ATI's next generation Crossfire (asymmetric GPUs)

[blanketyblank]

After reading the article looks like the first generation is simply "render every Xth frame on the feeble GPU." While guaranteed to increase benchmark scores this will stutter and/or have input lag like nothing seen before or since.

Think of it this way: GPU A can render a frame in 16 milliseconds (running at 60fps). GPU B takes 60ms (running at 15fps). If load balanced 4:1 you'd have to be willing to put up with 120ms of input lag (pre-rendering all 5 frames and then displaying them at a smoothed ~70 fps) not to have awful stuttering.

The grand vision of load sharing, if ever implemented, still holds lots of promise.

True, but I don't really think they would ever really do something like 1:4. More likely in those circumstances you'd just offload some other function to the weaker GPU like some of the tesselation or physics calculations. Otherwise I'd imagine you'd have two cards much closer in performance like say a 5770 and a 5850( bit less than twice as fast) where we'd see a 1:2 and prerendering 3 frames wouldn't be that bad.

 

[cbn]

- split frame rendering could be implemented with one GPU processing all the verticies, and the second for all pixels. Non-3D workload (i.e. DirectCompute, OpenCL) workloads could be scaled appropriately, and dynamically."

http://www.anandtech.com/show/2803/4

According to this article, it is the pixel shader that is the bottleneck to GPU processing times and "input lag".

If one GPU is dedicated to processing all the pixels....how much faster would it have to be compared to the smaller GPU in order for their appointed tasks to be completed at the same time?

Could we see ATI matching the size of its Fusion GPU to its largest discrete card in an effort to beat the Nvidia large die approach on both FPS and input lag?

As I understand things now a product like HD5970 with AFR would deliver higher playable frame rates than a product like Nvidia 480 GTX.....but suffer from more input lag. Could blending Fusion with a discrete Video card using SFR be a way to reduce "waste" while improving performance on both fronts?

 

[Apocalypse23]

A great article, really glad to see crossfire evolving, a few years ago, even when the 3870x2 was released, CF was impracticle, but with the 4xxx series and the 5xxx series, it evolved and still is.

 

[mhouck]

I wounder how this would work in dx11 games w/ say a 5870 and a 4870? would you have to run in dx10/10.1?

I might have missed it in the linked the article. I'm sure somebody will point it out if I did

 

[Dark4ng3l]

I wounder how this would work in dx11 games w/ say a 5870 and a 4870? would you have to run in dx10/10.1?

I might have missed it in the linked the article. I'm sure somebody will point it out if I did

Well of course it would have to run at the lowest common denominator. It's AFR if one card can't do DX11 then you can't do DX11....

 

[tweakboy]

It makes them feel good and looks good on them, with these interviews or previews.

Facts are its not out yet, and wont be out anytime soon. People go ahead get the 5980 dual GPU
or
Wait for Fermi to mature and buy a dual core furbi

 

[mhouck]

Well of course it would have to run at the lowest common denominator. It's AFR if one card can't do DX11 then you can't do DX11....

that's what I assumed. I can't imagine getting a new card to run w/ an old card and be forced to run my new card on old architecture. It's another incremental step in the right direction, but I don't know that the legacy hardware crossfire is really a great selling point. at least you will get improved raw power.