22-Way Comparison Of NVIDIA & AMD Graphics Cards On SteamOS For Steam Linux Gaming

[artivision]

Please show us evidence for your claims. Since you claim that the free drivers can outperform Catalyst in many cases you must have data to base your claims on.

http://www.linuxsystems.it

[duby229]

Please show us evidence for your claims. Since you claim that the free drivers can outperform Catalyst in many cases you must have data to base your claims on.

I don't have to, there are plenty of other people that already have. In fact it's common knowledge by now.

http://www.linuxsystems.it/2015/08/radeonsi-with-si-scheduler-humiliates-catalyst-in-all-tests/

[duby229]

Don't you read the news? AMD is doing this, is investing in the AMDGPU opensource driver, with also closed source support but user level only binary blob, for those that need certification or features still lacking in the open source ones.

AMD want to kill the close source driver, but cant do that quick enough. Developing drivers takes time... even intel, with huge resources invested in the linux drivers, still don't have drivers with all the features.

What i wanted to see here is this same test with the open source drivers and APUs also tested... also, both fast and slow CPUs, to see how much the CPU affects the performance

Yeah, personally I don't see any problem at all benching the OSS drivers for AMD hardware against nVidia's proprietary driver. For the life of me I just can't understand why Catalyst is still being used. Anybody with a brain will use the OSS drivers. If you use AMD hardware then it's the only thing that makes sense to use and therefore it's the only thing that makes sense to benchmark. I can understand if you have a specific reason that you want to benchmark Catalyst, but comparisons like this one clearly don't fit that bill.

[Michael]

Yeah, personally I don't see any problem at all benching the OSS drivers for AMD hardware against nVidia's proprietary driver. For the life of me I just can't understand why Catalyst is still being used. Anybody with a brain will use the OSS drivers. If you use AMD hardware then it's the only thing that makes sense to use and therefore it's the only thing that makes sense to benchmark. I can understand if you have a specific reason that you want to benchmark Catalyst, but comparisons like this one clearly don't fit that bill.

The open-source drivers (r600g included) can't even run all the tests done here.

[duby229]

The open-source drivers (r600g included) can't even run all the tests done here.

Yeah, that's currently true. Two that definitely won't run, but all the others have been patched or can run with overrides if you use the latest code.

[MoonMoon]

I don't have to, there are plenty of other people that already have. In fact it's common knowledge by now.

http://www.linuxsystems.it/2015/08/radeonsi-with-si-scheduler-humiliates-catalyst-in-all-tests/

The drivers with the si patches were in that benchmark tested with exactly one card. How that translates to "the driver in general" is faster in many cases (contrary to "the driver running with the HD7950") is beyond my understanding, calling this a simplification would be the understatement of the year. Not to mention that running not even 10% faster in most benchmarks is far from humiliating. Yes, I would also like to see benchmarks of this on a wide range of cards, but concluding from this benchmark anything in general goes way to far.

[pal666]

tests without real linux drivers(mesa) are pointless

[dungeon]

nvidia drivers, not only use use a option to enable better multi thread for openGL (with some risks, but usually works), but also are more CPU efficient.
Catalyst drivers are more CPU intensive and always single thread, so the less powerful is the CPU the worst the driver will perform

No, Catalyst are not always singlethreaded, for Metro driver enable threaded profile... Is the render singlethreaded or not depends on game used some are some are not, so that really deppend on a game some games are alreadeady threaded, well some are not... so, sometimes driver trying to do something about.

[dungeon]

I don't have to, there are plenty of other people that already have. In fact it's common knowledge by now.
http://www.linuxsystems.it/2015/08/r...-in-all-tests/

That particular test results is fake , no one else had not reproduced this kind of performance . Even author commented on irc i think, that something getting wrong there.

And no, radeonsi can't compare with current Catalyst driver where profiles are on. Maybe radeonsi can compare with year old 14.9 something driver, but also with any profile disabled. But yes, i can agree that radeonsi can compare to default Catalyst sometimes.

But good that you point out that, so we can look on Bioshock nvidia and catalyst results and compared it with this article... That was High and now Very High, humiliatedly horrific or golden fake, what do you think about... what happened there i dunno

Two months after on both nvidia and fglrx drivers Very High became faster then High or something

[dungeon]

@dungeon

Since when do you need a x86 CPU for Android gaming? Btw. XBox was x86, XBox 360 PowerPC, XBox One is x86. But Win 10 runs on ARM too, same base. PS3 was PowerPC+Cell, PS4 is x86. This all means nothing for next generation hardware. There has to be high GFX power for games, Nvidia is working on Pascal with HBM2 support. Could easyly combined with ARM cores - current Tegra X1 has already compute power similar to Intel i3 - I am sure AMD will face a huge competition - even Intel has stacked DRAM like the AMD chip for MS.

Well we will see who knows maybe even new VIA x86 can do the job, with right GPU, right API and more cores