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

[trek]

We use the open source multimedia stack in the pro driver as well.

and I can confirm that it is running well
previously mesa 10.3 had a minor glitch that, on my hardware, caused to skip the first half-second of video playback, but now it is fixed

[trek]

The truth has always been that if you want open drivers on a discrete GPU you need to buy AMD. But you will have to accept a penalty in terms of performance and features. If you're willing to do that, fine. But don't lie to yourself and everyone else here and pretend that sacrifice isn't being made.

I agree, but I would also add I'm totally happy to pay a little performance penalty to not sacrifice the stability/security/correctness of my system using the open source stack, instead of the uncertainty/unverifiability with the closed one

[LinuxID10T]

If we ever have a lot of free time I would be really tempted to work on a hacked-up set of drivers that cut out most of the API validation code, so we can have a more apples-to-apples comparison.

YES PLEASE! Not sure if you know, but is there more API validation in the OpenGL drivers than the DX drivers?

[dungeon]

YES PLEASE! Not sure if you know, but is there more API validation in the OpenGL drivers than the DX drivers?

Yeah, more ... Try to remove some shader while running AMD's GL blob, it will just throw at you exception. While DX continue to work, but being broken

[tomtomme]

Ah yup, seem you are right 8 out 15 are really 1080p
Why those are there then, so that anybody can ignore them?

exactly :P
the count then is: 5x4k vs 8x1080p

[tomtomme]

I can't believe a Radeon R9 Fury still can't beat a GeForce GTX 950 in games like Bioshock Infinite which has been out for years...

even nvidia results are kind of cpu bound at the tested resolution and the 1080/70 get slapped by the 980/70.
All that matters for such a game is 60fps (its not counter strike).
So the real problem here is: We need more demanding benchmarks / AAA games on linux to test.

[tomtomme]

If it was CPU limited, all cards would be hit proportionately. Starting at the low end and bringing that up is great. I still wait for the day when the ceiling gets raised to where it should be.

I'm not over-dramatizing things. The results speak for themselves. An enthusiast level card gets shown up but an entry level card.

I hate NVIDIA and won't ever buy their cards - definitely an AMD fanboy - but the results are hit or miss depressing still. Makes it hard.

There are real problems in life. Get depressed over those :P - 60 fps are fine for Bioshock infinite.
by the way: The bars state 24 fps as minimum for most nvidia and amd cards - is the port really that bad or is it an outlier? Would be nice to know the 5th percentile of the fps instead of the minimum to be sure.

[tomtomme]

I tended to run at 4K for the games that are knowingly rather weak on the GPU in order to stress them well but then for some games where it would be very low frame-rates, then like in past articles people complain why am I trying to run 4K benchmarks on a $100~150 GPU. I guess complaints either way, just like when I used to run the AMD binary driver always and people complaining I wasn't using open-source, now it's the opposite complaint

you always have to run everything dude: 720p 1080p 1440p 4k on every driver and every card with every game, even those without benchmark mode. Its super easy. just do that.

[vein]

could you try it with the phoronix test suite and post the results file?

Tried the benchmarking and got the same result as Michael did (there not posting it)...hmm...don't really know why since I got about 10fps more when playing...maybe it is like that for all cards...

[vein]

There are real problems in life. Get depressed over those :P - 60 fps are fine for Bioshock infinite.
by the way: The bars state 24 fps as minimum for most nvidia and amd cards - is the port really that bad or is it an outlier? Would be nice to know the 5th percentile of the fps instead of the minimum to be sure.

Actually, I agree completely...I mean, fury is doing 90fps with OSS drivers, it makes me very happy... honestly, I do not care at all if the 950 is doing better than that. What matters to me is that I can use OSS direver and got good framerate...

Only thing I would like to have is good support for OpenCL in OSS so I can do that coding on the same machine...made me a bit sad to hear that it has been abandoned...