AMD & NVIDIA: Open vs. Closed-Source Driver Performance

[phoronix]

Phoronix: AMD & NVIDIA: Open vs. Closed-Source Driver Performance

Continuing on from this weekend's open-source Nouveau vs. closed-source NVIDIA Linux driver performance are results now added in with showing AMD's open-source vs. closed-source driver performance with the same tests.

http://www.phoronix.com/vr.php?view=23995

[Ground0]

Nice Test. That's why i buyed a Premium Account :-)

[Michael]

Nice Test. That's why i buyed a Premium Account :-)

Thanks!

[LazarusIV]

Wow, AMD open-source drivers *destroy* nVidia open-source drivers... Didn't realize it was that bad. In a lot of cases, the AMD open-source driver outperforms the proprietary driver, pretty good deal! Seems like my choice to use Manjaro-stable with my R9 Fury was a good one! Phenomenal performance, especially considering the price I paid for the Fury!

[funfunctor]

TL;DR - Buy AMD Radeon, use a recent distro and do nothing to get working high performance graphics out the box with the open stack.

[patstew]

It seems like AMD have got to driver parity by gimping the proprietary driver as well as improving the open source one. As far as I remember, back in the fglrx days it performed similarly to windows (when it worked), but now it seems an RX 480 with either driver is competitive with a GTX 770, whereas on windows it trades blows with a GTX 980. That's quite a lot of performance (/money) to give up for the ability to edit the driver source code.

[duby229]

It seems like AMD have got to driver parity by gimping the proprietary driver as well as improving the open source one. As far as I remember, back in the fglrx days it performed similarly to windows (when it worked), but now it seems an RX 480 with either driver is competitive with a GTX 770, whereas on windows it trades blows with a GTX 980. That's quite a lot of performance (/money) to give up for the ability to edit the driver source code.

Unfortunately it's a bit more difficult than that. AMD's OpenGL performance is not as good as their DX performance even on Windows. And to exacerbate the problem most ports to linux underperform. The real solution is game engines written for linux ports in mind and Vulkan.

[bug77]

It seems like AMD have got to driver parity by gimping the proprietary driver as well as improving the open source one. As far as I remember, back in the fglrx days it performed similarly to windows (when it worked), but now it seems an RX 480 with either driver is competitive with a GTX 770, whereas on windows it trades blows with a GTX 980. That's quite a lot of performance (/money) to give up for the ability to edit the driver source code.

Nope, fglrx was never competitive with their Windows driver. Maybe in a few, selected tests, but ATI/AMD's OpenGL implementation has always been sub-optimal.

[cj.wijtmans]

It seems like AMD have got to driver parity by gimping the proprietary driver as well as improving the open source one. As far as I remember, back in the fglrx days it performed similarly to windows (when it worked), but now it seems an RX 480 with either driver is competitive with a GTX 770, whereas on windows it trades blows with a GTX 980. That's quite a lot of performance (/money) to give up for the ability to edit the driver source code.

i was going to say my 770 performs like the rx480 in these benchmarks and other options are a bit expensive. Upgrading is just not worth it for me yet unless i buy another nvidia card :/

[adakite]

Very very nice work! Thanks Micheal.

AMD open-source work is really nice, while Nouveau folks have done a great job, specifically for the Kepler line.

Beside game, I'd like to see similar comparison with computing task on both fp32 and fp64 as it seems that Nouveau allows this now AFAIK.