Updated nouveau benchmarks
Nouveau developers don’t stop to amaze me. Really.
Not only did they manage to reverse engineer one of the most complex chips is the world, but
the performance recently improved quite a lot:
This is low-mid end mobile Geforce 8400M GS card running at 1280×800.
|3D application||nvidia binary (FPS)||nouveau (FPS)||%||comments|
|0AD||48||32||66%||oasis map, few seconds into game|
|Nexuiz||78||50||64%||tutorial level, ‘normal’ graphic effect level|
|unig. tropics||26||16||61%||1st scene, different GL versions|
|Neverball||183||106||57%||1st level, corner.|
|unig. heaven||11||6||54%||1st scene|
|helloracer webGL demo||65||35||53%||just start, not move car, maximum|
|etracer||110||50||45%||secret valleys, few seconds into game|
|supertuxkart||100||45||45%||tux tollway, few seconds into race|
|glxgears||2907||880||30%||Yes, I know…|
|Torc||70||20||28%||On default track, few seconds into race|
A lot of things changed in nouveau meanwhile.
I switched to KDE, I no longer use compiz. I also benchmarked the games a bit differently,
thus don’t expect same fps for binary driver. These didn’t really regress, I just measure them a bit more accuratly
(but I really didn’t skew the result any way. Moreover I verified that blob didn’t use any image antialiasing, and set
its GL quality mode to performance).