Re-Testing NVIDIA's Threaded OpenGL Optimizations For Linux Gaming

[phoronix]

Phoronix: Re-Testing NVIDIA's Threaded OpenGL Optimizations For Linux Gaming

Back in 2012 with the NVIDIA 310 Linux driver series a threaded OpenGL optimization was added to the proprietary graphics driver. When this driver premiered we tested NVIDIA's Linux threaded OpenGL optimizations to mixed results. We're back now re-testing the OpenGL threaded optimizations to see if it makes any more of a difference now with modern Linux games and OpenGL workloads while using the latest 337.25 Linux driver.

http://www.phoronix.com/vr.php?view=20582

[AnAkIn]

__GL_THREADED_OPTIMIZATIONS is already set in all Source games, so the results should be exactly the same with and without it. Looks like the benchmarks are not really accurate.

[hajj_3]

you should try and create your benchmarks using html instead of flash. Lets kill off flash already.

[Michael]

you should try and create your benchmarks using html instead of flash. Lets kill off flash already.

I don't use any Flash for benchmarking results...

[mufasa72]

__GL_THREADED_OPTIMIZATIONS is already set in all Source games, so the results should be exactly the same with and without it. Looks like the benchmarks are not really accurate.

They were all within 5% of each other - might just be due to random variation. If this is the case it would be interesting to see the "optimization" turned off. Are we sure there isn't an @deprecated tag in front of it (it is from 2012...)?

[Guest]

It would be interested to see performance of games that use OpenGL -> DX translation (such as witcher2, dota2).
Actually these games on steam have it enabled by default afaik, so turning it off in their respective wrappers could see performance benefit/drop.

[Guest]

Dota2 [default - ON]

Ok, I have just tested dota2 with highest settings on my optimus laptop (using randr offloading feature).

I fired up the same demo at highest settings and difference between fps was quite massive (17fps->35fps, 1000 frames stat).
If you wanna try disabling it, open this file in your favourite text editor:
<pre>
..pathtosteamfolder../SteamApps/common/dota\ 2\ beta/dota.sh
</pre>

and change line:

<code>
# enable nVidia threaded optimizations
#export __GL_THREADED_OPTIMIZATIONS=1
</code>

into:

<code>
# enable nVidia threaded optimizations
#export __GL_THREADED_OPTIMIZATIONS=1
unset __GL_THREADED_OPTIMIZATIONS
</code>


Similar should be true for other valve games.

[Guest]

edit: Acutally fps was skewed by some settings in autoexec.bat (sorry about that):


ON MAX [AA, 1920x1200, everything maxed, VSYNC OFF]
fps with TO: 17
fps without TO:25

ON LOW [1920x1200 everything on lowest setting, VSYNC OFF]
fps with TO: 58 (range 15 worst-92 best)
fps without TO: 60 (range 38-86)

Conclusion: while my testing was imperfect and amateurish, I think that it has proven (on MY setup) that TO doesn't benefit performance of dota2.
Even at low settings where where fps of both settings was close to idea 60fps, dota2 command (cl_showfps) shows that best/worst framerate was much less stable than when TO was unset.

[mufasa72]

Conclusion: while my testing was imperfect and amateurish, I think that it has proven (on MY setup) that TO doesn't benefit performance of dota2.
Even at low settings where where fps of both settings was close to idea 60fps, dota2 command (cl_showfps) shows that best/worst framerate was much less stable than when TO was unset.

Sweet, thx! I've submitted a bug to NVIDIA - let's see where this goes... http://nvidia-submit.custhelp.com/ap...il/i_id/648369

[stqn]

I tried The Witcher 2 with "__GL_THREADED_OPTIMIZATIONS=1 voglperf32 20920" and the "fpsprint on" command, but I got the same FPS with and without threaded optims. Enabling or disabling compositing (Compton) also didn’t change anything, BTW. Around 20-25 FPS in 1680?1050 on low detail on a Core i3 550 + GTX 660.

PS: vsync disabled in the nvidia panel and in the game settings. Compton was set to vsync, on the other hand.