nVidia's CPU dream in a nutshell?

Dario D.

Gawd
Joined
Dec 8, 2004
Messages
582
I don't post threads very often, so I have license to start 2 nVidia CPU threads in a row. ;)

I saw this video of the Mythbusters at NVISION, talking about GPU's replacing CPU's (and nVidia's plans to do that) and realized I've fallen behind on my tech. Is there a definitive summary of everything that has been going on over the past year regarding nVidia's delve in CPU's?

I've been out of the loop, and am realizing that catching up means going through 500 articles, each that covers some individual, tiny piece of the puzzle, without an actual summary anywhere of what's been going on exactly, along with what nVidia's planning to do in the near future. (not even on Wikipedia - unless all they've ever talked about regarding CPU's is the Tegra "SoC" for handhelds, in which case I don't even really care)
 
Last edited:
There is no summary, because there's no NVIDIA CPU...yet.
The video you saw from NVISION was just trying to show how GPUs are much better for specific tasks that require a highly parallel execution, when compared to CPUs. That much is a given and there are already many examples of that. Look for CUDA applications and PhysX executed on GeForces, for more info.
Obviously, since it's a show from NVIDIA, it was also a PR stunt to tout their own products, at a time where Intel's imminent entrance in the desktop graphics cards market, which comes with no surprise. Every company does that.

As for the CPU itself, there are rumors that NVIDIA may actually create one, but there's nothing confirmed. NVIDIA itself, already talked about it and they said that they would try something when the time comes, but Intel is trying to fault their attempts, with licensing claims. You know, the usual Intel bullying, just to keep monopolizing the market. Again, no surprise there, since any company in their position would do it.
You can look at Tegra, which is NVIDIA's SoC (System on a Chip, ARM based) and is a very interesting product, with tremendous computing capabilities, given the size of the devices it's aimed at (handhelds).
 
I don't post threads very often, so I have license to start 2 nVidia CPU threads in a row. ;)

I saw this video of the Mythbusters at NVISION, talking about GPU's replacing CPU's (and nVidia's plans to do that) and realized I've fallen behind on my tech. Is there a definitive summary of everything that has been going on over the past year regarding nVidia's delve in CPU's?

I've been out of the loop, and am realizing that catching up means going through 500 articles, each that covers some individual, tiny piece of the puzzle, without an actual summary anywhere of what's been going on exactly, along with what nVidia's planning to do in the near future. (not even on Wikipedia - unless all they've ever talked about regarding CPU's is the Tegra "SoC" for handhelds, in which case I don't even really care)

You're missing the point a bit... the mythbusters never said anything related to either processor replacing the other. They were simply showing what the GPU does in layman's terms and explaining why it's important.
 
I doubt GPU's will ever replace CPU's simply because they lack the processing capabilities. GPGPU proves that "some" applications benefit from GPUs executing CPU code. In the GPU world everything is highly multithreaded. In the CPU world we havn't reached that stage yet. They are also designed for quite a different purpose. So for right now, the answer is no.
 
I think in the future we will see something like fusion taking over most of the rolls of CPU's, For example maybe 2-4 cpu cores to do the work that a GPU cant do, and then the GPU will do all the work that requires high amounts of parallelism. Of course in 5 years they probably won't call it a GPU except when referring to old graphic cards like the 7 and 1x series GPU's.

"I doubt GPU's will ever replace CPU's simply because they lack the processing capabilities"
Ok, I agree.
I doubt CPU's will ever replace GPU's simply because they lack the processing capabilities.

Of course both these statements work. I simply don't think either will replace the other, at least not in their current form.:)
 
I should read up on this. Anyone have any links?
All I can find is Intel and NV flaming each other. I want to see something on why NV thinks the CPU is going away.

I kind of think the same thing, but I'm thinking more along the lines of a change in how CPUs are architected rather than getting rid of them. Basically my thought is the various parts will eventually end up decoupled from each other, so instead of having a "core" with an integer unit, a floating point unit, and vector hardware they'll all be separate parts.
 
Thanks for the insight, all.

Well THIS is really interesting... I just saw this on Tom's Hardware Guide.
"The really interesting thing about OpenCL and DirectX is that OpenCL is going to form part of the Apple operating system (Snow Leopard) and DirectX (version 11) will form part of Windows 7," Grupta told CNET. "And what that essentially means to consumers is, if your laptop has an Nvidia GPU or ATI (AMD) GPU, it will run the operating system faster because the operating system will essentially see two processors in the system. For the first time, the operating system is going to see the GPU both as a graphics chip and as a compute engine," he said. Additionally, consumers using Windows 7 will see the GPU as a CPU in Task Manager.
If it's true that Intel is running out of breathing room in the CPU market (because CPU's are getting too fast for normal users to need new ones anymore), having OS's that utilize the GPU for running just regular PROGRAMS is going to really put the strain on, because then the CPU becomes even less significant.

Microsoft and Apple are really putting the strain on CPU's... and I like this new direction, because it means progress, and hopefully the eventual blending of the CPU/GPU - something AMD is well, well equipped to spearhead, since they own ATI now. (I don't buy Intel, ever since reading the book "Bill Gates Speaks", and want to see them burn, as punishment for their days of killing other companies by way of mere legal bills / frivolous lawsuits... I consider them founded on blood because of that)
 
Last edited:
Thanks for insight, all.

Well THIS is really interesting... I just saw this on Tom's Hardware Guide.

If it's true that Intel is running out of breathing room in the CPU market (because CPU's are getting too fast for normal users to need new ones anymore), having OS's that utilize the GPU for running just regular PROGRAMS is going to really put the strain on, because then the CPU becomes even less significant.

Microsoft and Apple are really putting the strain on CPU's... and I like this new direction, because it means progress, and hopefully the eventual blending of the CPU/GPU - something AMD is well, well equipped to spearhead, since they own ATI now. (I don't buy Intel, ever since reading the book "Bill Gates Speaks", and want to see them burn, as punishment for their days of killing other companies by way of mere legal bills / frivolous lawsuits... I consider them founded on blood because of that)

I find this an extremely interesting development as well. Really hoping it goes somewhere.
 
I'd have my doubts that any of that is true. Listing the GPU in task manager as a CPU is easy. Allowing code to be run on two physcally different CPU's is not. Not to mention there are potential communication issues. Normal code that we have is x86 or x86-64 but running code for both x86 and whatever GPU code is could result in massive communication issues. If you want intel to burn then you also want x86 to burn (as intel is the founder and owner of the x86 platform) and probably accounts for 95% of systems sold. So careful wishing, the only reason we have gotten as far in CPUs as we have is because of intel and AMD (split sector of intel) and the competition they create, remove that competition and companies are far more interested in profit rather than keeping the consumers interested.
 
The whole cGPU thing makes me think back to the time when intel tried having a separate math co-processor... Although a "cool" idea to increase performance will it actually end up picking up? We all know it ended up just being integrated into the core cpu in the end anyway.
 
the thing i'm worried about is bottlenecking. It seems to me (i might be wrong, and I probably am) that GPU's are advancing at such a high rate compared to CPU's that it is very difficult to keep a system balanced for optimal performance. For instance, for current 200 series gpu's its recommended to have an i7 processor, anything behind that is a bottleneck. I don't think this is true as I've been running a q6600 with a gtx285 and the gaming performance for me is fine. But with the 300 series around the corner, and the incredible amount of power they claim, I get the feeling any CPU will be a bottleneck until significantly new architecture is developed. Am I correct in this thinking, or is bottlenecking a gimmick/marketing ploy?
 
Yeah, I tend to agree...I dont think the future holds tons of co processors located all other the mainboard. That seems like going back in time, plus on top of that you add tons of latencies between all these co processors. Circuitry also becomes more complex. I envision that more and more things will be all going onto a single package or even a die. Just as they are heading that way now, I think eventually a PC will have all of its functionality on one chip and that attached to an IO board. All southbridges and northbridges gone.
 
the thing i'm worried about is bottlenecking. It seems to me (i might be wrong, and I probably am) that GPU's are advancing at such a high rate compared to CPU's that it is very difficult to keep a system balanced for optimal performance. For instance, for current 200 series gpu's its recommended to have an i7 processor, anything behind that is a bottleneck. I don't think this is true as I've been running a q6600 with a gtx285 and the gaming performance for me is fine. But with the 300 series around the corner, and the incredible amount of power they claim, I get the feeling any CPU will be a bottleneck until significantly new architecture is developed. Am I correct in this thinking, or is bottlenecking a gimmick/marketing ploy?

You know, to a certain point it is true CPU's will bottleneck the latest cards, but this is really only seen at lower resolutions. The newest cards are really designed to run the highest resolution gaming out there possible, as you will see once you bring the resolution down FPS at a certain point will top off. Thats where the CPU bottleneck hits. Its because at lower resolutions the latest cards rip through the graphics no problem and waiting for the CPU to feed them more info. But this is not the case on extremely high resolutions, the card is working its butt off while CPU is sitting around waiting for video card.
What I mentioned above is true to a point, because overall a outdated platform will hamper video card performance on nearly all resolutions.
 
Back
Top