Nahalem opinions

Well, an integrated memory controller should improve memory performance, but as we already see that doesn't help AMD processors much in games.

I think it will be good. Intel is biting off a lot here, but they'll pull it off. I wouldn't expect a P4 => Core 2 jump, but it should be decent.

As for prices, I could see them being high, especially early on with the first releases.

Actually, the IMC helps gaming a lot with K8, compare it to K7 which was the last FSB based AMD CPU, gaming performance is much improved.

Core 2 performs so well in games due to its massive cache, as well as having higher IPC. Would it perform even better with an IMC? You bet.

Intel is actually touting Nehalem as a bigger leap over Core 2, as Core 2 was over P4. This is under multithreaded performance, there are claims performance is 40 - 100% better, which would be incredible if even half true.

No more FSB, that seems it could provide a lot better bandwidth options.

at " accient FSB architecture of intel is somehow the only hold back in the memory subsystem.but in game and many application the memory bandwidth is already good enough.but i wonder if that will effect the overclocking of the CPU as everything is integrated into 1 chip as heat would be a problem.

also another thing i worry about is that would it be mature enough at the 1st generation.but the ist gen or Core 2 is very successful!hope that will continue.i wonder what the AMD Shanghai core is like!

how the overclocking would be done?would it be like the AMD's?different multiplier for CORE,"HT LINK",MEMORY many interesting thing to find out!

Bring back socket 7!

if Hyper threading is successful this time round and games and OS can take full advantage of it.then games could have much more physics in them as the CPU is able to process more data.but more then 2 core is still not much of a advantage at this time.

hope things will change sometime soon.

The single threaded performance will be the 100% gains and the multi threaded should be about 40%. But remember that this is not just a Core2 with a IMC slapped on it. Its a whole new architecture. The reason single threaded apps may see a big boost is b/c of the dynamic OC'ing Nehalem is supposed to have.

That means that when a single thread is being run other cores will under clock and the core processing the thread will overclock to attain maximum single threaded performance. So games could see a bigger boost than you think.

I would only hope this would go for the dual core optimized apps and games. Imagine having 2 cores at 2GHz on a 3GHz CPU and the other 2 at 4GHz just for that game like Crysis.

Personally I think it will turn the tables on AMD. AMD will no longer have the Nitive Quad or IMC bit. It will be whatever performs better on an IPC and price level.

And that likely has AMD bleeping their pants.

AMD strong memory lead really helps in the Server arena.
Historically we go XEON for most general purpose servers, but our large VM boxes are all AMD due to their memory performance. If Intel significantly closes that GAP, AMD will be really hurting.

They got tagged last quarter in that area, but I think will do much better this one now that they are finally shipping the B3 Server chips. But things will start to look real tough again late this year.

Nehalem is going to be a sad disappointment, I'm afraid, because of advertising people and fanboyism.

It'll work beautifully, it'll blow everything else out of the water...

...and because the market will have believed the lies, the PR and the hype, it'll be 'crap' because 'it was supposed to be more (insert adjective here) but it isn't!'

Seriously, though, Intel getting off the shared-bus architecture is a good thing. That has really been holding them back.

8 cores? Rofl , seems to me that Intel is assuming the general public are morons , lol."Slap more cores on it and charge 3 times the amount for it and we'll be rich!"But seriously as we all know the software for multi core proc's are not even close to catching up.This 8 core monster would be a great server chip , but in all reality it'll offer minor performance gains in gaming and general use.They're jus pulling what they did in the P4 days , Hype the Ghz to sell chips.


IMC is a "about damn time" idea , i'm gonna err on the negative side until they release the chip and we get numbers , but for now it's looking like a server chip hyped as the next uber goober dodad that we all have to buy to be l33t.

Indeed.

There has been talk of them releasing an MCM Shanghai - called Montreal. It uses a hypertransport link to communicate between cores.

*see here => http://download.amd.com/Corporate/ [...] ystDay.pdf





More interestingly... Propus. Now a question - how much of the die is taken up by... say... 6MB of L3 cache?

Would it be possible to put 4 Propus chips on the one die? That would give you a full 16 core CPU... on the one socket.

The question of course would be data exchange using only 4x Hypertransport3 links - you are limited to a single socket server unless your willing to take the performance hit on double hops for information between cores.

Really, you'd need 8x links for a dual socket workstation, 16x for a quad socket (but thats 64 cores!!!)

Why do you assume Intel are aiming their 8 threaded CPU at the desktop market?

Amiga , ok you got me , i am assuming they will market it for the elite gamer group , because they are still trying to sell the skulltrail platform for gaming.My logic is , if folks are still buying Macs because of slick marketing and a elite "I got bigger Ghz then you" attitude , why would Intel not try to cash in on the next big chip?

As has been said before, understand the clientele before you make such sweeping comments....
I have two work machines that are QX6850 based, so thanks for calling me a moron (I chose the spec for them)
OCing is not an option for a work machine as I can't afford for any problems to them, hence no Q6600 with a 450MHz FSB, just a stock QX6850!

Once again, don't be so naive. Nvidia aren't playing dead! Understand economics before you make such a ridiculous comment! Nvidia aren't waiting for ATi or Intel, their shareholders would go ballistic if they were preventing them (the shareholders) from making money. Nvidia are falling in share prices (as is the rest of the tech market, in general) because the US is currently heading for recession. As TGGA said on other thread (where he agreed with me), If Nvidia could sell a $600 card, they would be. Profits are only going to fall this year (I think, but it's not an uncommon opinion) as people are going to be stung by mortgage payments/car finance and any other outstanding finance.

Pretty much the same comments to you as I said to the other poster. Intel aren't holding back 45nm quads, they're having problems somewhere even if tehy don't admit it. Lets look at it subjectively, 45nm ahs been signed off for a while and production is in process, so the millions of $ (if not a billion or more) that have been spent on development for the 45nm CPU's is currently going to waste while they wait for AMD to release faster Phenoms...??
Not in a year of sundays! Intel need to make money off those CPU's to recover the vast investment for their development. No way are they going to sit on a fully-developed product (while it is becoming an obsolete asset) to let their competitors catch up... Basic economics!
Anyhow, onto the OP's question. Personally I think Intel will (and are) struggling with implementation of the all new IMC, native quad & octo and HT Link on a new architecture. It's a big thing to do on a new process and there is a lot that could go wrong. Anyone heard of the AMD Phenom, what problems did that go through, considering AMD already had HT & IMC nailed from K8, they stumbled into a lot of problems with native quad. I am in no way saying that Intel are going to fail, more that I think they're claims over performance are a little premature. If they can implement all those new features on a new process then all the credit to them, I can't help but think they've bitten off a lot in one go with Nehalem!

It is my opinion that Nehalem does not exist yet. Therefore, it is too early to form much of an opinion. The Phenom looked GREAT on paper, it looked like a real Core2-killer, but we all know the results of the end product.

Um Nehalem is not a "Paper Tiger". Does this look it doesn't exist yet?

http://youtube.com/watch?v=LP5Pp-Z [...] re=related

http://www.intel.com/technology/ar [...] iid=search

Its real baby and is working. So don't doubt it without proff.

I cant believe Ive read thru this whole thing and havnt seen the obvious question. Like many an Intel fan has rubbed it in the face of AMD fans, you have to buy the more expensive ram, better timings with your IMC based cpu. Wont that also hold true for Nehalem?

I know. We may soon see some in house benchmarks, which I will take with a grain of salt of course. But I was happy to see this at IDF. It just shows its coming and that excites me to see new technology.

This is true with any CPU. Even with AMDs. In order to truly take advantage of Phenoms features you need a AM2+ mobo. In order to not bottleneck the CPU with older technologies you need the latest mobo and chipset.

Its a universal system and that goes for current Intel Core2 parts as well. A X38 chipset supports up to 4 PCIe 2.0 x16 lanes where as a P35 will only support one.

This is just the way it is.

Lukebird , if i had a company credit card and told to buy a computer , i would stuff it full of moronic gear because it's faster and i'm not throwing my own cash in a pit and burning it lol.To me it's just Intel hyping there new chip , and i'll sit here and wait for some actual numbers.

To the folks that say Nehalem will be trash for gaming/any gamer that wants a nehalem is a fool....

YOU ARE THE FOOL.

Either you simpley dont understand or believe what Intel has said about this arch. It has been said that it will be a vast improvement over core2. The same kind of jump we saw from the p4s to core2. Or even more. Im dont feel like looking it up, even more so for folks that have been posting pure bullcrap in this thread.

Its not about how many fricking cores a cpu has. Its not about how many Ghz it can be clocked it. It is all about the performance clock per clock. You people are making it out to be just a core2 quad with hyperthreading. IT IS A NEW ARCH. A Nehalem quad @ 2ghz MAY able to crush all the crush all in the current quads AND DUALS in anything. If it destroys core2 clock for clock and can clock anywhere near the current clock speeds, it will simply wipe the floor with it in everything, INCLUDING GAMES. If its alot faster clock for clock then core2 then it will simply destroy them. It will be the best gaming cpu period,

I think many of you are thinking nehelam is going to do the same thing AMD did. Release there next cpu gen that does perform a couple hundred mhz better then the old cpu. But doesnt clock as high to make up the difference. What Intel has been saying is it will be much better then the last gen. If you believe that is the key. AMD preached the same thing and looked what happened.

That is fine. I have a wait and see approach myself. I am just addressing a good deal of the utter crap in the thread

With any new socket design and technology you want to wait a few months before bothering. I would definitely recommend waiting until there's a refresh on this new socket design. I believe it's called socket H.

Give it 6~18 months.

What I'd like to see, in the future, from both AMD and Intel, is dynamic clocking on the fly, where the application tells the CPU how many cores it needs and at what clock, and the multicore CPU responds accordingly.

One problem I could see is when clocks ramps down on their own because they're not being used. I'm still trying to figure out if I have a hampered 3870x2 because both CCC and GPU-Z see one GPU @ 421 with a core clock of 421 and the second GPU @ 421 with a core clock of 850. I guess I could lug another monitor over and set it up so I can see GPU-Z while running 3DMark 06, The Witcher or LOTR Online.

I can see the posts from people later on that their quad or octa core is acting funny, with some cores clocked at 2.0 and others at 2.8 or 3.0. I think we'll need a whole generation of new monitoring software just to make sense of it; identifying a core that's stuck underclocked that might mean a CPU is defective.

At any rate, the thing that will hurt Deneb is hyperthreading. Without the advantages of AMD's hypertransport over a fsb, the extra virtual cores will mean that Nehalem will do 3DMark better than Deneb and perhaps game demo's and benchmarks optimized for more cores. Hyperthreading will probably only appear on the EE models, but it has an affect on branding the lower end CPU's in the public's eye.

Considering the switch to a new socket, I might try a Deneb on a trusty AM2 board that supports it over either AM3 or a $250 Nehalem without hyperthreading on an overpriced new socket.

I work in a data center, but what matters to me when I buy a CPU is how it performs at home, not how it does in a server. Unless I actually built a graphics workstation for my wife, I need a good general home CPU that can do everything I throw at it, even if a bit slower than the top gaming rigs people here have, and I sometimes run other applications in the background when I game.