4-Way Nouveau Benchmarks On Linux 3.12, Mesa 9.3

[phoronix]

Phoronix: 4-Way Nouveau Benchmarks On Linux 3.12, Mesa 9.3

Earlier this week on Phoronix were the results of an 11-way AMD Radeon GPU comparison using the Linux 3.12 kernel and Mesa 9.3-devel. Coming out today are the results for a four-way NVIDIA GeForce graphics card comparison using the Nouveau driver on Linux 3.12 / Mesa 9.3-devel. Sadly, this comparison is smaller than the open-source Radeon comparison since the reverse-engineered Nouveau driver was very buggy in its current configuration.

http://www.phoronix.com/vr.php?view=19147

[Dukenukemx]

How bad is this re-clocking issue to resolve? This issue has been going on for a while now.

[schmidtbag]

How bad is this re-clocking issue to resolve? This issue has been going on for a while now.

well it took AMD's own developers a relatively long time to fix it on their own hardware. I don't expect a bunch of volunteers who have to reverse-engineer everything to accomplish it any time soon.


Personally what I would find interesting is how the tests would compare to the proprietary drivers if the GPUs were set to their lowest clock setting. That would get us a good impression on how close the nouveau devs are in terms of performance.

[UraniumDeer]

well it took AMD's own developers a relatively long time to fix it on their own hardware. I don't expect a bunch of volunteers who have to reverse-engineer everything to accomplish it any time soon.


Personally what I would find interesting is how the tests would compare to the proprietary drivers if the GPUs were set to their lowest clock setting. That would get us a good impression on how close the nouveau devs are in terms of performance.

Michael actually did that last year, showing that in most cases the Nouveau-driver matched the proprietary driver clock for clock.
I believe this is one of many reasons that some of the first documentation that Nouveau-devs asked for after Ritgers announcement, was power management and clock, but the reply was that that was probably some of the 'more difficult areas' for them to address.

Edit:
It would be interesting with a new one though
Here's the article:

http://www.phoronix.com/scan.php?page=article&item=nvidia_nouveau_clocks&num=1

[schmidtbag]

Michael actually did that last year, showing that in most cases the Nouveau-driver matched the proprietary driver clock for clock.
I believe this is one of many reasons that some of the first documentation that Nouveau-devs asked for after Ritgers announcement, was power management and clock, but the reply was that that was probably some of the 'more difficult areas' for them to address.

Edit:
It would be interesting with a new one though
Here's the article:
http://www.phoronix.com/scan.php?pag...u_clocks&num=1

Ah yes, I forgot about that one. But it'd be nice to see if newer hardware has gained any advantages.