Message boards :
Number crunching :
Help with config to run multiple gpu wu in non-lunatics installation
Message board moderation
Author | Message |
---|---|
Ray Stone Send message Joined: 13 Dec 12 Posts: 6 Credit: 24,847,186 RAC: 5 |
Can someone help. I'm trying to configure my gpu to run multiple SETI WU on a non-lunatics installation. I realize this is not a new question but after hours looking through crunching posts, I can't seem to find the answer that works. Have tried configuration via app_config.xml (see below) with no change (only 1 WU runs on GPU). I suspect The name I've given in the app_config may be wrong (setiathome_v7) or I have to have an additional/different file?? Is the mapping of names shown in Boinc Manager to names in app_config defined anywhere? I've been receiving SETI@home v7 7.00 (cuda50) work units (name that shows in BOINC Manager) Its got to be a Seti configuration issue because I've configured Einstein for 2 simultaneous WU [cuda32]). I wont be changing hardware anytime soon so would like to make what I have work... additional hardware expenditure prohibited as a condition of marriage Boinc Ver 7.4.36 (X64) Seti computer info: 7433523 The following is my machine description from the event log 3/1/2015 2:08:04 PM | | CUDA: NVIDIA GPU 0: GeForce GTX 660 Ti (driver version 347.52, CUDA version 7.0, compute capability 3.0, 3072MB, 2753MB available, 2810 GFLOPS peak) 3/1/2015 2:08:04 PM | | OpenCL: NVIDIA GPU 0: GeForce GTX 660 Ti (driver version 347.52, device version OpenCL 1.1 CUDA, 3072MB, 2753MB available, 2810 GFLOPS peak) 3/1/2015 2:08:04 PM | | Host name: (omitted) 3/1/2015 2:08:04 PM | | Processor: 6 AuthenticAMD AMD FX(tm)-6300 Six-Core Processor [Family 21 Model 2 Stepping 0] 3/1/2015 2:08:04 PM | | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 htt pni ssse3 fma cx16 sse4_1 sse4_2 popcnt aes f16c syscall nx lm avx svm sse4a osvw ibs xop skinit wdt lwp fma4 tce tbm topx page1gb rdtscp bmi1 3/1/2015 2:08:04 PM | | OS: Microsoft Windows 7: Ultimate x64 Edition, Service Pack 1, (06.01.7601.00) 3/1/2015 2:08:04 PM | | Memory: 15.95 GB physical, 31.90 GB virtual 3/1/2015 2:08:04 PM | | Disk: 1.82 TB total, 1.48 TB free 3/1/2015 2:08:04 PM | | Local time is UTC -8 hours 3/1/2015 2:08:04 PM | SETI@home | Found app_config.xml <app_config> <app> <name>setiathome_v7</name> <gpu_versions> <gpu_usage>0.5</gpu_usage> <cpu_usage>0.2</cpu_usage> </gpu_versions> </app> </app_config> |
Bill Butler Send message Joined: 26 Aug 03 Posts: 101 Credit: 4,270,697 RAC: 0 |
I am not expert at this but here is some basic troubleshooting. 1. Check that you actually have some GPU work units in your queue. The system has been intermittent. For example, I have cruched through all my WU's, am dead in the water, waiting for the system to give me some more work. You may have nothing to do. 2. In BOINC manager, check under the Activity tab. Make sure "Suspend GPU" is NOT checked. 3. Regarding app_config.xml. 3A. Add <max_concurrent>N, i.e. let your system have a maximum of N simultaneous tasks. 3B. I am going to temporariy boost the number of GPU tasks. You can back off again later, if this works. So..., your app_config.xml file will look like this. <app_config> <app> <name>setiathome_v7</name> <max_concurrent>16</max_concurrent> <gpu_versions> <gpu_usage>.25</gpu_usage> <cpu_usage>.04</cpu_usage> </gpu_versions> </app> </app_config> Finally, to turn on app.config.xml without a reboot, on BOINC Manager menu, pull down Advanced tab and click on "Read config files". Look in the log file (Advanced tab, Event Log...) and see if BOINC actually read the app_config.xml file. (I think you are OK on this because the log you posted found it.) This much I clearly understand, ... additional hardware expenditure prohibited as a condition of marriage "It is often darkest just before it turns completely black." |
rob smith Send message Joined: 7 Mar 03 Posts: 22228 Credit: 416,307,556 RAC: 380 |
First, on the computer concerned using BOINC manager, advanced view/tools set "on multi-processor ssytems use at most" to 87% (assuming its your 6-core AMD) - this reserves one CPU core to feed the GPUs, which is essential on an AMD. Then create an app_config.xml file in the right place Here's an app_config.xml that will work: <app_config> Use a TEXT editor like Notepad and copy the quoted text into a new text file. Save as app_config.xml in your BOINC/projects/seti folder (which by default is C:\ProgramData\BOINC\projects\setiathome.berkeley.edu ) Going through the lines: <gpu_usage>.5</gpu_usage> - what fraction of the GPU each task is to take <cpu_usage>.075</cpu_usage> - how much CPU is set aside for the GPU When commissioning my phenom with a GTX760 I did some timings: Single shortie - 6-7 minutes two shorties at once - 8-10 minutes each three shorties at once 12-14 minutes each four shorties at once >30 minutes and system very unresponsive (This was before I ran the Lunatics, so you can use these figures as a good guide for your system) Bob Smith Member of Seti PIPPS (Pluto is a Planet Protest Society) Somewhere in the (un)known Universe? |
Ray Stone Send message Joined: 13 Dec 12 Posts: 6 Credit: 24,847,186 RAC: 5 |
setup app_config.xml as suggested. Still only one task at a time. I have about 50 SETI@home v7 7.00 (cuda50) taskes in ready to start state. Turned on coproc_debug and can now see in event log the following which leads me to believe that "setiathome_v7" is wrong value in Name element of xml. Can you confirm the name value? According to values you suggested should be looking for 0.25gpu not 1.0gpu 3/1/2015 9:46:35 PM | SETI@home | Found app_config.xml 3/1/2015 9:46:36 PM | SETI@home | [coproc] NVIDIA instance 0; 1.000000 pending for 27jl12ab.32488.2112.438086664206.12.109_1 3/1/2015 9:46:36 PM | SETI@home | [coproc] NVIDIA instance 0: confirming 1.000000 instance for 27jl12ab.32488.2112.438086664206.12.109_1 3/1/2015 9:46:36 PM | SETI@home | [coproc] Insufficient NVIDIA for 27jl12ab.32488.2112.438086664206.12.148_0; need 1, available 0 3/1/2015 9:47:36 PM | SETI@home | [coproc] NVIDIA instance 0; 1.000000 pending for 27jl12ab.32488.2112.438086664206.12.109_1 3/1/2015 9:47:36 PM | SETI@home | [coproc] NVIDIA instance 0: confirming 1.000000 instance for 27jl12ab.32488.2112.438086664206.12.109_1 3/1/2015 9:47:36 PM | SETI@home | [coproc] Insufficient NVIDIA for 27jl12ab.32488.2112.438086664206.12.148_0; need 1, available 0 |
rob smith Send message Joined: 7 Mar 03 Posts: 22228 Credit: 416,307,556 RAC: 380 |
Ray - have a look at the version I posted - it was copied directly from my Phenom/GTX760 system which is very similar to yours. Take particular note of the timings (and the RAC of about 17,000 - which is inflated by about 10% by the use of optimised apps). Bob Smith Member of Seti PIPPS (Pluto is a Planet Protest Society) Somewhere in the (un)known Universe? |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
Have you suspended Einstein and/or all other projects? |
BilBg Send message Joined: 27 May 07 Posts: 3720 Credit: 9,385,827 RAC: 0 |
Computer: http://setiathome.berkeley.edu/show_host_detail.php?hostid=7433523 ... which leads me to believe that "setiathome_v7" is wrong value in Name element of xml. Can you confirm the name value? <name>setiathome_v7</name> is OK: http://setiathome.berkeley.edu/forum_thread.php?id=76754&postid=1644471#1644471 According to values you suggested should be looking for 0.25gpu not 1.0gpu "should be looking for 0.5gpu" Do you save app_config.xml as ANSI? Try log flag <unparsed_xml> Â - ALF - "Find out what you don't do well ..... then don't do it!" :) Â |
BilBg Send message Joined: 27 May 07 Posts: 3720 Credit: 9,385,827 RAC: 0 |
Have you suspended Einstein and/or all other projects? You suppose that maybe one SETI@home + one Einstein task run on the GPU but this seems not to be the case. His log say that SETI@home task 'need 1' GPU (not 0.5): 3/1/2015 9:47:36 PM | SETI@home | [coproc] NVIDIA instance 0: confirming 1.000000 instance for 27jl12ab.32488.2112.438086664206.12.109_1 3/1/2015 9:47:36 PM | SETI@home | [coproc] Insufficient NVIDIA for 27jl12ab.32488.2112.438086664206.12.148_0; need 1, available 0 Â Â - ALF - "Find out what you don't do well ..... then don't do it!" :) Â |
Ray Stone Send message Joined: 13 Dec 12 Posts: 6 Credit: 24,847,186 RAC: 5 |
I reread all the posts and noticed the comment on saving in ansi. Since its my dev box, I'd edited and saved the file using Visual Studio 2013 and VS saved the file in UTF8. When I changed to ansi, everything started to work... Thanks to BilBg for the tip 3/4/2015 6:31:15 AM | SETI@home | [coproc] NVIDIA instance 0; 0.500000 pending for 02fe13ac.25297.10701.438086664198.12.163_0 3/4/2015 6:31:15 AM | SETI@home | [coproc] NVIDIA instance 0; 0.500000 pending for 02fe13ac.25297.10701.438086664198.12.164_1 3/4/2015 6:31:15 AM | SETI@home | [coproc] NVIDIA instance 0: confirming 0.500000 instance for 02fe13ac.25297.10701.438086664198.12.163_0 3/4/2015 6:31:15 AM | SETI@home | [coproc] NVIDIA instance 0: confirming 0.500000 instance for 02fe13ac.25297.10701.438086664198.12.164_1 3/4/2015 6:32:03 AM | SETI@home | [coproc] NVIDIA instance 0; 0.500000 pending for 02fe13ac.25297.10701.438086664198.12.163_0 3/4/2015 6:32:03 AM | SETI@home | [coproc] NVIDIA instance 0; 0.500000 pending for 02fe13ac.25297.10701.438086664198.12.164_1 3/4/2015 6:32:03 AM | SETI@home | [coproc] NVIDIA instance 0: confirming 0.500000 instance for 02fe13ac.25297.10701.438086664198.12.163_0 3/4/2015 6:32:03 AM | SETI@home | [coproc] NVIDIA instance 0: confirming 0.500000 instance for 02fe13ac.25297.10701.438086664198.12.164_1 3/4/2015 6:33:03 AM | SETI@home | [coproc] NVIDIA instance 0; 0.500000 pending for 02fe13ac.25297.10701.438086664198.12.163_0 3/4/2015 6:33:03 AM | SETI@home | [coproc] NVIDIA instance 0; 0.500000 pending for 02fe13ac.25297.10701.438086664198.12.164_1 3/4/2015 6:33:03 AM | SETI@home | [coproc] NVIDIA instance 0: confirming 0.500000 instance for 02fe13ac.25297.10701.438086664198.12.163_0 3/4/2015 6:33:03 AM | SETI@home | [coproc] NVIDIA instance 0: confirming 0.500000 instance for 02fe13ac.25297.10701.438086664198.12.164_1 3/4/2015 6:34:03 AM | SETI@home | [coproc] NVIDIA instance 0; 0.500000 pending for 02fe13ac.25297.10701.438086664198.12.163_0 3/4/2015 6:34:03 AM | SETI@home | [coproc] NVIDIA instance 0; 0.500000 pending for 02fe13ac.25297.10701.438086664198.12.164_1 3/4/2015 6:34:03 AM | SETI@home | [coproc] NVIDIA instance 0: confirming 0.500000 instance for 02fe13ac.25297.10701.438086664198.12.163_0 3/4/2015 6:34:03 AM | SETI@home | [coproc] NVIDIA instance 0: confirming 0.500000 instance for 02fe13ac.25297.10701.438086664198.12.164_1 |
BilBg Send message Joined: 27 May 07 Posts: 3720 Credit: 9,385,827 RAC: 0 |
VS saved the file in UTF8. When I changed to ansi, everything started to work... Pity that BOINC do not give any warning in such cases. It say "Found app_config.xml" but probably see the file as empty. I wonder if log flag <unparsed_xml> (or some other log flag?) will make BOINC give some warning on UTF8 file? (e.g. "app_config.xml do not configure any apps (file empty or in wrong format)" or "[unparsed_xml] п»ї") I use old BOINC which do not support app_config.xml, somebody with current BOINC have to test.  - ALF - "Find out what you don't do well ..... then don't do it!" :)  |
Ray Stone Send message Joined: 13 Dec 12 Posts: 6 Credit: 24,847,186 RAC: 5 |
I wonder if log flag <unparsed_xml> (or some other log flag?) will make BOINC give some warning on UTF8 file? I tried both ansi and utf8 on another machine with unparsed_xml turned on. in both cases log simply reports found app_config.xml - with the ansi version working and the utf8 version not working. If BOINC mgr is changed, it probably should report what it found in the log - its rather useless for debugging this scenario. I suppose it would catch mal-formed xml but that's one reason I typically use VS for xml editing. sigh... I noticed also that with 2 seti tasks running on GPU, it reports 0.15 or 0.27 cpu (values observed on 2 different machines - not the 0.075 value in xml which was the same on both) and 1 GPU each (not 0.5 as is in xml). Einsten on the other hand reports exactly what I'd configured. I played around for a while with various values but I couldn't find any way to get the manager to report any other values via app_config. I don't care -- it works now. |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
Going to go out on a wild guess. You changed the app_config.xml and then had Boinc read the Config files without turning off BOINC or restarting the computer? Sometimes when you do that, it doesn't change the display in the Task Manager window. So it could be that you are really running 0.075 CPU + 0.5 GPU. Easiest way to tell is to just suspend BOINC, exit it and then relaunch it and restart crunching and see if that corrects the display. Rarely do you need to reboot the computer. Either way, you got it to work so that's the main thing. You might consider giving the work units more CPU usage. That's an AMD chip in the original computer you mentioned. They tend to need more CPU than the Intel computers. I would probably recommend 0.35 CPU Happy Crunching.. Zalster |
BilBg Send message Joined: 27 May 07 Posts: 3720 Credit: 9,385,827 RAC: 0 |
I would probably recommend 0.35 CPU With 2 task per GPU the "0.35 CPU" will not do anything (BOINC can't 'reserve' part of a core: 2 * 0.35 = 0.7 < 1 ) "0.5 CPU" will free one CPU core (one less CPU task) "0.99 CPU" will free one CPU core (Truncate(1.98) = 1) "1.0 CPU" will free two CPU cores  - ALF - "Find out what you don't do well ..... then don't do it!" :)  |
©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.