Message boards :
Number crunching :
Why is BOINC not using both of my Nvidia GPU's?
Message board moderation
Author | Message |
---|---|
blueviking Send message Joined: 14 Aug 01 Posts: 6 Credit: 114,409,447 RAC: 20 |
Ever since the "reboot" a few days ago, when SAH came back online again after the week or so of absence, my BOINC manager shows me that only one of my GPU's is being used. This is the message when I start up the BOINC manager: NVIDIA GPU 0: GeForce GTX 260 (driver version 306.97, CUDA version 5.0, compute capability 1.3, 896MB, 8381200MB available, 875 GFLOPS peak) NVIDIA GPU 1 (not used): GeForce GTX 260 (driver version 306.97, CUDA version 5.0, compute capability 1.3, 896MB, 802MB available, 875 GFLOPS peak) So, for some reason it appears that my second video card is not being used. I don't think I had a problem BEFORE the outage a week ago, but I can't be certain. Is it possible that something happened with BOINC, or the SETI servers, during the outage, so that my second GPU is not being used now? Or, maybe it's a problem with one of my video cards? I have re-booted a few times, and re-started BOINC manager a few times to see if it would fix the problem, but to no avail. Any help will be appreciated. |
Jord Send message Joined: 9 Jun 99 Posts: 15184 Credit: 4,362,181 RAC: 3 |
BOINC uses only the best GPU that it detects, and it determines which is best on the following factors, in decreasing priority: 1 - compute capability 2 - software version 3 - available memory 4 - speed You may have two GTX 260s there, but the first is being detected as having 8381200MB of memory, whereas the second only has 896MB. So the first one is definitely superior. ;-) You've been bitten by the memory detection bug that's in a lot of BOINC versions. The first reliably working fix for this bug is in 7.0.36 or .38 (I forget which). So if you don't mind running a beta BOINC, you can upgrade to the newest around, 7.0.40. A change log for it and download links to it can be found in post 46649 of the BOINC 7 Change Log Thread. |
Bill G Send message Joined: 1 Jun 01 Posts: 1282 Credit: 187,688,550 RAC: 182 |
From older posts that I used to correct this on my computer: It's just a line to be added to your cc_config file. 1. exit BOINC 2. go to ProgramData\Boinc directory 3. find cc_config file and edit it with notepad 4. find at the end line use_all_gpus>0</use_all_gpus> and change it to use_all_gpus>1</use_all_gpus> If the option is not there, add it. 5. save it If you do not have a cc_config file, you will have to create it. If you have to create it you need to add the following exactly as shown: <cc_config> <options> <use_all_gpus>1</use_all_gpus> </options> </cc_config> If you will notice, the two cards are not recognized at being identical, hence the problem. SETI@home classic workunits 4,019 SETI@home classic CPU time 34,348 hours |
blueviking Send message Joined: 14 Aug 01 Posts: 6 Credit: 114,409,447 RAC: 20 |
I did this, and now it appears to be working (fine?). It recognizes both GPU's and gave them both a CUDA task to process. The odd thing is that these are the same two, identical video cards that I've been running in my machine for almost 3 years now, and I don't think I've ever encountered this problem in the past... and obviously something is wrong with BOINC, because it's reading that first card as having over 8 Terabytes of memory available (wouldn't it be great to have a video card with TERAbytes of memory?). Anyway, thanks for the fix! From older posts that I used to correct this on my computer: |
©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.