09-20-2012, 03:44 AM
The only game that truly needs it is Mario Sunshine. Anyway, in 8 hours from writing this post (the time is now 1:38PM EST) I'll just go ahead and start the test with the game that has the most votes.
09-20-2012, 03:44 AM
09-20-2012, 06:01 AM
This should test the strength of the overall system with Dolphin very well, but I have an idea that will work well for GPU benchmarking:
D3D9, 4x IR, 9x SSAA, and any other GPU intensive enhancements/accuracy changes besides 3d. The game I'm thinking of isn't the most popular (Super Maril Strikers), but I think it has some of the most demanding animations (specifically the super strike; it always gives me a massive slowdown on 4x IR even without other enhancements).
The speed difference is massive.
D3D9, 4x IR, 9x SSAA, and any other GPU intensive enhancements/accuracy changes besides 3d. The game I'm thinking of isn't the most popular (Super Maril Strikers), but I think it has some of the most demanding animations (specifically the super strike; it always gives me a massive slowdown on 4x IR even without other enhancements).
The speed difference is massive.
09-20-2012, 08:01 AM
Animations are based on vertex transformations, which don't scale with resolution. What you really want is a game that requires intensive pixel shaders to emulate.
And I wouldn't try 4x IR + 9xSSAA because that could cause some video cards to run out of video ram and crash the application. 4x IR + 4xSSAA or 3x IR + 4xSSAA would be a lot more reasonable. The latter should work on cards with 512MB of video ram without issues.
And while a GPU benchmark would be cool we already did one I think it's more important that we turn this into a cpu oriented benchmark (minimal graphics settings). Considering that we already have some data on GPUs but virtually none on CPUs. Then again this is all up to SS.
And I wouldn't try 4x IR + 9xSSAA because that could cause some video cards to run out of video ram and crash the application. 4x IR + 4xSSAA or 3x IR + 4xSSAA would be a lot more reasonable. The latter should work on cards with 512MB of video ram without issues.
And while a GPU benchmark would be cool we already did one I think it's more important that we turn this into a cpu oriented benchmark (minimal graphics settings). Considering that we already have some data on GPUs but virtually none on CPUs. Then again this is all up to SS.
09-20-2012, 08:27 AM
(09-20-2012, 08:01 AM)NaturalViolence Wrote: [ -> ]Animations are based on vertex transformations, which don't scale with resolution. What you really want is a game that requires intensive pixel shaders to emulate.
And I wouldn't try 4x IR + 9xSSAA because that could cause some video cards to run out of video ram and crash the application. 4x IR + 4xSSAA or 3x IR + 4xSSAA would be a lot more reasonable. The latter should work on cards with 512MB of video ram without issues.
And while a GPU benchmark would be cool we already did one I think it's more important that we turn this into a cpu oriented benchmark (minimal graphics settings). Considering that we already have some data on GPUs but virtually none on CPUs. Then again this is all up to SS.
It's not just an animation; it also introduces a bunch of brightly lit clouds around the character (which I would think means a lot of lighting and shading) and causes my computer to slow to a crawl (like 10 FPS) every time.
But yeah your right on the fact that GPU benchmarking is done for the most part. Although the GPU list you made you have admitted isn't perfect, so it could use some more work if we have time to deal with it.
09-20-2012, 08:44 AM