus, Dell, MSI, and more in the coming months.

It’ll also come to the Microsoft Surface Laptop Ultra, with Acer and Gigabyte expected to follow. Sadly, because it’s not on sale yet, we don’t have an idea of what the final prices on these machines will be. Expect them to be high.

The M5 Pro, Apple’s latest and greatest laptop chip, can be found starting at $2199. That’s a steep cost, but it only gets steeper if you’re looking for more memory or storage.

Specs

What’s in a chip?

As you might expect, Nvidia’s new chip is built to take advantage of the company’s expertise across the likes of AI and graphical prowess.

To that end, the company hasn’t been shy about reaching out to partners like Microsoft and Adobe. RTX Spark, thanks to its development in partnership with Microsoft, will allow the first Windows PC models “purpose-built for personal agents”. OpenClaw users, rejoice!

Nvidia has also been working with Adobe to make its suite of professional apps, like Photoshop and Premiere, run even better on RTX Spark. The result (with the caveat of us not having tested it yet) is two times the AI and graphics performance.

It’s worth noting that Apple also trots out Adobe numbers when talking about its Mac chips, stating the M5 Pro and Max models offer a 35% increase in capabilities from the prior generation.

We’re expecting big things for GPU performance, too. The Blackwell architecture from the RTX 50-series of GPUs is found in the onboard GPU, and that means plenty of CUDA cores that could, at least in theory, mean features like DLSS 4.5 could come to RTX Spark.

Apple does boast improved graphical performance for the Mac lineup with each iteration, but it would be fair to say that its ecosystem doesn’t support as many titles. With that in mind, gamers will likely prefer the RTX Spark.

The CPU on Apple’s chip supports up to 64GB of RAM, but the RTX Spark can reach as high as 128GB. Expect to pay handsomely for it, but it might make it a no-brainer if you’re using your laptop for complex AI compute tasks.

Bottom line

Watching these two chips go head-to-head in the next generation of devices (MacBook Ultra, anyone?) is going to be fascinating, but it seems pretty clear-cut to me where the dividing lines will be.

Anyone who wants raw power for things like gaming or demanding creative work is going to be leaning towards the RTX Spark. Meanwhile, those looking for streamlined efficiency and battery life will still be suited to a MacBook. While Nvidia's chip will win on raw power, it will still run Windows inside a laptop developed by a third party. Apple's unification will be its remaining strength.

More from Tom's Guide

• Best of Computex 2026: The 17 greatest gadgets from the world’s largest computing show
• I gamed for 250 hours on AMD Radeon RX 9070 GRE — it’s a $549 nightmare for Nvidia’s RTX 5060 Ti
• I just tested the MSI Claw 8 EX AI+ — it leaves ROG Xbox Ally in the dust with breakthrough Intel Arc G3 power

I’ve tested Nvidia’s RTX Spark laptop chip (spoiler alert: it’s mindblowing), and now I’ve gone hands-on with every single laptop launching in just a few months time to set up a ranked list.

Eight premium notebooks are launching — each with their own unique features and quirks that I’m keen to share, so that when the launch date arrives, you know which one is best for you. Of course, this is personal preference, and certain systems I put in a lower grade may be better suited for you!

These are all incredible-feeling laptops, but grading them allowed me to get brutally honest with where I’d put them. So without further ado, let’s get into it.

While the Microsoft Surface Laptop was used in the demos, other models were turned off. This was something Nvidia required to ensure journalists (like me) don’t sneakily benchmark the laptops! That being said, these notebooks are basically pre-existing laptops but with RTX Spark inside and some tweaked cooling, so I can use past experience to fill in some gaps.

Nvidia RTX Spark laptops: Tier list

S-Tier: Microsoft Surface Laptop Ultra and Asus ProArt P14

I’ve already shared my experience with the Microsoft Surface Laptop Ultra — it’s a spectacular-feeling laptop in terms of a premium build quality, fantastic mini LED screen, tactile keyboard and a utilitarian aesthetic.

But the big showcase feature here (that no other Spark system has) is that breakthrough touchpad, which is able to communicate back to you with haptics across the UI. It’s a unique sensation like when snapping apps to the sides of the screen.

That being said, it is a 16-inch beast of a system, and the reason why I love my 14-inch MacBook Pro is the power paired with portability. Enter the Asus ProArt P14, and when I say Apple’s pro notebook is in trouble, this is the most direct competition to it in terms of premium build, fantastic ergonomics, port array and that RTX Spark chip.

Throw in an OLED display, and you’ve got a phenomenal ultraportable that I’d be willing to trade off a little bit of battery life to get — but hopefully Nvidia’s “all-day battery life” comes true to make this less of a fear.

A-Tier: MSI Prestige Flip N16 AI+, Asus ProArt P16 and HP Omnibook X 14

Kicking off with the Prestige Flip N16 AI+, I’ve already gone hands-on with the Intel version of this system back at CES 2026, and I’m a huge fan of it. This is the only 2-in-1 in the lineup, and that flippable hinge has a smooth glide and strong staying power at all angles.

On top of that, the touchscreen OLED panel on the pre-existing model is gorgeous; the keyboard is nicely tactile; and the touchpad is absolutely massive. Plus, there’s a stylus stowed in the bottom of the shell magnetically.

Then there’s the ProArt P16 from Asus, which is the same as that P14, but only bigger. If 16 inches is more than your speed, this is a fantastic option. But I’ll always opt for something more portable.

And finally in the A-tier is the HP Omnibook X 14, which immediately tickled my inner love for 14-inch laptops. Premium build quality, good I/O and a top notch OLED panel up top — so why the lower grade? Well, it comes down to that keyboard.

Whenever I’m typing on laptop boards, I’m more of a fan of a subtle dish that my fingers can fall into, whereas the keys on here are much more pronounced. It definitely has its fans, but I’m a little more mixed towards them.

B-Tier: HP Omnibook 16 Ultra, Lenovo Yoga Pro 9N and Dell XPS 16 Creator Edition

Now for the B-tier. Again, not any real slight, and this is my way-too-early impressions based on holding the systems. But personal taste made this an easy decision.

The HP Omnibook 16 Ultra felt like the heaviest laptop of the bunch, and while it has all those same winning specs as other options like an OLED display and a ton of I/O, that keyboard has those odd raised keys again and the edges can feel a bit sharp on the wrists when leaning on this to type.

Next, the Lenovo Yoga Pro 9N. From my time going hands-on with the Yoga Pro 9i (the same laptop shell), it’s a real prosumer piece of hardware with a nice aluminum build and those classic Lenovo ergonomics (a phenomenal keyboard for example).

But amongst a bunch of laptops that have some real unique elements that make these devices shine, it’s a rather unremarkable option. Not to say that’s a bad thing — I’m all for laptops that nail the fundamentals, but some just reinvent those fundamentals.

And then, there’s the Dell XPS 16 Creator Edition. In many ways, this is very much a Surface Laptop Ultra alternative, and another chapter in the book of Dell’s XPS apology tour for ditching the brand for a weird year back in 2024. The screen’s great, the touchpad is massive, and the I/O is plentiful.

My only main gripe is the keyboard — chiclet keys with no gaps between them. In my time testing XPS 16s of old, it took me a while to get used to a board like this, and I found myself often losing my place and making a couple of spelling errors.

Outlook

Of course, there's a lot more we need to find out first — not just actually testing these systems, but the big question of price. In a Q&A session, Nvidia CEO Jensen Huang did say this is the top-tier chip as part of a whole family.

That does mean we could get cheaper, lower-end models too. But looking at what we tested, we can get a somewhat decent guess of the price based on the fact its essentially the same silicon (and RAM) as the DGX Spark mini AI PC — a device that costs $5,000.

So if you want the top-tier model, this could get real pricey! But all I can say is I hope there are more reasonably priced RTX Spark laptops, which in the midst of RAMageddon may be tricky.

More from Tom's Guide

• Best of Computex 2026: The 17 greatest gadgets from the world’s largest computing show
• I gamed for 250 hours on AMD Radeon RX 9070 GRE — it’s a $549 nightmare for Nvidia’s RTX 5060 Ti
• I just tested the MSI Claw 8 EX AI+ — it leaves ROG Xbox Ally in the dust with breakthrough Intel Arc G3 power

Eight laptops are launching with Nvidia RTX Spark inside at Computex 2026, and the Microsoft Surface Laptop Ultra is easily one of the best — so beautifully designed it almost makes Spark feel like a side character.

Gunning for the 16-inch MacBook Pro, Microsoft doesn’t miss with a seriously luxurious build, great I/O for pros, fantastic ergonomics (including a breakthrough touchpad) and a mesmerizing display.

And you’ve seen what it will be able to do from all the RTX Spark testing I’ve done. This will not be a cheap laptop with that chip inside, so Microsoft has rolled out the red carpet in utilitarian aesthetics. Let me tell you about it.

Microsoft Surface Laptop Ultra: Specs

A seriously good looker (and worker)

Make no mistake about it — the Surface Laptop Ultra is built different. On the face of it, this is a Surface Laptop if it started bulking for a few months. But once you get into the details, you realize it’s so much more than that.

It starts with a beefy utilitarian presence atop any desk. This is a hefty machine, but those premium materials and the added bulk makes this feel ready to power through any task you throw at it — just like a MacBook Pro does.

The similarities also extend to the port selection, but chalk this up as a win for the Ultra with its additional USB-A for wider support. They keyboard feels gorgeous to type on, which is something that Microsoft has really excelled in in the past — giving you a nice tactility to every key press that makes it feel intentful.

But the real star of the show on that bottom deck is the 30% larger haptic touchpad. It’s an absolute ocean with a smooth gliding surface, with every click giving me those same snappy sensations I love from the MacBook Pro’s trackpad. But as an addition here is the haptics that actually communicate back to you.

Built directly into Windows 11, you’ll feel small hints of feedback as you do key things like snap apps to sides of the screen. On top of that, the company is working with key developers to give you those same sensations in third party apps like moving your cursor over clips in a video timeline.

It’s a feeling that when returning to my MacBook Pro, I felt like something was missing! Top it all off with that gorgeously color-accurate and smooth PixelSense display up top that can get seriously bright, and you’ve got a nice-looking beast that’s ready for anything.

Monstrous performance

Speaking of ready for anything, that’s where I get to the fun stuff inside. You’ve seen a lot of the software stuff it will be able to do from the Microsoft Build announcements — in short, expect a lot of agentic AI features to take advantage of that RTX Spark chip.

And as you saw in my testing of Spark itself, there’s a lot to be had here from the massive local models changing the dynamic of how you interact with a computer from it being a tool you lays idle when you’re not using it, to an active assistant machine.

But let’s compare apples and oranges and put it up against the MacBook Pro workload. You want your laptop to get the pro stuff done speedily both on and off the charger. After flying around the most dense Unreal Engine map while connected and disconnected, that’s easily proven true. There’s also tie-ins with Adobe to use much more of that built-in RTX 5070-level GPU and turboboost creator tasks beyond levels I’d seen in equivalent Apple laptops.

But the real killer app for me is the ability to use the same laptop to work by day and play by night. Alan Wake 2 is built natively for Arm and can run at 1600p resolution with ray tracing and DLSS 4.5 ray reconstruction at buttery smooth frame rates. Even games that are not built for this architecture like Pragmata go through the Prism emulator layer and play smoothly too. And yes, both of these were on and off the charger too!

The big unanswered question here is power efficiency, which could be affected by that emulation layer. Speaking to Nvidia, they’re not giving anything away, but did give me a cheeky smile when I started talking about what all-day battery life really means. We could be in for something special here.

Outlook

Ignoring the larger USB-C port mystery (my theory is it’s a magnetic breakaway USB-C socket that replaces the missing Surface Connect), what you’re left with here is a phenomenal laptop that feels powerful.

The aesthetics are clean, the ergonomics are fantastic, the display is mesmerizing and the power under the hood is awesome! Of course, there’s a big question on price, which rumors here at Computex point to a starting cost of $3,000.

This is definitely for the pros who know how to make the most of that power, and no longer is the MacBook Pro the default option at this price. A new player has come to town.

More from Tom's Guide

• I spoke to Nvidia CEO Jensen Huang about RTX Spark — he tells us why this is less of a laptop chip and more like 'R2-D2,' and shares future plans on N2X and N3X
• I gamed for 250 hours on AMD Radeon RX 9070 GRE — it’s a $549 nightmare for Nvidia’s RTX 5060 Ti
• I just tested the MSI Claw 8 EX AI+ — it leaves ROG Xbox Ally in the dust with breakthrough Intel Arc G3 power

So, to start my thoughts on testing Nvidia RTX Spark laptops, some context is needed.

My daily driver is the M5 MacBook Pro for its lightning fast performance and power efficiency to be just as fast on and off the charger, but I have to carry a Steam Deck for the gaming side of it. Enter RTX Spark, and that compromise is gone.

With Microsoft in tow, working in lockstep to rebuild Windows 11 for Nvidia's agentic vision for the future of computing, these are going to be some truly mighty notebooks that may very well revolutionize the Windows laptop and truly stand on their own as the future of computing.

But even if you’re not interested in the AI features, Nvidia has created the silicon that actually can kill the MacBook. No joke. Let me explain.

Nvidia RTX Spark: Adobe Premiere Pro boost

Nvidia worked closely with key creative app developers to ensure serious speed increases with that on-board GPU. Adobe Premiere Pro is 2x faster in key tasks that you can automate with AI, like cutting down massive clips. Meanwhile, in Unreal Engine 5, a massive cityscape of trillions of polygons can be loaded and moved around with ease.

And even better? All of this can be done both while plugged in unplugged. That crucial MacBook Pro advantage is no longer limited to Apple’s notebooks.

With these big specs (128GB RAM in this economy!?) these particular variants will be limited to only the absolute pro creators, but it’s not too far of a stretch of the imagination to see lower-end versions of these chips being more than enough for putting your projects together.

Nvidia RTX Spark: Gaming

This is the main weakness of any MacBook. Don’t get me wrong, they do offer game support, and apps like GameHub can navigate the compatibility issues and give you some form of PC gaming. But none of it compares to a DLSS-armed RTX Spark system.

Whether it’s Arm-native games like Alan Wake 2 taking full advantage of DLSS 4.5 ray reconstruction for some eye-catching visuals, or games that are being emulated like Pragmata, you’re getting a buttery smooth experience across the board. Nvidia wouldn’t show us FPS numbers, so based on eyeing it, I saw both these games in excess of 100 FPS on these laptop screens (which I assume are 120Hz).

The questions I don’t have answers to yet are how much that emulation harms performance (if at all), and whether battery life is impacted. But based on early impressions, these are laptops you can work on and play on.

Nvidia RTX Spark: Going agentic

Let’s get into Team Green’s mission to “reinvent the PC” to begin with. I got to try out some demos of some of its agentic features.

First, there was a mutli-device OpenClaw situation of talking to your computer. This was done on a DGX Spark mini PC, but given the silicon is practically identical to RTX Spark (just that the latter is optimized for Windows), you can do this too. And the idea of your computer being something that you can work with from anywhere no matter what device came into view.

From complex things like coding a website landing page to simpler things like translating a menu and picking the right dish with low salt if you have high blood pressure, the long memory context that can be saved and loaded into the massive memory on tap here is significant.

Second, there was the multi-app pipeline of getting something done. I got to recreate the demo from the keynote where a griffin was AI generated in Photoshop based on a sketch, turned into a 3D model and animated via ComfyUI (the visual way of connecting AI models together).

The speed was seriously impressive, though you did see the limitations of the current local AI models, where the legs would disappear in what I assume was some hallucination madness.

Then came the developer side of things. I won’t go into the vibecoding bit of it (I don’t really agree this is the way to create sites and apps), but what I will talk about is the system taking agentic control of the website. The last time I tested something agentic on a web browser was Opera’s Browser Operator, which would take a screenshot and then send it to an LLM in the cloud to devise the next move.

All happening securely on the system, the speed difference is night and day — tearing through a site and filling in a form in mere seconds. This is a glimpse of what telling your laptop to just do something can be.

And yes, the bird creation turned into video definitely screams AI slop, but the principle is there. You can imagine the natural language prompts with an image to tweak the colors and contrast, or turning your pencilled ideas into something that can inspire a real-life idea.

Also, shout-out to the tweaks being made to Windows 11 too! Attaching the taskbar to the side of the monitor is a revelation to getting it out of the way and focusing on your work.

The Nvidia RTX Spark laptops are seriously good

Then there’s the laptops themselves. I won’t spend too much time talking about them here, as I want to rank them in a future piece (spoiler alert: that Asus ProArt P14 is calling to me). But to match that expected high price, there are some seriously sleek hardware options coming.

One thing you’ll notice is that these are very familiar shells with maybe a tweak for cooling here and there. This lines up well with what Nvidia CEO Jensen Huang told me about wanting users to “go on the journey at your own pace” when it comes to the agentic side of things. So to completely flip the laptop build would probably terrify some folks.

For the 2-in-1 crowd, the MSI Prestige N16 Flip AI+ is a seriously nice system, Microsoft’s Surface Laptop Ultra is a luxurious device, HP’s Omnibook 14X has an amazing keyboard, and the Dell XPS 16 Creator Edition gives you a nice upmarket feel. These partners have truly pulled out all the stops to provide some primo shells for Nvidia’s silicon.

RTX Spark outlook

I was speaking with our video producer extraordinaire Paul Antill in an uber around Taipei, and he asked about whether RTX Spark could beat the MacBook. To me, that question is always a bit of a poison pill, but one that is fair — every tech news outlet is guilty of saying “this laptop is good, but it’s not quite as good as the MacBook.”

But now, finally, after years and years of waiting for an Apple killer, Nvidia has stepped up and done the damned thing. AI slop laptop potential aside (lapslop?), this is a phenomenal chip that delivers big on making an extremely powerful machine both on and off the charger with serious power efficiency, while also being a peak gaming machine, too.

More from Tom's Guide

• I spoke to Nvidia CEO Jensen Huang about RTX Spark — he is ‘willing to work’ on an RTX gaming handheld, N2X and N3X are already planned and the chip is 'more like R2D2’ than a laptop CPU
• I gamed for 250 hours on AMD Radeon RX 9070 GRE — it’s a $549 nightmare for Nvidia’s RTX 5060 Ti
• I just tested the MSI Claw 8 EX AI+ — it leaves ROG Xbox Ally in the dust with breakthrough Intel Arc G3 power

So in case you’ve been living under a rock, Nvidia just changed the face of consumer computing with RTX Spark — a new all-in-one chip that Team Green aims to “reinvent the PC” with. It’s a bold mission and I wanted to know more, so I talked to CEO Jensen Huang about it in a Q&A at Computex 2026.

What does reinventing the PC actually look like? What does the future of this silicon look like? How does Huang fancy his chances against Apple silicon? And could we see Nvidia take this, pair it with its gaming handheld expertise with Nintendo Switch chipsets, and make a PC handheld? It was a long conversation, so grab a cuppa and find out what he had to say!

Jensen wants to turn your PC into R2-D2

I think the most fascinating part of this talk came down to the grand vision of where Huang and Nvidia see the world of computing going. You’ve already heard about the want to reinvent the PC with agents, but he goes into more detail here.

“I believe that today, the computer sits there waiting for us to use it. In the future, when we leave it, we’ll be talking with it all the time,” Huang expanded. “I’ll be chatting with my agent on WhatsApp. It’ll be talking back to me. It’ll even call me!

“That is the personal computer of the future. Tell me that’s not R2-D2. Tell me that’s not robotics. Tell me that’s not cool.”

And looking back to Nvidia’s keen interest and investment in OpenClaw and innovating upon it for themselves suggests just how this vision comes together. Agentic AI via the cloud is slow, whereas with the right-trained model built locally into Windows, this could be a breeze. That is what RTX Spark is capable of.

That being said, Huang did acknowledge that the transition will be tricky, when asked the question about whether this should even be called a PC at all. Nvidia and Microsoft want to keep the foundational experience familiar, and make RTX Spark systems “100% awesome at everything you expect the PC to do,” so you can “go on that journey at your own pace” with the agentic side of things.

The roadmap has already been set

To clear up any confusion, Huang was quick to confirm that N1X is RTX Spark, saying, “N1X was the project code name.”

But he went a little further into the family plans for the architecture, and its future too! “N2X and N3X are already planned, and N1X is called N1X because it has a smaller version called N1. We’re going to expand our family. We’re going to extend this architecture for a very long time.

And pointing back to Nvidia’s historical track record of taking care of its software over a long term, users can expect to keep their RTX Spark systems in their homes for 5-10 years. “Just like my home theater system,” Huang compared in probably one of the more humorous comparisons to home computing I’ve heard.

But this puts it right in line with what kind of longevity you can expect from Apple Silicon products. Speaking of…

RTX Spark vs Apple Silicon

In group Q&As like this, it’s best to make friends with the person next to you to double up on what you want to know. Shout-out to Arsh Garwal for stating the fact that RTX Spark is “basically entering Apple Silicon’s home turf,” and asking Huang what winning looks like for Nvidia in comparison.

“Apple’s ecosystem is excellent as you know, and they have a world-class silicon roadmap. But that’s not our focus,” Huang responded. “Our focus is to reinvent the PC. Windows is of course improving, but the basic architecture of a PC has largely been the same now for about 40 years, and we want to reinvent it.”

Of course, this is a subtle dodge of the question here, as Huang talks about how Apple has the ecosystem advantage (which it really does). But it goes to show the focused mission going on here.

Can Nvidia have its own Nintendo Switch now?

I got the chance to ask Huang a question about gaming — specifically handheld gaming. Besides the goal to “reinvent the PC” with Microsoft, this is an Arm chip that has the same graphical capabilities of an RTX 5070 notebook card, while being much more power-efficient.

Nvidia claims you can get 100 fps in Indiana Jones at 1440p, and with it being the full Team Green silicon, you’ve got access to all that DLSS 4.5 goodness with dynamic multi-frame generation. On top of that, Nvidia’s silicon is inside arguably the best-selling gaming handheld in the Nintendo Switch and Switch 2.

And with Intel Arc G3 taking PC handhelds to the next generation of performance with its beasty integrated GPU and XeSS 3 AI trickery, it got me thinking. So I asked him whether he envisions a future where some form of this RTX Spark silicon comes to a gaming handheld.

“If somebody wants to do it, we'll work with them on it,” Huang responded. “But right now, we're really focused on doing something that is just such a big deal — reinventing the PC after 40 years.” Understandable given the lofty goals, but he goes on. I’ll just put the transcript here and analyze afterwards, as he touches on a lot of details.

“This is such a gigantic project. We worked with hundreds of people that worked inside the company for years now. Remember all of these applications in the Windows world and x86 world, we have to prove that they work fantastically. And all the games, they have to have anti-cheat turned on. All of that stuff is really really hard. There's a reason why today's PCs are so fragmented — that things that are great for Adobe aren't great for gaming, or things that are great for gaming aren't great for AI. The world has a very hard time uniting that into one platform, and we felt that we had the ability to do that from a technology perspective, from a software architecture perspective, and an ecosystem reach. And over the course of the last couple years between us and MediaTek, all the PC OEMs and Microsoft, you're going to hear more about it later today [at Computex]. I mean, just the mountain of people just so that we could reshape this computer to get it ready for Agentic AI.”

Now, there’s a whole lot of talk about the bigger mission here, and it’s clear that while Nvidia is willing to work on handheld gaming, that’s not the key focus here. But he does hint at some clear hurdles here, such as making sure this is an all-in-one platform that works great across everything.

Specifically calling out anti-cheat compatibility is an interesting side point. When speaking with Team Green separate of this Q&A, the team did say that native Arm anti-cheat is “one of the biggest challenges,” and that they are “working with developers to bring support for major anti-cheats like Easy Anti‑Cheat, BattlEye, and Denuvo among others.”

And since it’s my birthday today, he decided to do something nice for me, as you can see!

Why now?

It’s a fair question. We heard about Nvidia entering the laptop silicon space for years and with the can getting kicked down the road every time, I honestly thought it would just subside. Plus, if we’re being real, the vast majority of Nvidia’s money comes from its data center business. So why now?

“We don’t really have to choose. The real question is can we make a contribution? If it’s only marginal, we won’t do it,” Huang responded. “If you get the chance to reinvent the single most important tool for humanity, and reinvent it in the age of AI, we’re not going to sit around and not get it done”

With physical AI only being “only a couple of years away,” he believes having a laptop primed for this will be critical to the future of computing as a whole. It’s a big bet, and the next few years will show whether it pays off.

More from Tom's Guide

• I gamed for 250 hours on AMD Radeon RX 9070 GRE — it’s a $549 nightmare for Nvidia’s RTX 5060 Ti
• I just tested the MSI Claw 8 EX AI+ — it leaves ROG Xbox Ally in the dust with breakthrough Intel Arc G3 power
• I just held the most beautiful laptop, but I’m not sure Vincent van Gogh would be happy with his name connected to AI

Nvidia has stormed Computex 2026 with the official announcement we all knew was coming: the company's first all-in-one laptop silicon called the RTX Spark Super Chip family.

Analysts are calling this Nvidia's most disruptive move in a decade, and my colleague Jason England likens it to the moment Apple unveiled its own M1 chip back in 2020. By moving beyond GPUs into integrated processors, Nvidia is throwing the gauntlet down to Qualcomm, Intel and AMD.

Nvidia CEO Jensen Huang revealed RTX Spark during his keynote in Taipei, and the first laptops powered by the new chip should arrive in the fall. Beyond that, we can expect to see the RTX Spark appearing in mini PCs and small form-factor desktops from the likes of Dell, HP, Lenovo, Asus, MSI, Acer and Gigabyte.

Huang noted that Nvidia and Microsoft have been working away on this for the last three years to "reinvent the PC for the first time in 40 years," aiming to move towards a responsive, proactive AI experience. Microsoft itself is coming out of the gate with an RTX Spark-powered Surface Laptop Ultra that could be a MacBook Pro killer.

Here's the list of the first eight RTX Spark-equipped laptops, along with a few details we know about them.

Microsoft Surface Ultra

Microsoft hasn't confirmed many details about its upcoming RTX Spark-powered Surface Laptop Ultra, but we know it'll have a 15-inch mini-LED touchscreen with 2,000 nits of peak brightness.

It will benefit from a large haptic touchpad and all the required ports: HDMI, USB-C, USB-A, SD card and a headphone jack.

Microsoft claims this will be “the most powerful Surface Laptop ever built” when it arrives later this year. Microsoft also claims it can run AI models locally and that the chip delivers power comparable to a mobile RTX 5070 GPU.

Asus ProArt

Asus is planning to launch two RTX Spark-powered ProArt laptops later this year: the ProArt P16 and the ProArt P14. Both laptops will feature 120Hz OLED touch panels in the 16:10 format (one will be 16-inch, one 14-inch) and be geared towards creative professionals.

Memory will be up to 128GB LPDDR5X 9400, while storage can be configured up to 2TB on the P16 and 1TB on the P14. Regardless of which model you choose, you'll get a full suite of ports: 3x USB-C, 1x USB-A, 1x HDMI 2.1, 1x headphone jack and 1x standard SD.

MSI Prestige N16

MSI is set to launch the 2-in-1 Prestige N16 Flip AI Plus, which, it says, will represent "a new generation of premium thin-and-light PCs.”

Sadly, we've got no further information on the specs, dimensions, or — crucially — the price. All we can say is that this will be a 16-inch laptop with a 3,840 x 2,160 OLED panel and a peak brightness of over 1,000 nits.

Dell XPS 16

Dell will launch an RTX Spark-equipped version of the XPS 16, which likely won't deviate much from the current model in terms of outward design. We can expect an aluminum 16.3-inch workhorse with three USB-C ports, an HDMI port, an SD card slot, and a headphone jack.

Price and specifications haven't been revealed yet.

HP OmniBook and Ultra 16

HP has confirmed two forthcoming RTX Spark laptops, the OmniBook X 14 and Ultra 16, but stopped short of announcing any specific details. The company says it will create "the world’s thinnest RTX Spark, built for powerful performance" and has confirmed it will follow up on the laptops with a compact desktop in the future.

“Our expanded portfolio pairs compact, powerful hardware with pre-configured environments and open-source toolchains to eliminate setup friction and accelerate the path from idea to execution," said Samuel Chang, senior vice president and division president, Consumer Personal Systems at HP.

Lenovo Yoga Pro 9n

Lenovo's flagship 16-inch laptop will also be getting the RTX Spark treatment, but we haven't had any details about exactly what changes to expect. Like the other laptops on this list, Lenovo's creator-focused 16-inch machine boasts a durable aluminum chassis and configurable specs tailored to your needs.

One area ripe for improvement would be battery life, and we'd hope that the RTX Spark-equipped Yoga Pro 9n can surpass the 9.45-hour battery life we recorded on the Intel-powered Lenovo Yoga Pro 9i.

Bottom line

In addition to announcing that over 30 laptops will come with the RTX Spark chip, Nvidia also revealed that Adobe is rebuilding Photoshop and Premiere Pro from the ground up for compatibility.

Suffice it to say, the above laptops will kick off a silicon war in the fall that could be good news for consumers. Especially for gamers and creative video editors. The Spark's Blackwell-architecture GPU, with 6,144 CUDA cores, brings Windows-on-Arm compatibility with Nvidia's gaming suite, including DLSS 4.5. You're going to get thin, light laptops capable of running triple-A games at 100 FPS with real-time ray tracing.

Finally, these laptops will be positioned to move Windows into a truly agentic AI future, with RTX Spark delivering 1 Petaflop of local AI performance. In a nutshell, it'll have the power and efficiency to work on the user's behalf across various apps to achieve the desired outcome.

So we're potentially looking at laptops with a thin-and-light form factor, elite battery life, and the power to crush desktop-class AI development, heavy creative production, or cutting-edge gaming. The only question is: how much will these devices cost when they arrive?

More from Tom's Guide

• I just tested the MSI Claw 8 EX AI+ — Intel Arc G3 is a breakthrough for handheld gaming, but at what cost?
• Intel Arc G3 is here to shake up handheld gaming, and AMD should be terrified — Acer Predator Atlas 8, a new MSI Claw and more launching soon
• I built a touchscreen MacBook Pro using this snap-on accessory, and I'm baffled why Apple hasn't made its own for years

The biggest news at Computex is official, as Nvidia’s CEO Jensen Huang just announced RTX Spark Super Chip family — the company's first-ever all-in-one laptop silicon that takes on Intel, AMD, and Qualcomm in a big way with a GPU equivalent to an RTX 5070.

Set to release this fall, this is not your ordinary ARM chip. The RTX Spark is packed with CUDA cores, which is sure to make that integrated GPU one helluva gaming monster. Plus, Spark is set to change the way you interact with Windows PCs forever with an AI agentic twist. I will be testing the chip extensively during my time out here at Taipei, and talking to a couple of the engineers behind it.

But for now, I’ve taken a deep dive into the specs tables and found what could be a breakthrough moment similar to when Apple brought the M1 to the world back in 2020.

Nvidia RTX Spark Super Chip specs

Nvidia RTX Super Chip devices

Nvidia is expecting a huge lineup of systems to launch with RTX spark — with plans to release over 30 laptop models and well over 10 desktop variations, targeting a range of price points.

But as of now, there are a whole lot of devices launching this fall with the RTX Spark Super Chip inside. On the laptop side, we have:

• Asus ProArt P14 and P16
• Dell SPX 16
• HP OmniBook
• Lenovo Yoga Pro 9
• Microsoft Surface Ultra
• MSI Prestige N16

These devices will be as thin as 14 millimeters, and as light as three pounds — a serious upgrade over the chunky Nvidia gaming laptops you're used to.

But they're not stopping at laptops. In an announcement that probably woke up Anthony Spadafora (our mini PCs king) from his nap, RTX Spark will be coming to small form-factor desktops that are "small enough to disappear on a desk."

These will be made by Dell, HP, Lenovo, Asus, MSI, Acer, and Gigabyte.

Explain it with pizza

This is partly inspired by the tasty ‘za I found while typing this, but also shoutout to Nvidia for specifically pointing out that I just “love to explain stuff with pizza.” You asked for it, and now you’re getting it!

In the laptop space, you’ve got various ways of serving pizza.

• The standard meal (an x86 SoC): This is like ordering a plain cheese pizza. It works perfectly fine for everyday eating, but if you want heavy flavor (think intense gaming or AI generation), it lacks all the necessary toppings. Don’t get me wrong, Intel and AMD have been adding some new garnish, but they’re still a little hampered by architectural limitations.
• The combo meal (x86 CPU + dedicated GPU): For that premium flavor, you have to order the standard pizza plus a massive side of premium toppings. Passing these ingredients back and forth across the table does the job, but it takes up way more space (think thicker gaming laptops) and requires a whole lot of power.
• The efficiency slice (Snapdragon X Elite): Qualcomm is already dealing in making efficient single-crust pizza using a similar recipe (Arm architecture). It’s great for saving power, but Nvidia's world-famous secret sauces (CUDA and RTX) can’t be matched.

So in this backdrop, I can only describe the RTX Spark Super chip as a “super pizza.” Instead of ordering a meal with lots of separate sides, this is the all-in-one slice baked onto a single, highly efficient crust. Half of the pizza is that custom CPU, and the other half is loaded with incredibly spicy, high-end toppings in that Blackwell GPU.

But the magic is in the cheese (the memory). Instead of dividing the cheese between the crust and toppings, Spark uses up to 128GB of “unified cheese” that the entire pizza shares instantly at blazing speeds (five times faster than standard PCIe Gen 5). And because it’s baked so efficiently, you can take this super pizza anywhere without it getting cold (all-day battery life).

…am I stretching this analogy? Absolutely. Do I regret it? No (hi Nvidia if you’re reading this).

Going agentic

With this increase in RTX Spark's brute force, Nvidia is working with Microsoft to shift personal computing on Windows from a passive tool to a proactive agent.

This is a one petaflop AI superchip... For context, some quick math I did while sitting in the keynote, the AMD Ryzen AI Max+ 395 can do around 60 Teraflops. This thing is built differently. To go back to the pizza analogy, not only can it make a mean slice, but it can probably feed it to you, too.

It can work across apps on your device — stringing together different tools like any user would to get the job done. For example, an artist can provide a sketch and a mood board, then use Photoshop not only to generate the art locally but also to turn it into a 3D image and animate it into a video. And this is all done without ever needing cloud compute.

Or if you’re a gamer, you can use G-assist to tweak critical settings like monitor optimization, or (if you’re streaming) tell it to operate your lights, mute the mic, and turn on your be right back scene in OBS.

Essentially, it can take complete control of your system — even the mouse and keyboard to navigate as a human user would. Keeping this all secure is Nvidia OpenShell, which brings enhanced security to ensure your model use works locally and privately. It can even mask personal information before ever using a cloud-based AI model to do anything.

For the players

But of course, there are big gaming capabilities here too. This is Nvidia, after all. Team Green says this falls into the same performance class as an RTX 5070 laptop GPU but uses significantly less power to reach that level.

That means you could be playing the likes of Cyberpunk, Doom, or Indiana Jones at 100 FPS at 1440p resolution. And given its ARM, the RTX Super Chip is capable of this level of performance both when plugged and unplugged.

Of course, battery life will vary based on how much you throw at it, but when I asked Nvidia, a tweak of maximum frame rate and graphics settings can optimize this nicely for long gameplay sessions on flights.

Now, what about the emulation part of it? General apps are included in this, too, but gaming has been the big test for Windows on Arm, and that Prism emulation layer used to translate apps written for other CPUs to this new architecture. Well, Nvidia’s thought of that too.

The company is working with a large list of developers to either port games natively to Windows on Arm or ensure they run well through the emulator. And with that hefty GPU, anything that’s particularly taxing on that will have almost no negative effect on performance.

Plus, Nvidia has collaborated with all first- and third-party anti-cheat software providers to ensure competitive multiplayer games launch without a hitch.

Nvidia RTX Spark vs Apple M silicon vs Snapdragon X2 Elite

So, since I’m calling this Nvidia’s M1 moment, how does it stack up with Apple laptop silicon? While I’m at it, I'm gonna throw Snapdragon into this comparison too, as they’re all similar-ish in Arm architecture.

But the crucial differences show that it’s not about the tools you have, but rather how you use them. You see, Apple silicon isn’t just about the brute force of the hardware; it’s also about its tight integration with software that extracts that performance. In gaming, for example, the MetalFX layer has enabled MacBook Pros to do some pretty great things, like run Cyberpunk 2077.

And while Snapdragon X2 Elite is level pegging in certain benchmarks, there are still some moments where it can feel like a bit of a blunt instrument against Windows 11 — especially in that GPU department.

But with RTX Spark, this is Nvidia we’re talking about. That integrated GPU supports DLSS 4.5, so you can bet your bottom dollar that the out-of-the-box gaming performance is going to be clearly superior to Apple and Snapdragon.

On top of that, Team Green and Windows are rewriting the rules on local agentic AI, so you’re getting a heap of AI coworkers too, with those CUDA cores for the raw brunt.

Nvidia RTX Spark vs Intel Core Ultra Series 3 vs AMD Ryzen AI 400 series

Why did I put these separate from the face-offs with Snapdragon or M5? Well, it comes down to the different ways these chips compute stuff. Nvidia, Apple and Qualcomm all have Arm architectures, which do things very differently to the x86 chips from Intel and AMD.

That being said, the 30+-year-old x86 architecture does bring some material benefits over ARM — namely, zero app compatibility issues whatsoever. It’s a challenge that Microsoft and Nvidia are overcoming with a separate Arm-specific version of Windows 11, alongside working closely with developers. That could narrow the library you can play!

However, for what it may lack in this area, it absolutely makes up for in a sweeter blend of performance and power efficiency. While x86 is all about multitasking its way to doing things you ask it to (not the best on that battery), ARM tackles one task at a time very quickly. That’s easier on the power while bringing high speeds to the table, too.

Outlook

It’s been a long road since the first time we heard rumblings about this chip. In fact, I even put this in my Computex 2025 predictions! But now it’s here, I think it’s fair to say that RTX Spark is far beyond any expectations I had.

Not only is it a beastly integrated GPU for monstrous performance and impressive power efficiency (especially in gaming), but the software tie-ins fundamentally change how a Windows laptop works with you. This is going to be an interesting few months as we count down to the fall launch.

Now the wait is on for an Nvidia RTX Spark gaming handheld…that would slap so hard!

More from Tom's Guide

• I just tested the MSI Claw 8 EX AI+ — Intel Arc G3 is a breakthrough for handheld gaming, but at what cost?
• Intel Arc G3 is here to shake up handheld gaming, and AMD should be terrified — Acer Predator Atlas 8, a new MSI Claw and more launching soon
• I built a touchscreen MacBook Pro using this snap-on accessory, and I'm baffled why Apple hasn't made its own for years

The Microsoft Surface Laptop Ultra is the company’s answer to the MacBook Pro. Announced during Nvidia’s Computex 2026 presentation, this laptop is among the first to feature the newly announced Nvidia RTX Spark chip. It promises to be powerful enough to handle creative workloads, development, local AI use, and gaming.

Besides the Laptop Studio, Surface devices are generally designed for average use. That won't be the case with the Surface Laptop Ultra, as it will pack up to 128GB of unified memory. Microsoft also claims it can run AI models locally and that the chip delivers power comparable to a mobile RTX 5070 GPU.

If the Surface Laptop Ultra is everything Microsoft claims, it will easily be the most powerful Surface device yet — and could give the likes of the MacBook Pro stiff competition. Here’s everything you need to know about the Microsoft Surface Laptop Ultra.

Microsoft Surface Laptop Ultra: Design

The Microsoft Surface Laptop Ultra will include some of the features you’d expect from a premium notebook. For starters, it has a 15-inch mini-LED PixelSense Ultra display that can reportedly reach 2,000 nits of peak HDR brightness, making it the brightest display of any Surface device ever. Microsoft says it will have strong color accuracy for creative work.

You’ll also get a full set of ports, including HDMI, USB-C, USB-A, an SD card reader, and a headphone jack. The haptic touchpad is the largest Microsoft has ever used on a Surface, and the laptop is designed to run quietly even under load.

The Surface Laptop Ultra will reportedly have “all-day” battery life, which is a claim I’m eager to verify in our lab tests. Expect it to come in Platinum and Nightfall finishes.

Microsoft Surface Laptop Ultra: Nvidia RTX Spark

The Microsoft Surface Laptop Ultra announcement alone is noteworthy, but it’s even more interesting because it’s one of the first laptops to feature the Nvidia RTX Spark chip. In case you were wondering, yes, it’s the processor long known as the “N1X.” Different moniker aside, the chip should be a major boon for the Surface Laptop Ultra.