GPUs processing power

I stumbled on the values of processing power for some GPUs ..... & I was wondering how can that be translated to gaming performance?

The thing that arouse that question is something I found pretty strange & that is:

Mobility Radeon HD 5870: has 1120 GFLOPs
while
GeForce GTX 285M: has merely 576 GFLOPs

..... so the the GTX 285M benches close to the 5870 while it has half its processing power

 

Uh, two different things. Floating point operations are not the same as polygon output.

The Processing Power is for stream processing, look up ATi Stream for more information.

One thing it could be used in the future is for better physics, and not the lame hardware accelerated from Nvidia.

 

Both GPUs have different architecture, you can't compare them directly.

 

Short answer: these values tell you nothing about gaming performance.

 

Not quite true.

The thing about DX11 I think developers are most excited about is not Tessellation but Direct Compute. Direct Compute will utilize this computational power to offload some work usually done on CPU to GPU. For example, Depth of Field is a Direct Compute aspect which for some make gaming a better experience.

But you need a beefy GPU to really do this, like a desktop HD5870 and GTX 480. The Mobile HD5870 can do tessellation and some direct compute operations but looking at 30 FPS not the 60 FPS many gamers want.

The Mobile HD5870 proven to utilize Direct Compute though. Bad Company 2 has a lot of soft shadowing, Depth of Field, Deferred Lighting, SSAO and these are all things that are done in DX11 Direct Compute.
- Some users have noted DX11 runs better than DX10 because of Direct Compute utilizing the computational power of the GPU otherwise not done with DX10.

So yes it does matter for gaming and in future games will matter even more. Hav0k, Bullet Physics and other physics engines utilize GPGPU power also.

 

No, just check out the OP's initial post. The number of gigaflops for processing power on itself is meaningless for games. There are many more important factors that determine game performance, which you need to take together to have any sort of comparison between cards.

 

I find the topic kindda interesting but im an ignorant can u mention some of the factors ????
I think the only gain from overclocking a GPU's core is processing power ... correct me if im wrong
I take it that the processing power is completely unaffected by the memory, & memory plays a vital part in performance based on the bus width & bandwidth

 

The main factors for ranking GPUs would be first the bus width (64 to 256-bit), then memory type (DDR2 to GDDR5). A 64-bit DDR2 card will never be as powerful as a 256-bit GDDR3 card, even if it has a million shaders, 2GB of memory, and clocked at insane speeds (due to bandwidth limitations). The next important thing would be the games themselves, as certain games perform better on nVidia, while others with ATI. Due to different architectures, coming shaders and clock speeds between companies isn't easy but the old rule of thumb was 5 shaders in an ATI card was around 1 shader for nVidia (though I have found that lately, this doesn't always hold up).

 

^ And that is about what I know also. I think the bottomline is that benchmarks of actual games between cards is the easiest way to translate into GPU power.