Message boards :
Number crunching :
Setting is for only GPU tasks, am getting CPU tasks
Message board moderation
Author | Message |
---|---|
Send message Joined: 12 Apr 08 Posts: 621 Credit: 161,934,067 RAC: 0 |
Why this is happening I do not know... but all of a sudden I am getting stray MW CPU tasks on at least one system even when it is set to only do GPU tasks (I have both ATI and Nvidia equipped systems)... See this task as an example ... my stupid system did ask for work for the CPU, but the server should have ignored the request because of the server side settings ... or ... Did the database access get turned off for misc scheduler requests? Only thing I can think of ... |
Send message Joined: 31 Jan 08 Posts: 5 Credit: 17,641,479 RAC: 0 |
same you, my settings should be ignore CPU, for use Nvidia GPU, but no: see this: 28/02/2010 17:13:22 Starting BOINC client version 6.10.17 for windows_x86_64 28/02/2010 17:13:22 log flags: file_xfer, sched_ops, task 28/02/2010 17:13:22 Libraries: libcurl/7.19.4 OpenSSL/0.9.8k zlib/1.2.3 28/02/2010 17:13:22 Data directory: D:\BOINCx64 28/02/2010 17:13:22 Running under account Fabien-BOINC 28/02/2010 17:13:22 Processor: 4 GenuineIntel Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz [Intel64 Family 6 Model 15 Stepping 11] 28/02/2010 17:13:22 Processor: 4.00 MB cache 28/02/2010 17:13:22 Processor features: fpu tsc pae nx sse sse2 pni 28/02/2010 17:13:22 OS: Microsoft Windows Vista: Business x64 Edition, Service Pack 2, (06.00.6002.00) 28/02/2010 17:13:22 Memory: 7.00 GB physical, 10.92 GB virtual 28/02/2010 17:13:22 Disk: 110.49 GB total, 23.77 GB free 28/02/2010 17:13:22 Local time is UTC +1 hours 28/02/2010 17:13:22 NVIDIA GPU 0: GeForce GTX 260 (driver version 19621, CUDA version 3000, compute capability 1.3, 896MB, 477 GFLOPS peak) 28/02/2010 17:13:22 Milkyway@home Found app_info.xml; using anonymous platform 28/02/2010 17:13:22 Not using a proxy 28/02/2010 17:13:22 Version change (6.10.18 -> 6.10.17) 28/02/2010 17:13:22 Milkyway@home URL http://milkyway.cs.rpi.edu/milkyway/; Computer ID 23238; resource share 100 28/02/2010 17:13:22 Reading preferences override file 28/02/2010 17:13:22 Preferences limit memory usage when active to 3583.13MB 28/02/2010 17:13:22 Preferences limit memory usage when idle to 5374.69MB 28/02/2010 17:13:22 Preferences limit disk usage to 10.00GB 28/02/2010 17:13:22 Running CPU benchmarks 28/02/2010 17:13:22 Suspending computation - running CPU benchmarks 28/02/2010 17:13:54 Benchmark results: 28/02/2010 17:13:54 Number of CPUs: 4 28/02/2010 17:13:54 2271 floating point MIPS (Whetstone) per CPU 28/02/2010 17:13:54 7193 integer MIPS (Dhrystone) per CPU 28/02/2010 17:14:02 Milkyway@home update requested by user 28/02/2010 17:14:03 Milkyway@home Sending scheduler request: Requested by user. 28/02/2010 17:14:03 Milkyway@home Requesting new tasks for CPU 28/02/2010 17:14:08 Milkyway@home Scheduler request completed: got 0 new tasks 28/02/2010 17:14:08 Milkyway@home Message from server: No work sent 28/02/2010 17:15:09 Milkyway@home Sending scheduler request: To fetch work. 28/02/2010 17:15:09 Milkyway@home Requesting new tasks for CPU 28/02/2010 17:15:14 Milkyway@home Scheduler request completed: got 0 new tasks ................... |
Send message Joined: 12 Apr 08 Posts: 621 Credit: 161,934,067 RAC: 0 |
Um, as long as you don't get tasks issued you are fine. Because of the bonehead way they slipped in new features the system blindly asks for work that it should not be allowed to run. And it will ask for GPU work from every project over and over again even when we all know that some projects will never have GPU work. So, this is a server side issue, I ask for work that I shouldn't because of the bonehead client, but the server should not send me tasks because of my server side setting. Unfortunately, there is an error someplace and I am still, on occasion, getting MW CPU side tasks. |
Send message Joined: 12 Nov 07 Posts: 2425 Credit: 524,164 RAC: 0 |
Unfortunately, there is an error someplace and I am still, on occasion, getting MW CPU side tasks. They are the same tasks and can be run on either cpu or gpu. Doesn't expecting the unexpected make the unexpected the expected? If it makes sense, DON'T do it. |
Send message Joined: 31 Jan 08 Posts: 5 Credit: 17,641,479 RAC: 0 |
i have 2 PC, on the other i have a HD4770, and i receive many wus for this GPU. the settings are same. |
Send message Joined: 12 Apr 08 Posts: 621 Credit: 161,934,067 RAC: 0 |
Unfortunately, there is an error someplace and I am still, on occasion, getting MW CPU side tasks. I know the tasks are the same. But, I am not interested in using my CPU time on MW or Collatz in that I can get far more production on the GPU. So, I have the server side settings set to no CPU tasks... yet I still see them issued against my desires... And that is the point ... There is little utility in having a setting that is ignored. Even if the ignoring happens only on occasion. |
Send message Joined: 6 Apr 08 Posts: 2018 Credit: 100,142,856 RAC: 0 |
Unfortunately, there is an error someplace and I am still, on occasion, getting MW CPU side tasks. That occasion could be costly if the unwanted CPU MW/Collatz WUs jump in and stop other WUs from completing on time and result in wasted work with loss of credit. I too would rather have control of what's going on and prefer not to see the random activity that can take place in BOINC manager. |
Send message Joined: 12 Apr 08 Posts: 621 Credit: 161,934,067 RAC: 0 |
That occasion could be costly if the unwanted CPU MW/Collatz WUs jump in and stop other WUs from completing on time and result in wasted work with loss of credit. I too would rather have control of what's going on and prefer not to see the random activity that can take place in BOINC manager. Not so much of an issue for me... I use 1.5 day cache and have fast enough systems that I am not skirting the edge on that ... but, still, I want to have my CPU go to the other 50 projects out there that don't have GPU applications. |
Send message Joined: 18 Feb 09 Posts: 158 Credit: 110,699,054 RAC: 0 |
This has been happening to me off and on for the last few weeks. Its quite annoying. Its not that I don't think the science is worthwhile on MW, I just don't want to waste 10 hours on a Wu that i could have done in 90 seconds. I've got WCG medical research running on my CPU, and I really don't like MW butting into that. It's only been MW that has done this to me thus far. They seem to come in spurts, like I got three of them today. :p |
Send message Joined: 27 Dec 07 Posts: 35 Credit: 1,432,926 RAC: 0 |
That occasion could be costly if the unwanted CPU MW/Collatz WUs jump in and stop other WUs from completing on time and result in wasted work with loss of credit. I too would rather have control of what's going on and prefer not to see the random activity that can take place in BOINC manager. There is no RANDOM ACTIVITY in the BOINC Manager, it goes by the resource share you give each program. Sometimes it looks random when one program gets more than its share of time one day and less another day but it all works out to what your setting are in the long run. |
Send message Joined: 22 Apr 09 Posts: 38 Credit: 27,377,932 RAC: 0 |
Hi, I'm quite new here and my WU's are done by CPU, even tried an SSE3 optimized app., but can't see much difference. :( And the host with an ATI HD5770, now doing Collatz C., doesn't work because it's single precision, bummer. Also have a QX9650 & 9800GTX+ & GTS250, running XP64, but isn't supported, by default??? Seeing the RAC, some have, there has to be an NVidida (Optimized) app. , too? |
Send message Joined: 12 Apr 08 Posts: 621 Credit: 161,934,067 RAC: 0 |
Seeing the RAC, some have, there has to be an NVidida (Optimized) app. , too? The GPU applications run quite well ... slightly better (in my case) on ATI cards than on Nvidia (CUDA)... time for a task on CUDA is above 10 minutes each and on my less expensive ATI cards the time is as low as 1.5 minutes, 3 minuts and change on my slower ATI cards... Collatz scales about the same way. The main advantage of CUDA is that it is more widely supported by projects with PG, Einstein, GPU Grid, and SaH (and a couple of others) having CUDA apps and no ATI application in production ... there are efforts to make ATI applications for several of these projects but it is slow going ... I think the furthest along may be GPU Grid (for what that is worth) ... |
©2024 Astroinformatics Group