By using this site, you agree to our Privacy Policy and our Terms of Use. Close
Soleron on 30 June 2012
Edit Thread
kanageddaamen said:
Soleron said:
kanageddaamen said:

...

 

You're not getting it. I know how a GPU works and what all the parts do, and the effect on performance. All that is nothing to my actual point that a "single processor" operating on a 2x larger screen is still 2x the work for a completely different image of equal complexity being on the other half (notice I said it's NOT 2x the work for a blank screen, and infer that there is a gradient of work between those two if some is shared), while the OP article claims it's insufficient as if 2 GPUs would be better when they are not due to overhead.

Also thre is no such thing as a fragment shader or vertex shader any more. All the computational units can perform both tasks.

I must not be getting it.  I never said anything about a blank screen.  I said, with concrete reasons that, that a "single processor" operating on a 2x larger screen of equal complexity will not require twice the work.  It simply doesn't work that way.  My estimate is that it would probably take about 30-40% more time (ie work) to process a screen double the size and of equal complexity, provided that the rendering isn't being synced to the screen refresh rate.  If it is already running at 60 FPS, its possible you will see zero slowdown.

The bolded part MAY be true for some architectures, but not all, and who cares.  "Fragment shader" is much simpler than "computation unit programmed to run fragment shader operations"

It does take 2x the time because nothing is shared, not loading, triangles, shaders. It's like virtualising two GPUs on one physical board. Obviously if you start sharing the data between two screens it will reduce but I was never referring to that. Your "estimates" are unsubstantiated.

The architecture of the Wii U is known, it's post-R600 AMD and hence unified-shader.

All of which is irrelevant to the point that one GPU > two, always.