NVIDIA 375.10 vs. Linux 4.8 + Mesa 13.1-dev AMD GPU Benchmarks

[duby229]

Not really no. The game isn't CPU limited because NVIDIA drivers for that game are not CPU limited. What is CPU limited are the AMD drivers in the context of this game.

You did not say that.



AMD must be universally as good or better in everything for success.

You can see the 950 beating the Fury in CS:GO @ 1080p also. It's not just one game.

You can see the same in Metro - most of the benchmarks in this round are evidence against your claim.

Which is exactly why I advise to run the benchmarks at settings that reach close to refresh rate and then do a visual fidelity comparison. You will see AMD win.

[Xen0sys]

Which is exactly why I advise to run the benchmarks at settings that reach close to refresh rate and then do a visual fidelity comparison. You will see AMD win.

Is there an automatable metric for testing visual fidelity? Sounds like that might be a valuable addition to the testing suite.

As mentioned, I require no sway for AMD - that's already my company for the foreseeable future. Their dedication to openness is very valuable.

[duby229]

Is there an automatable metric for testing visual fidelity? Sounds like that might be a valuable addition to the testing suite.

Michael implemented that years ago. It can already, what it lacks is a automated facility to find the best settings that reach close to refresh rate. But it already can compare visual fidelity.

[Michael]

Is there an automatable metric for testing visual fidelity? Sounds like that might be a valuable addition to the testing suite.

I've had such 'image quality comparison' capability in PTS itself for over 7 years... http://www.phoronix.com/scan.php?pag...nvik_iqc&num=1

But it relies upon being able to get a precise frame from the game/engine and unfortunately not aware of any reliable paths for doing so except for e.g. Xonotic where thanks to their engine devs was enlightened about a way to get screenshots to happen upon precise events that are reproducibly accurate.

[Michael]

Michael implemented that years ago. It can already, what it lacks is a automated facility to find the best settings that reach close to refresh rate. But it already can compare visual fidelity.

It's trivial for me to add a PTS module to automatically test all available test profile options until finding say the highest result that's below a certain threshold or to find all option combinations above a certain threshold. Super easy. What is harder (though technically not that hard but still not my area of expertise) is a way of effectively visually presenting all of that information in still a universal manner for being able to show all of the data nicely in a visual way.

[Xen0sys]

I've had such 'image quality comparison' capability in PTS itself for over 7 years... http://www.phoronix.com/scan.php?pag...nvik_iqc&num=1

But it relies upon being able to get a precise frame from the game/engine and unfortunately not aware of any reliable paths for doing so except for e.g. Xonotic where thanks to their engine devs was enlightened about a way to get screenshots to happen upon precise events that are reproducibly accurate.

What if you just captured the entire benchmark in video format and do a match that way? May be some difficulty still in syncing the frames but you could probably figure out an offset.

[Xen0sys]

It's trivial for me to add a PTS module to automatically test all available test profile options until finding say the highest result that's below a certain threshold or to find all option combinations above a certain threshold. Super easy. What is harder (though technically not that hard but still not my area of expertise) is a way of effectively visually presenting all of that information in still a universal manner for being able to show all of the data nicely in a visual way.

Sounds like some kind of scoring system would be necessary. If the visual fidelity could be accurately transcribed into numerical values that is.

[Michael]

What if you just captured the entire benchmark in video format and do a match that way? May be some difficulty still in syncing the frames but you could probably figure out an offset.

Probably would be too much work to get that infrastructure setup for what it's worth unless some company/peoples were sponsoring such work, especially not familiar much myself with GStreamer or other video frameworks for coding such functionality. Plus complexities of avoding video compression, etc interfering with the results.

[sdack]

This thread...

Just don't forget the numbers for Windows/DirectX, please. There is still some room for all to improve.

[Xen0sys]

This thread...

Just don't forget the numbers for Windows/DirectX, please. There is still some room for all to improve.

Who's battling who exactly? Sounds pretty out of context.