Perseus Arm Survey - very low performance with new nvidia drivers

AgentB
AgentB
Joined: 17 Mar 12
Posts: 915
Credit: 513211304
RAC: 0

RE: Thanks for testing,

Quote:

Thanks for testing, standing by for results :-).


My gtx460 results remain the same (compared with 325.15). No error / invalids / power and temps the same.

Quote:

This could have some important implications for E@H: If this new driver (when finally released) should really work well with our code, we would finally be in a position to provide BRP apps that are compiled against CUDA 5.5 runtine and linked with the CUDA 5.5 FFT lib, something that is high on our wish list. In theory this should also help performance.

Soon time to visit Albert at home!

Will CUDA 5.5 lose or gain any nVidia cards?

ExtraTerrestrial Apes
ExtraTerrestria...
Joined: 10 Nov 04
Posts: 770
Credit: 580235504
RAC: 141806

RE: Looking at the recent

Quote:

Looking at the recent phoronix benchmark displaying compute performance of AMD versus NVIDIA, I would highly question anyone who states they are getting good performance with NVIDIA on Linux. Seems only AMD GPUs benefit from Linux right now.

http://www.phoronix.com/scan.php?page=article&item=nvidia_geforce_gtx650&num=8


In addition to what Bernd said: Luxmark has always favored AMDs greatly, be it for hardware or OpenCL reasons. This has nothing to do with Linux and not much with other programs.

MrS

Scanning for our furry friends since Jan 2002

Anonymous

I have installed the 330.13

I have installed the 330.13 drivers on Ubuntu 12.04, kernel: 3.5.0-18-generic, with a NVIDIA 650Ti.

I installed from a ppa and not by downloading the "run" file from NVIDIA.

How does one determine the performance with the new drivers? Do you compare the run-time/cpu-time columns of "valid" tasks under the old driver against these same values under the new drivers? Or is there a better way?

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2143
Credit: 2967218693
RAC: 701693

RE: How does one determine

Quote:
How does one determine the performance with the new drivers? Do you compare the run-time/cpu-time columns of "valid" tasks under the old driver against these same values under the new drivers? Or is there a better way?


I think comparing the timings is the best (most reliable and accurate) way there is.

People talk about RAC, but that is so slow to stabilise and subject to so many external influences (such as time before validation) that it's hard to draw conclusions.

ExtraTerrestrial Apes
ExtraTerrestria...
Joined: 10 Nov 04
Posts: 770
Credit: 580235504
RAC: 141806

RE: How does one determine

Quote:
How does one determine the performance with the new drivers? Do you compare the run-time/cpu-time columns of "valid" tasks under the old driver against these same values under the new drivers? Or is there a better way?


If you compare WUs which get the same amount of credits this is valid. If you're looking for small changes aveage over 10 - 20 WUs.

If the credits and amount of work per WU vary it gets more complicated. In this case I always suggest to calculate a "projected RAC", which are the credits per day which you'd get if you could crunch undisturbed. But this shouldn't be neccessary for Einstein GPU WUs.

MrS

Scanning for our furry friends since Jan 2002

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.