Questions and Answers :
GPU applications :
Use one GPU only
Message board moderation
Author | Message |
---|---|
Aus_Karlos Send message Joined: 28 Apr 10 Posts: 16 Credit: 2,679,860 RAC: 0 |
Hey guys im having a problem with BOINC. I have two GTX280 cards installed running CUDA driver 3.0. They are currently not in SLI, however my attempts at choosing "Device 0" to fold has failed. As soon as i load up BOINC manager two GPU tasks start. I have a cc_config.xml file in C:\ProgramData\BOINC. <cc_config> <options> <ignore_cuda_dev>1</ignore_cuda_dev> <no_gpus>0</no_gpus> </options> </cc_config> Also without the "Use GPU always" checked the GPU clients wont start, even if my PC is in idle (Can tell from Everest on my G15) I want BOINC to only use Device 0 all the time but its just not working for me. Is there a command line i can use? Messages- 1/05/2010 6:52:24 PM Re-reading cc_config.xml 1/05/2010 6:52:24 PM Re-read config file 1/05/2010 6:52:24 PM Config: ignoring NVIDIA GPU 1 1/05/2010 6:52:24 PM log flags: file_xfer, sched_ops, task 1/05/2010 6:51:34 PM SETI@home Restarting task 07mr07ag.22170.481.12.10.29_0 using setiathome_enhanced version 609 1/05/2010 6:51:34 PM SETI@home Restarting task 07mr07ag.22170.481.12.10.184_1 using setiathome_enhanced version 609 Assuming that the _0/1 at the end file string relates to Device 0 and 1. |
Gundolf Jahn Send message Joined: 19 Sep 00 Posts: 3184 Credit: 446,358 RAC: 0 |
Assuming that the _0/1 at the end file string relates to Device 0 and 1. Nope. The _0/_1 suffix denote the number of the WU replication. How many tasks get started if you don't ignore any device? How many tasks get started if you ignore device 0 and device 1? Gruß, Gundolf PS: There are also some logging options for cc_config.xml, like <coproc_debug>, but they tend to fill up your log rapidly. Computer sind nicht alles im Leben. (Kleiner Scherz) SETI@home classic workunits 3,758 SETI@home classic CPU time 66,520 hours |
Geek@Play Send message Joined: 31 Jul 01 Posts: 2467 Credit: 86,146,931 RAC: 0 |
Your line............ <ignore_cuda_dev>1</ignore_cuda_dev> will ignore the second cuda device replace it................ <ignore_cuda_dev>0</ignore_cuda_dev> to ignore the first cuda device Boinc....Boinc....Boinc....Boinc.... |
BilBg Send message Joined: 27 May 07 Posts: 3720 Credit: 9,385,827 RAC: 0 |
Sometimes BOINC starts (bug?) 2 tasks on one GPU check by GPU-Z temperatures and % load for both GPUs Temperature Monitoring Programs http://setiathome.berkeley.edu/forum_thread.php?id=59292 Restart BOINC and post the first 20-30 lines of messages On the page for your computer only one GPU is shown: Coprocessors - NVIDIA GeForce GTX 280 (1005MB) driver: 19745 http://setiathome.berkeley.edu/show_host_detail.php?hostid=5388577 . Â - ALF - "Find out what you don't do well ..... then don't do it!" :) Â |
Gundolf Jahn Send message Joined: 19 Sep 00 Posts: 3184 Credit: 446,358 RAC: 0 |
On the page for your computer only one GPU is shown... That's new! At the time of my previous post, there have been two of them shown ([2]). Gruß, Gundolf |
Aus_Karlos Send message Joined: 28 Apr 10 Posts: 16 Credit: 2,679,860 RAC: 0 |
Looks like BOINC wasn't closing down completely on my system. So i did a reboot in the end and now its ignoring GPU 1. Thanks for the help guys. |
Aus_Karlos Send message Joined: 28 Apr 10 Posts: 16 Credit: 2,679,860 RAC: 0 |
Might aswell keep it in the same thread. Ive found something really interesting some of my GPU WU have begun with the flag "high Priority". Even though there is still 15days till deadline but the really interesting bit is when they enter High Priority my card completes them in under 5mins. Where as normal priority does them in 1hr30min. Noticeable side-effect is my GPU sits at 80c with fan at 70%. Where in normal mode its at 60c with 55% fan speed. To be honest 80c on a GTX280 is in the norm, how can i force all GPU tasks to start up in high priority regardless of deadline? |
Jord Send message Joined: 9 Jun 99 Posts: 15184 Credit: 4,362,181 RAC: 3 |
Looks like BOINC wasn't closing down completely on my system. See BOINC and science applications continue running when I exit BOINC Manager. |
Jord Send message Joined: 9 Jun 99 Posts: 15184 Credit: 4,362,181 RAC: 3 |
how can i force all GPU tasks to start up in high priority regardless of deadline? You don't want to do that. High Priority means that these tasks are in trouble, that they're eligible to miss their deadline. This is usually because you have too much work on your system. Seeing that you run BOINC 6.10.43, a version that's been pulled due to a severe bug in the GPU code, where BOINC will continue to download work for the GPU even while the GPU is hung or has work stuck in its memory, I think your system is overworked. When some tasks take an hour and a half and some take 5 minutes, that is because of the different angle ranges. The so called very low angle range (VLAR) and very high angle range (VHAR) tasks have problems being run on a GPU. They either finish very slowly, or they get stuck. You best uninstall this version of BOINC and go back to 6.10.18, which is the recommended version again. And stay on that one until the next version comes out, whichever one that is and whenever that is. |
Aus_Karlos Send message Joined: 28 Apr 10 Posts: 16 Credit: 2,679,860 RAC: 0 |
Reverting back to v18, the BOINC manager no longer recognises the following option <ignore_cuda_dev>1</ignore_cuda_dev> How do i tell the manger to ignore my second card? |
Jord Send message Joined: 9 Jun 99 Posts: 15184 Credit: 4,362,181 RAC: 3 |
Ah sorry, that core client configuartion option didn't become available until 6.10.19 So get that version from here: 6.10.19 (32bit version) 6.10.19 (64bit version) |
Aus_Karlos Send message Joined: 28 Apr 10 Posts: 16 Credit: 2,679,860 RAC: 0 |
Ok done. One more question, is it worth installing Lunatics optimisation? Are there risks involved from installing this. Like computation errors? |
Jord Send message Joined: 9 Jun 99 Posts: 15184 Credit: 4,362,181 RAC: 3 |
Normally there aren't any problems, but even the Lunatics app has problems with VLAR and VHAR tasks. It's not very much faster on the GPU, as far as I understand. Sometimes even slower than the stock app. You can do it, but my advise is to run your present cache empty (set Seti to No New Tasks), before you do that. Just in case something does go wrong and you lose all your present work in progress. You wouldn't be the first (and sadly also not the last). ;-) |
©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.