Is it my imagination...
27 Nov 2014 13:27:59 UTC
Topic 197830
(moderation:
Greetings,
... or is Einstein@Home a memory hog? Before running, my RAM usage is about 30%. When running SETI@Home, it jumps to about 35 to 40%. When running Einstein@Home, my RAM usage pole vaults to 70+%. Is this normal for Einstein? And, should I not worry about it if it is?
Thank you in advance! :)
Keep on BOINCing...! :)
CAPT Siran d'Vel'nahr XO
USS Vre'kasht NCC-33187
Siran's website: [ ONLINE! ]
Is it my imagination...
)
When I look at SIV to see it says that it is only using 122 MB but if you look at the stderr report it says it uses closer to 500 MB (4XXmb)
So it's not your imagination.
Some speculate that is why it is difficult to run more than 1 Work Unit per GPU as it quickly exceeds the Memory of the GPU while adding more work units.
Happy Crunching....
Zalster
RE: When I look at SIV to
)
Greetings Zalster,
Thanks! :) I won't worry about the RAM usage then.
Keep on BOINCing...! :)
CAPT Siran d'Vel'nahr XO
USS Vre'kasht NCC-33187
Siran's website: [ ONLINE! ]
RE: When I look at SIV to
)
So why does Einstein ignore my GPU utilization settings of 0.33 in project preferences and also my 0.33 GPU usage settings in app_config.xml? Even though this is what I have set, the project only will run 2 instances of the GPU apps on each card of mine.
(0.2 CPUs + 0.5 NVIDIA GPUs) is on the status line of each running E@H GPU task.
I have 4GB of RAM available on each card. SIV has shown the maximum usage on each card to only get to about 1.4GB so there is plenty of room. Both SETI and MilkyWay obey the 0.33 GPU usage settings, why won't Einstein. I have updated project settings numerous times, stopped and restarted the client many times and even looked into sched_reply_einstein.phys.uwm.edu.xml file which has this entry 0.33 and still only 2 instances of Einstein GPU tasks each will run on my 4 cards in 2 machines. The only thing I haven't done is reset the project which I have read usually does nothing. Does anyone know why this is happening?
Cheers, Keith
RE: So why does Einstein
)
An oddity of the Einstein project-preference specified GPU job multiplicity is that it only propagates to implementation on a particular host when that host successfully downloads a new WU of the type in question after the preference settings has been updated.
I don't know what to expect when you combine project preference specification with app_config.xml setting. I'll hazard a guess the app_config.xml setting attempt was faulty in some way.
Have you tried rebooting the
)
Have you tried rebooting the system? Not boinc but your computer.
I know when I changed my app_config.xml and had Boinc Manager reread the config files that it still continued to use the old settings.
I found that I needed to completely shutdown and reboot the system in order for the system to delete the old settings and accept the new parameters.
Worth a shot.
Happy Crunching...
Zalster
Yes, the systems have been
)
Yes, the systems have been shut down and restarted since I updated the preferences settings. In response to the other poster, I have copied the relevant section of the app_config files from the other projects for the GPU usage. The syntax is correct for Einstein. Also, the GPU usage section of the scheduler file shows the correct 0.33 usage. I have let both systems crunch down to zero tasks onboard several times and refilled, typically 6 tasks at a time of a mix of BRP4G and BRP5 tasks, so the system has requested new task types several times now after the upgrade to the GTX 970's. I have reread the config files many times and the response in the log file shows no errors. What I can't figure out is where the old 0.5 GPU usage setting is hiding in whatever project system file. It must be reading it in from somewhere. I've examined every XML file for Einstein in my BOINC and project directories. I can only surmise it is coming from some type of E@H server file for my account. I've suspended the project for a while and resumed thinking that might cause the project servers to reload my new GPU preferences but that didn't work. The only thing I haven't tried is to reset the project. Maybe that is the next thing to try.
Cheers, Keith
What about your cc_config?
)
What about your cc_config? if you use one. I think I would suggest posting your app_config so we can look at it and see if we can spot the problem before you reset the project. My 2 cents
Zalster
and please post your Boinc
)
and please post your Boinc startup from the Event log.
Claggy
RE: So why does Einstein
)
If you set the GPU utilization settings on the website and use an app_config.xml file, the latter will over-ride the former.
You should post your app_config.xml because the problem would seem to lie there. I can assure you from personal experience that there is no problem running many more than two concurrent GPU tasks. I think the limit is 10x, not that I'm suggesting you go there :-).
I do it for my hosts using website preferences (different venues) to get a standard 2x or 3x mode of operation and then over-ride with app_config.xml if I want a particular host to do 4x or more. If you can't get more than 2x, do you have a tag somewhere set to 2?
Cheers,
Gary.
RE: RE: So why does
)
Thanks for the info on the supersedance of the local app_config over the website preferences. I should have guessed that since I knew that local preferences overrode website settings. To answer all the suggestions of the other posters, here are the startup lines from the log and the contents of E@H app_config.xml file.
11/28/2014 11:30:00 AM | | Starting BOINC client version 7.4.27 for windows_x86_64
11/28/2014 11:30:00 AM | | log flags: file_xfer, sched_ops, task
11/28/2014 11:30:00 AM | | Libraries: libcurl/7.33.0 OpenSSL/1.0.1h zlib/1.2.8
11/28/2014 11:30:00 AM | | Data directory: C:\ProgramData\BOINC
11/28/2014 11:30:00 AM | | Running under account Keith
11/28/2014 11:30:00 AM | | CUDA: NVIDIA GPU 0: GeForce GTX 970 (driver version 344.65, CUDA version 6.5, compute capability 5.2, 4096MB, 3766MB available, 4257 GFLOPS peak)
11/28/2014 11:30:00 AM | | CUDA: NVIDIA GPU 1: GeForce GTX 970 (driver version 344.65, CUDA version 6.5, compute capability 5.2, 4096MB, 3844MB available, 4257 GFLOPS peak)
11/28/2014 11:30:00 AM | | OpenCL: NVIDIA GPU 0: GeForce GTX 970 (driver version 344.65, device version OpenCL 1.1 CUDA, 4096MB, 3766MB available, 4257 GFLOPS peak)
11/28/2014 11:30:00 AM | | OpenCL: NVIDIA GPU 1: GeForce GTX 970 (driver version 344.65, device version OpenCL 1.1 CUDA, 4096MB, 3844MB available, 4257 GFLOPS peak)
11/28/2014 11:30:00 AM | SETI@home | Found app_info.xml; using anonymous platform
11/28/2014 11:30:00 AM | | Host name: Keith-Windows7
11/28/2014 11:30:00 AM | | Processor: 8 AuthenticAMD AMD FX(tm)-8350 Eight-Core Processor [Family 21 Model 2 Stepping 0]
11/28/2014 11:30:00 AM | | 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
11/28/2014 11:30:00 AM | | OS: Microsoft Windows 7: Home Premium x64 Edition, Service Pack 1, (06.01.7601.00)
11/28/2014 11:30:00 AM | | Memory: 15.90 GB physical, 48.68 GB virtual
11/28/2014 11:30:00 AM | | Disk: 238.47 GB total, 181.74 GB free
11/28/2014 11:30:00 AM | | Local time is UTC -8 hours
11/28/2014 11:30:00 AM | Einstein@Home | Found app_config.xml
11/28/2014 11:30:00 AM | Milkyway@Home | Found app_config.xml
11/28/2014 11:30:00 AM | SETI@home | Found app_config.xml
11/28/2014 11:30:00 AM | | Config: use all coprocessors
11/28/2014 11:30:00 AM | Einstein@Home | URL http://einstein.phys.uwm.edu/; Computer ID 3967953; resource share 50
11/28/2014 11:30:00 AM | Milkyway@Home | URL http://milkyway.cs.rpi.edu/milkyway/; Computer ID 257518; resource share 450
11/28/2014 11:30:00 AM | SETI@home | URL http://setiathome.berkeley.edu/; Computer ID 5741129; resource share 500
11/28/2014 11:30:00 AM | Einstein@Home | General prefs: from Einstein@Home (last modified 15-Nov-2014 13:59:24)
11/28/2014 11:30:00 AM | Einstein@Home | Host location: none
11/28/2014 11:30:00 AM | Einstein@Home | General prefs: using your defaults
11/28/2014 11:30:00 AM | | Preferences:
11/28/2014 11:30:00 AM | | max memory usage when active: 8141.67MB
11/28/2014 11:30:00 AM | | max memory usage when idle: 14655.00MB
11/28/2014 11:30:00 AM | | max disk usage: 2.00GB
11/28/2014 11:30:00 AM | | max CPUs used: 6
11/28/2014 11:30:00 AM | | (to change preferences, visit a project web site or select Preferences in the Manager)
11/28/2014 11:30:00 AM | | Not using a proxy
11/28/2014 11:30:01 AM | Milkyway@Home | Starting task de_81_DR8_rev_8_4_00001_1413455402_17832414_0
11/28/2014 11:30:01 AM | SETI@home | Sending scheduler request: To fetch work.
11/28/2014 11:30:01 AM | SETI@home | Requesting new tasks for CPU and NVIDIA GPU
11/28/2014 11:30:02 AM | SETI@home | Scheduler request completed: got 0 new tasks
11/28/2014 11:30:02 AM | SETI@home | Project has no tasks available
11/28/2014 11:30:07 AM | Milkyway@Home | Sending scheduler request: To fetch work.
11/28/2014 11:30:07 AM | Milkyway@Home | Requesting new tasks for NVIDIA GPU
11/28/2014 11:30:09 AM | Milkyway@Home | Scheduler request completed: got 0 new tasks
11/28/2014 11:30:09 AM | Milkyway@Home | Not sending work - last request too recent: 27 sec
Here is app_config.xml:
einsteinbinary_BRP4G
Binary Radio Pulsar Search (Arecibo, GPU) v1.39 (BRP4G-cuda32-nv301)
0.33
0.5
einsteinbinary_BRP5
Binary Radio Pulsar Search (Perseus Arm Survey) v1.39 (BRP5-cuda32-nv301)
0.33
0.6
I used to have a max_concurrent entry in app_config.xml but I removed it for the new cards and reduced app_config.xml down to its simplest form for running 3 tasks per card.
Thanks in advance, Keith