Looks like these are going to run very very long, I hope they produce good results and award mega points as well :-)

GTX 480 WinXP32, BOINC 7.0.28, only 3% complete after 35 minutes. GPU utilization 94% and MEM usage very low at 7%-8% so it looks like it is running as expected. I'll post back when it's done some time tomorrow :thumbsup:
____________
Thanks - Steve

Credits for them are written there:

http://www.gpugrid.net/forum_thread.php?id=3086#26403
____________
DSKAG Austria Research Team: http://www.research.dskag.at

Looks like these are going to run very very long, I hope they produce good results and award mega points as well :-)

GTX 480 WinXP32, BOINC 7.0.28, only 3% complete after 35 minutes. GPU utilization 94% and MEM usage very low at 7%-8% so it looks like it is running as expected. I'll post back when it's done some time tomorrow :thumbsup:

One of this very long Noelia workunits finished on one of my GTX 480@800MHz in 10 hours and 55 minutes. I've received 98100 credits for it. Another one is running on my GTX 590@625MHz for 7 hours and 27 minutes, it's progress is at 65.154%, it will finish in about 11 hours and 30 minutes. A third one is running on my GTX 580@825MHz, it took 1 hour 27 minutes and 15 seconds to reach 20%, so it will finish in 7 hours and 16 minutes (probably more, because these very long Noelia tasks tend to slow down as they progress). They run at 85% GPU usage.

Ugh.

W7 w/ 680 is running at 77%

Any way we can increase the GPU usage on these tasks. That's pretty low.

Or at least tweak future tasks?

What is Aero? How do you disable it? Rick




____________

Thats the animated gui in win7 ;)
____________
DSKAG Austria Research Team: http://www.research.dskag.at

And after runing one of these WUs on my GTX 670 Win7 x64 I see that the runtime is much longer than PAOLA WUs but awards much less points (super rough guestimate is that we should be getting about 120K per WU instead of 68K). While we all have our reasons for crunching and I'd like to think I lean towards the altruistic side of the conversation net, points don't cost the project anything so I don't see why you would not align them with the others?
____________
Thanks - Steve

I'll try to remember to provide better details in the future :-)
Both systems are optimized for GPUGrid crunching with 1 thread free, no systems changes were made between the runs being compared.

WinXP 32 on a GTX 480:
rundig10_run8-NOELIA_smd-0-5-RND7514
70,774.31 seconds runtime granted 67,350.00 credits

compared to:
1XFG_40_10-PAOLA_HGAbis-24-100-RND0147
24,075.72 seconds runtime granted 86,850.00 credits



Win7 x64 on a GTX 670:
rundig5_run7-NOELIA_smd2-3-5-RND5863
23,389.33 seconds runtime granted 68,100.00 credits

compared to:
1XFG_39_2-PAOLA_HGAbis-31-100-RND1613
18,260.10 seconds runtime granted 86,850.00 credits[url][/url]
____________
Thanks - Steve

Hi Snow Crash,

We looked into it on our end, checking for human error and running on several cards (one similar to yours). The running times were within 30% here, so the huge difference you're seeing is very strange. Therefore, it is perhaps something on your end. Do you use your system to crunch for other projects, in particular CPU projects? Do you use your computer while you crunch? The NOELIA_smd work units require more CPU usage, and if the resources aren't there or are being used by another project/application, it could explain the slowdown.

Nate

I apologize if I came off sounding like an ascii, that was never my intention and I will endeavor to start my threads as a question rather than postulating that I know what is going on.

I've been trying to think around and around what the issue might have been.
I regularly monitor my systems to make sure I have not downclocked so I'm confidant it was not that, my BOINC config has 1 thread free, and the GPU utilization was actually fairly good. Trying to keep these things in mind along with the results of the testing you have done, and the searching I have since done on other people's results I can only *assume* that the WinXP actually processed on my GTX 295 and not my GTX 480 which would be possible if the WU was from the lower end of the processing range required crunching for Paola's experiments (I was so darn sure of myself ... open mouth, insert foot.

That being said, the Win7 system only has 1 GTX 670 so there was no card mix up on my results comparisons and that mach9ine is a headless cruncher that I don't use for anything else (well until the summer is over and I can move out of my basement and back into my study :-).

I have found other results run on GTX 680s under Win7 and roughly the points per second ratio looks to be about 3 to 1 on various WUs from these series but other WU types come in around 4 to 1 so I still think there is a substantial disparity. If that is within the acceptable range for the project then that's exactly what it is and I won't ask you to look into this any further.

Thanks again for listening and looking into this, I'm sure you had better things to do with your time. I'll try to be clearer and more thorough next time I think something may not be quite right.

Sincerely,
Steve

Ok, glad you got it figured out. I didn't see the 295 in any of your hosts, but that would make sense.

I have found other results run on GTX 680s under Win7 and roughly the points per second ratio looks to be about 3 to 1 on various WUs from these series but other WU types come in around 4 to 1 so I still think there is a substantial disparity. If that is within the acceptable range for the project then that's exactly what it is and I won't ask you to look into this any further.

WinXP 32, BOINC 7.0.28, nvidia driver 301.42, EVGA GTX 480 OCd to shaders 1512 and rock solid stable no downclocking. This was riunning with no user activity as I was at work.

Caught the WU in the screenshot taking an exceeding amount of time. All other WUs process fine, even the most demanding PAOLA_HGA WUs take less than 7 hours but this NOELIA task was at just over 9% after alomst 2.5 hours. I'll keep crunching everything else for GPUGrid but I'm going have to abort this.


____________
Thanks - Steve