CPUs as an addition to GPUs? (and not the other way around...)

[petrusbroder]

I thought this is a interesting read over @ ars technica. Would certainly do for crunching!

 

[biodoc]

Interesting read. Thanks for the link, Peter!

 

[rabrittain]

Hmmmm ..., very interesting!!

 

[ZipSpeed]

Nvidia has a fairly strong relationship with Stanford so let's hope the trend continues when these ARM based systems are released.

 

[Cogman]

Bwhahaha!

Ohh, Nooss, Nvidia is releasing an ARM processor! It is the end of the world for Intel and microsoft.

Do people actually read what they are writing? Yes, nVidia is developing a CPU, Great. That doesn't mean that all the sudden they will release something that will run everything faster than intel can. The most they can hope is to swipe away highly parallel work loads, and even that is somewhat sketch as there are some things that GPGPUs simply suck at.

 

[ZipSpeed]

Bwhahaha!

Ohh, Nooss, Nvidia is releasing an ARM processor! It is the end of the world for Intel and microsoft.

Do people actually read what they are writing? Yes, nVidia is developing a CPU, Great. That doesn't mean that all the sudden they will release something that will run everything faster than intel can. The most they can hope is to swipe away highly parallel work loads, and even that is somewhat sketch as there are some things that GPGPUs simply suck at.

I don't think people are expecting Nvidia to displace Intel or even AMD for that matter. Just another alternative, and isn't it a good thing when alternatives are available? Nvidia is simply looking at other ways to survive with their traditional businesses such as chipsets and low-end GPUs being eaten away by fusion products from Intel and AMD.

 

[Cogman]

I don't think people are expecting Nvidia to displace Intel or even AMD for that matter. Just another alternative, and isn't it a good thing when alternatives are available? Nvidia is simply looking at other ways to survive with their traditional businesses such as chipsets and low-end GPUs being eaten away by fusion products from Intel and AMD.

Don't get me wrong, it is a good thing to have more competition. But the way the article is written you would have thought nVidia had just release a product that outperforms the intel platform in every aspect.

That is what I was laughing at. This article is touting the "Death of Wintel" which is particularly funny, all based on nVidia announcing that it is making a processor. This article blows the announcement WAY out of proportion for what it really is.

Nothing would please me more then to see nVidia release a product that is comparable/better than intel and AMD CPUs.

 

[petrusbroder]

Bwhahaha!

Ohh, Nooss, Nvidia is releasing an ARM processor! It is the end of the world for Intel and microsoft.

Do people actually read what they are writing? Yes, nVidia is developing a CPU, Great. That doesn't mean that all the sudden they will release something that will run everything faster than intel can. The most they can hope is to swipe away highly parallel work loads, and even that is somewhat sketch as there are some things that GPGPUs simply suck at.

I agree. BUT: much of the crunching we do is highly parallel work - that is why quite a few of the projects (PrimeGrid, Collatz, GPUGRID, Folding@Home and others) use GPUs for crunching ...
Thus a reasonably strong CPU to feed and coordinate the GPU (or GPUs) is quite a good idea. And the CPU does not have to be the fastest one either.

 

[Cogman]

I agree. BUT: much of the crunching we do is highly parallel work - that is why quite a few of the projects (PrimeGrid, Collatz, GPUGRID, Folding@Home and others) use GPUs for crunching ...
Thus a reasonably strong CPU to feed and coordinate the GPU (or GPUs) is quite a good idea. And the CPU does not have to be the fastest one either.

For distributed computing, yes, this does make a lot of sense. For just about every other computing, it is more meh. But again, this is an announcement that nVidia is making a CPU, and the article writer goes insane about how this spells the end of wintel as we know it. A little bit of an over-reaction for a product that hasn't been fully developed.