how is the CPU graphics driver overhead on AMD cards in GPU-bound situations, vs. NV?

[soccerballtux]

We've seen some pretty condemning evidence of driver inefficiency, AMD cards topping out at 95FPS in GPU bound situations while Nvidia managing all the way up to 120FPS.

I'm wondering if the CPU is a bottleneck at 60FPS, when the card is. Theoretically, the CPU overhead in use at 90fps would be 50% higher than at 60fps.

In other words, maybe it doesn't matter.

 

[psolord]

I've read about this and it has worried me a lot.

I switched from 7950 crossfire to GTX 970 and I have to say that I am very impressed with the 970.

I have seen some very high framerates in games that have been known to be quite cpu limited but I cannot really say that this is due to the better Nvidia cpu overhead.

I cannot do a fair comparison since the 7950s need to be both in operation to come close to 970 performance and that by definition increases the cpu usage.

What I could do however, is find games that both cards can run at 60fps and compare the cpu graph from task manager, or msi afterburner or use resource manager or something.

Would this work?

Also is the problem of the driver or the architecture? I mean will Fiji be more cpu efficient or it will be the same?

 

[Silverforce11]

In GPU bound situations?

Obviously the overhead leads to the CPU working more to achieve a particular FPS. But a quad core intel won't have issue keeping most games GPU limited.

This is one of the reasons why total system power looks worse on AMD. When measured on the GPU alone, R290/X is ~225-240W, not much more than 780ti, but in total system power, the gap grows to 50-75W delta. The reason is the CPU in the AMD system needs to work harder.

They definitely need to stop wasting time with Mantle and get a proper multi-thread rendering for DX11, its still going to be widely used once DX12 gets here.

 

[Unoid]

let's all just be honest. we all know nvidia is better, websites say so. heck even danish cats say so.

SCIENCE is settled!

 

[futurefields]

when I play Advanced Warfare my 2500k @ 4.2ghz bottlenecks my GPU to about 50% usage before frameratr drops occur. Nvidia users dont experience this.

 

[happy medium]

Is it possible that my Q9550 @ 4ghz, with direct x 12 optimizations and a Nvidia card will last a bit longer?

My rig is so old its getting rusty , and it still games.

 

[bystander36]

I'd like a study on it, because I know it isn't true all the time. A while back, Tom's Hardware tested a bunch of games for this, and it did show it favored Nvidia, but that doesn't mean it is still true today.

I noticed CS:GO favors AMD in this department.

 

[soccerballtux]

Is it possible that my Q9550 @ 4ghz, with direct x 12 optimizations and a Nvidia card will last a bit longer?

My rig is so old its getting rusty , and it still games.

I think even with an AMD card you'd be fine.

 

[Spjut]

I'd like a study on it, because I know it isn't true all the time. A while back, Tom's Hardware tested a bunch of games for this, and it did show it favored Nvidia, but that doesn't mean it is still true today.

I noticed CS:GO favors AMD in this department.

I'd also like to see overhead comparisons between the HD 5000/6000 vs Fermi.