Questions and Answers :
GPU applications :
how do i get more cuda work units?
Message board moderation
Author | Message |
---|---|
cheifgaffer Send message Joined: 13 Mar 02 Posts: 1 Credit: 140,975 RAC: 0 |
my systems keep running out of cuda work units & when requesting an update nothing happens. it just says " requesting 0 seconds of work, reporting 0 completed tasks" does my system have to wait for the Astropulse work units to finish before it will be allocated more cuda work units? hope someone can shed more light! :) |
john deneer Send message Joined: 16 Nov 06 Posts: 331 Credit: 20,996,606 RAC: 0 |
does my system have to wait for the Astropulse work units to finish before it will be allocated more cuda work units? I suspect that you have only a very small amount of extra work set in your preferences, and that therefore Boinc will indeed not ask for extra work before the astropulse units have almost finished. You could try to set a larger cache (extra work setting in your local preferences). If the cache setting is larger than the time needed for the astropulse units, boinc should start asking for new work. And that would then most likely be cuda type units. Regards, John. |
Dawn Raison Send message Joined: 6 Jun 99 Posts: 3 Credit: 4,317,068 RAC: 0 |
I see similar on occasion - it is frustrating to know that there are many GPU cores laying idle just because the scheduler wont issue them more work... Try suspending the astropulse WU's, wait a few mins and the try to force an update. |
BigWaveSurfer Send message Joined: 29 Nov 01 Posts: 186 Credit: 36,311,381 RAC: 141 |
In short, I have to 'abort' all of but one of my AP files, then it will download a TON of CUDA files. I 'think' I have found a reason, and assume they are working on a fix for it.... The 'to completion' is messing EVERYTHING up! My AP usually finishes in around 75 hours, yet it has an estimated time of triple that. Now, when I am crunching 'normal' CUDA units that take about 30 minutes my system seem to continue to download new WU's with no problem......However..... when CUDA is crunching the low angle CUDA units that take close to 2 hours suddenly my AP 'to completion' time shoots through the roof, over 400 hours!!!! I have mine set to keep 10 days of extra work, but when I have 4 AP units and the system thinks it will take almost 2 months to finish nothing downloads. Like I said above, I 'abort' all but one AP unit and watch the flood of new CUDA units download. Seems to be a manual fix for the problem, hopefully they are working on a fix for the software. |
Matthew Heys Send message Joined: 6 Apr 00 Posts: 1 Credit: 24,392 RAC: 0 |
I'm getting the same issue and I think BigWaveSurfer has hit the nail on the head, it's estimating that an AstroPulse is going to take 16 days and it's trying to do two of them, so even though I have 10 days of work set-up it won't download anything. I had to abort the AstroPulses and hit update :-( it would be nice to have two separate counters, one for GPU and one for CPU |
Sterling_Aug Send message Joined: 27 Sep 02 Posts: 54 Credit: 14,105,725 RAC: 0 |
Another nice feature to add to the BOINC wishlist would be: Now that we have CUDA work, the limit of 500 WUs per day needs to be increased to 2,000 or maybe 5,000 WUs. |
Fred W Send message Joined: 13 Jun 99 Posts: 2524 Credit: 11,954,210 RAC: 0 |
Another nice feature to add to the BOINC wishlist would be: Why?? In fact the limit is not 500 per day - it is 100 per day per CPU core + 400 per day per GPU. My overclocked quaddie and top-end vid card never get anywhere near crunching the WU's that are available. F. |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.