35-Way NVIDIA/AMD Proprietary Linux Graphics Driver Comparison

[brosis]

Sad, but nice (think Judd Nelson, Breakfast club)

Compared to the Nvidia (Linux and Windows) and ATI (Windows) driver performance, these results are sad.

Compared to prior Radeon open source implementations, they are nice.

No. No sadness here.

http://openbenchmarking.org/result/1406061-KH-BINARYDRI44,1406022-KH-OPENSOURC58&compare

Nvidia binary is of equal performance M$/LINUX, then
Xonotic 1920x1080, Ultimate,
GTX 760 proprietary - 140 fps, 185? minimal price in Europe.
Radeon HD 6870 (opensource radeon) - 110.96, 124? minimal price in Europe.
Radeon HD 6870 (catalyst) - 104.94

Sad is only one thing: Michael does not test the same cards over same tests. Although I understand that some cards might not work with opensource well, he should still test them or set 0 score as N/A.
But because he issues mostly different tests to compare linux:catalyst/radeon//nvidia/nouveau and omits M$:catalyst/nvidia, the tests are not very good to compare.
He does huge work, but this is sad.

[mitcoes]

Add some 720p (1k) and some arch to the tests please ot name it phorobuntu

I have just buy a NVIDIA GTX 750 Ti
I tested it at 720p (1k)
in Ubuntu 14.04 default kernel, Half Live Lost Coast test 100 fps
in Manjaro 3.14 kernel 150 fps

It was not the same kernel, but I bet the Manjaro kernel settings are faster than Ubuntu ones. In a new Xubuntu it takes seconds to start Chrome and not at all in Manjaro.

You asked if we would like to see other distros tested and Manjaro had a good share of the votes, and it is a good benchmark for Ubuntu if they are doing something not good enough compared with other distro.

Last but not least, usually Linux boxes are not state of the art ones, so 720p (1k) game benchmarking that will give twice or more FPS than 1080p (2k) and would make some games playable or fun to play and the video quality is very similar, would be very interesting to read. Soon enough test will be only 4k and it would be better if we can read 1k +2k + 4k test together to watch if the GPU scales well with that test and at what resolution it is playable that game with each card.

[Lysius]

What happened to the GTX 460 results for TF2?

[deppman]

http://openbenchmarking.org/result/1...OURC58&compare

Xonotic 1920x1080, Ultimate:
Model ----------- Open - Proprietary
Radeon HD 5450 8.43 14.27
Radeon HD 5770 69.92 88.03
Radeon HD 5830 94.33 96.56
Radeon HD 6450 16.50 19.54
Radeon HD 6570 35.81 47.28
Radeon HD 6770 69.59 89.56
Radeon HD 6870 110.96 104.94
Radeon HD 6950 119.75 122.32

This is very nice.

Sad, but nice (think Judd Nelson, Breakfast club)

Compared to the Nvidia (Linux and Windows) and ATI (Windows) driver performance, these results are sad.

Compared to prior Radeon open source implementations, they are nice.

[brosis]

Sorry, I didn't get it :P

Well, it seems to me that you are bashing AMD a little, but overall I agree with you. Hopefully the open-source drivers will improve their performance even more!
I can't wait the day when mesa will beat fglrx in most of the benchmarks

http://openbenchmarking.org/result/1406061-KH-BINARYDRI44,1406022-KH-OPENSOURC58&compare

Xonotic 1920x1080, Ultimate:
Model ----------- Open - Proprietary
Radeon HD 5450 8.43 14.27
Radeon HD 5770 69.92 88.03
Radeon HD 5830 94.33 96.56
Radeon HD 6450 16.50 19.54
Radeon HD 6570 35.81 47.28
Radeon HD 6770 69.59 89.56
Radeon HD 6870 110.96 104.94
Radeon HD 6950 119.75 122.32

This is very nice.

[Michael]

Great work Michael! You've obviously put a lot of effort into making these huge comparisons, and the community really appreciates it. Hopefully these great reports will help phoronix attract more subscribers.

Thanks. Sadly, the articles barely subtracted many new subscribers nor tips, so won't be doing such large comparisons regularly.

[inhuman4]

Great work Michael! You've obviously put a lot of effort into making these huge comparisons, and the community really appreciates it. Hopefully these great reports will help phoronix attract more subscribers.

[pinguinpc]

Because this testing took over a week to do, the 340 driver was just released this week.....

Ok

Thanks for respond

But if you can try test wine with nvidia and amd hardware (thankfully you have amd/nvidia cards of various power/performance segments) with propietary driver, this test its really need and dont see on web

At least good test

[krasnoglaz]

Awesome work! I really like new power consumption graphs.

When Valve releases Left4Dead 3, I hope it will become a good benchmark to really load modern GPUs. Until then Unigine Heaven is the test to look at. Simpler games just cap out at ~200 fps even on lower speced cards. Also it would be nice to see how GPUs work with anisotropy antialiasing enabled, because I don't play without them (4x minimum for both).

[oleid]

So it's not that grim anymore for AMD, as soon as you take the power consumption under consideration. Nice!