Why AMD had to buy ATI

While I agree that a Fusion-esque APU is the future, discrete graphics cards won't be disappearing for a long time, as long as there are gamers that will buy them and they perform better than whatever is integrated. I think the Inquirer is exaggerating things a bit here.

nVidia is far from dead.

Hopefully it's better then the ast GPU Intel released. I mean they're talking about 16x faster than G80.

I'd ike o see how they manage it considerng the power reqs for G80/R600.

They are saying it will be x86 "mini-CPUs" like Cell or XBox360 CPU. It will be interesing to see, but hopefully no one really thinks nVidia can be beat.
R700 is already set to be modular and nVidia will undoubtedly want to really out-distance Intel after this announcement. And considerng the time between G70 and G80, Intel will NEVER catch up.

if anything it'll be like ATi where ntel releases a chip that's faster than nVidia current and nVidia releases a new monster that lows the competitin away.

nVidia has the most experience in 3D GPUs and Intel - no matter the money they throw at it - will not be able to make up the 10+ year headstart.

sure it sounds like a monster, but how well will it really push the pixels? who knows. it sounds like it will be one helluva server add on type chip not a gpu.

but hence the thread title... AMD did what it had to do early....to be viable in 7 years...

Employees by themselves won't do it and if they(employees) steal manuf or trade secrets, Intel can be prosecuted.

They will never REALLY catch nVidia or even ATi.

Upon reading this I have concluded that it is not outside the realm of possiblility for intel to become a powerhouse in graphics. With the expertise of 3D Labs engineers and intels very deep R&D pockets, it would be very foolhardy to become complacent about intel becoming a player in the GPU market. This event was probably what triggered the eventual merger of AMD/ATI.

I wonder how much it cost intel to obtain this level of graphics expertise? I don't know, but I bet it was alot less than 5 billion dollars.

its still amazing to me AMD bought ATi

don´t panic ...if you think that the inquirer is incorrect


http://www.xbitlabs.com/news/video [...] 30933.html

Employees by themselves won't do it and if they(employees) steal manuf or trade secrets, Intel can be prosecuted.

They will never REALLY catch nVidia or even ATi.

That won´t do Nvidia any good though. If intel decided to go that route (and intel isn´t quite the nice guy if it comes to laws and morals) they will damage Nvidia to a degree that will make it worth it for them paying any fine. The worst case scenario would be they compete with nv, devalue them by stealing their tech and then buy them... 8O

So, I will speculate that nVidia will eventually buy the company with x86-license, VIA.

Employees by themselves won't do it and if they(employees) steal manuf or trade secrets, Intel can be prosecuted.

They will never REALLY catch nVidia or even ATi.

Only a Sith lord Deals in absolutes...

People keep talking about the deal as if integrated gpu was what it was all about.
What about the math co-processor function of today"s gfx cards. What about having a physics core in the processor, or the drivers to use the same setup in either way?
Ati had a great chipset unit and some great coders. These people might be able to help Amd out with it's problem getting SSE working properly.
Then again, if one of the new fabs is a little slow on work, it might be handy.
Bottum line though, is that Ati is a good investment.
If Amd is ever going to catch up with Intel, they need to diversify.

Do not mock the power of the Dark Side.

you are assuming that fusion-type performance will not be able to compete with discrete at some point

Discrete cards are upgradable....GPU needs grow faster than CPU needs. This is the same reason why you have slots in your PC case.

Ok, I have small amount of time.

"I, for one, will be severely disappointed if we are still tapping on QWERTY keyboards designed at the time of the US Civil War to keep fast typists from sticking their mechanical levers."

I would be disappointed too. I realize AMD's FUSION initiative is really focused on graphics at this point, however I believe they also have a Torrenza initiative that has a number of disparate businesses claiming interest. That led me to believe AMD could add-in a variety of coprocessors. Think for a minute ... graphics processing uses stuff from image processing which uses stuff from signal processing. Hmm, signal processing, like DSP. Not sure if AMD has contacted TI yet, or vice versa. If so, then good-bye keyboard. Catch my drift?

Hey, anybody from AMD, care to hire me yet? 8)


--M

I do not see this as problem, they will simply add SSE6 or something like that to architecture. Actually, I think they will do that for general purpose CPUs sooner than this GPU goes out.

Personally, while following is just speculation, there is one aspect of this architecture that has VERY strong advantage - if they do it right, GPU code will be simply the part of regular x86 program, I mean it will be sitting in normal memory (GPU will cache it to hide differences), it will be produced like normal code, just there will be perhaps special StartGPUThread API function.

If this is true, this thing will be awesome. Forget about DirectX or OpenGL in that case. This is about ray-tracing.

Mirek

Check this out and see how nicely it would fit into CPU/GPU tech! 8)

Hi all,

Speaking about the future is always interesting.

Here's my input:

I think that moving the graphics pipeline onto a single chip is a good idea, i mean with the prospect of having multiple cores on a single chip and the ever increasing cost multiple computer components eg. The 8800GTX,

Moving everything onto a single chip would not only save the overhead and latency associated with any fancy busses like AGP or PCI Express, but it would also lead to a more elegant solution for home users. It would also prevent incompatibility problems between components. If companies did bother to do some extensive testing they would also find an optimum point for their hardware configuration eg. 16 GB of ram is enuff for the processing power on this particular chip model.

Ultimately the computer would consist of two components: the Bigchip (Processor cores, ram, any other stuff) and the Mother board with a peripheral interface chip, so upgrading would be simple, if u want more usb or SATA ports you get a new motherboard, if you want more power u get a new processor chip (Bigchip). This would also lead to a smaller number of pins that will need to connect to the motherboard since the cpu has onboard ram and only needs to connect to the motherboard for input/output and access to storage devices.

Ultimately everyone has the a limited amount of Silicon real estate in their computer, but if you can redistribute it to make it work far more efficiently then that's obviously the best way to go. ie. If you put it all onto one big chip you save siliconused for overhead of busses and component interconnects and use it most efficiently.

It seems to me that while the past year revealed consoles that resemble computers, it seems quite appropriate that computers in future resemble consoles in the sense that they are marketed as one single unit that has each component within designed to work optimally with the rest. Consoles of the past have proven to be far more efficient than their PC equivalents for example if one tried to get a pc with 32 MB of ram to run Final Fantasy X i doubt you would have any success. Yes a PS2 has 32 MB of ram. Are PS2's just magical? I think not.

In reference to graphics processors, I think that the current architecture will be around for a while, the way that it has been setup is quite good for a graphics pipeline, i mean the cpu moves the polygons around and the graphics card keeps drawing them periodically. All the current games use it and the only reason to change would be if the integrated solution out did the current arcitecture, this is still a fair way off.

One last thing: I do not like Intel's architecture of the 16/8/4 core processor as stated in the inquirer, Many processors simply connected in a ring bus. I think that arcitecture like that of the Cell Processor is much better suited to home computers since the operating system can be loaded into a main processor that can implement management functions and can manage the other smaller co-proceseors like processing resources, easier to program for, and easier to manage many cores.

In this future respect AMD buying ATI was an incredibly smart move, especially if Intel and nVidia relations start straining, however the benefits are still a while off.

Personally I see DSP on the desktop in some sort of Fusion brainfart well before Cell. But that's just me.

I'd love to see Bigchip and Mama in my desktop, and I agree that it is the way that the industry is going. However, I don't know if most PC buyers are ready to buy the one-size-fits-all over the modular philosophy, at least right now. PS2s get along fine on 32MB RAM since that is what the software is designed to handle. Over on the Windows (and Mac) side, we're seeing RAM requirements skyrocket. I'm looking at getting 8GB in my next system just so I can feed Photoshop 4GB and run the rest of my stuff, including OS in the other 4GB. Now it wasn't too long ago that an 8GB HD was a big un! Now I need that much in RAM. Much of the humongous RAM requirements are due to lazy coding and mediocre programming, but some just can't be helped. When I ask Photoshop to manipulate a 2.5GB file where do I expect it to keep the data? As for cell, I can't help but agree with verndewd that the writing the code for it is a bear, but then again, I'm not a programmer so I can't say first-hand. I don't see the ATI buy as being that smart. I see that Wall St. has voted with its wallet and severely punished AMD stock since the buy. And when it comes to the bottom line, nobody quite knows it like the boys on Wall St.!