Suddenly One of my GPU don´t run e@h...
20 Nov 2012 22:35:21 UTC
Topic 196636
(moderation:
Yes, one of my two Nvidia GTX 550 TI in SLI mode, don´t compute tasks to I am just nervous, because I love computing...
The AIDA64 software detect the two Graphic Cards normally, and I go to Nvidia Control Panel and suddenly SLI mode was Turn off ...I turn On again, and reset my computer, and Switch On again Boinc...And now, only one of my graphic cards are computing...
And the second Card doesn't appear in the computer of user...And before yes, appear the two graphic cards...
http://einsteinathome.org/account/130870/computers
I need help, because all is aparently normal, but boinc don´t take de second card and SLI mode...
which may have been misconfigured?
Thanks in advance and greetings from Spain :)
Suddenly One of my GPU don´t run e@h...
)
Hallo Jonatha!
I oberved similar behaviour. I´m running a GTX550Ti and a GT440. The Gt440 stops sometime suddenly crunching and seem to be ok by all tests. But sometimes, after switching off/on it comes back to work and than, even after days of normal working, disapears again. Also I´m for an explanation for this behaviour. But there is one big difference between us. I´m crunching on my GPUs MilkyWay@Home, as for unknown reason tasks of BRP4 are ending up with errorson my rig. If I put this togesther, it seems to me a problem of BOINC. But I tried also older versions of BOINC like 6.12.34 without success.
Hopefully, someone else has usefull hints.
Kind regards and happy crunching
Martin
Thank you very much Martin,
)
Thank you very much Martin, really i am very worried with that, I am seeing that the 50% of my SLI is turned off...
I also think that is a boinc problem...If the situation remains same, I will try to uninstall and reinstall
Thanks for answer :)
I think something may have
)
I think something may have gone wrong with the BOINC/SLI relationship. Yesterday I turned my SLI back on to play a game for awhile, and then I just happened to be watching when BOINC resumed afterwards. One of my 570's vanished completely and the other two were running SLOW...like 4 hours projected time to complete a WU instead of 25 minutes.
I shut off SLI and everything snapped back to normal.
It is suposed that BOINC
)
It is suposed that BOINC shouldnt be affected by the setting about SLI...
But this was not always true and it can change when new drivers appears...
Ussually the best thing it to leave SLI disabled while crunching (you dont need to take out the SLI bridge if there is one) just disable SLI on de nvidia control panel.
Old drivers were mirroring the memory of SLI'ed GPUs which caused long runtimes and sometimes too much memory usage... recent drivers were not affecting CUDA apps or OpenCL when SLI was enabled... but who knows...
Also, there is some issue with the last versions of BOINC in which even identical GPUS are recognized different due to a bug in memory recognition and then BOINC just uses the one that reports more memory... but if this were the case you should see both GPUs in the event log, and one of them should be shown as "(not used)". The workaround if this is the case is to instruct BOINC to use all the GPUs via the cc_config file.
By the way, there was a recent upgrade of Microsoft that signaled as important an upgrade of the video drivers and most computers aout there installed that without even prompting to the user... If your drivers were upgraded by the automatic install that Microsoft does, you may have a lot of weird issues... To solve this, just download the drivers you want to use from the nvidia page and install them using the advanced option and select the "do a clean install" option...
Hallo Horacio, hallo
)
Hallo Horacio, hallo Jonatan!
Thank you for answering.
First, I don´t use SLI and it happens.
Seconde, in the Eventlog just after startup I find this note (not used).
Third, I did de/reinstall BOINC without effecting this.
Forth, could you, Horacio, please show us what to introduce where into the config file. Sorry, I´m no programmer.
I´ll try this just after sleeping for some hours.
Kind regards and happy crunching
Martin
RE: Yes, one of my two
)
Please post your Boinc startup messages from the Event Log, the first 20 to 30 lines will do,
Claggy
These are the lines who
)
These are the lines who appears in the event log:
22/11/2012 0:20:20 | | No config file found - using defaults
22/11/2012 0:20:20 | | Starting BOINC client version 7.0.28 for windows_x86_64
22/11/2012 0:20:20 | | log flags: file_xfer, sched_ops, task
22/11/2012 0:20:20 | | Libraries: libcurl/7.25.0 OpenSSL/1.0.1 zlib/1.2.6
22/11/2012 0:20:20 | | Data directory: C:\ProgramData\BOINC
22/11/2012 0:20:20 | | Running under account Jonatan
22/11/2012 0:20:20 | | Processor: 8 GenuineIntel Intel(R) Core(TM) i7-2600K CPU @ 3.40GHz [Family 6 Model 42 Stepping 7]
22/11/2012 0:20:20 | | Processor: 256.00 KB cache
22/11/2012 0:20:20 | | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss htt tm pni ssse3 cx16 sse4_1 sse4_2 syscall nx lm vmx tm2 popcnt aes pbe
22/11/2012 0:20:20 | | OS: Microsoft Windows 7: Professional x64 Edition, (06.01.7600.00)
22/11/2012 0:20:20 | | Memory: 15.98 GB physical, 31.95 GB virtual
22/11/2012 0:20:20 | | Disk: 1.82 TB total, 1.36 TB free
22/11/2012 0:20:20 | | Local time is UTC +1 hours
22/11/2012 0:20:20 | | NVIDIA GPU 0: GeForce GTX 550 Ti (driver version 306.97, CUDA version 5.0, compute capability 2.1, 5120MB, 8381103MB available, 745 GFLOPS peak)
22/11/2012 0:20:20 | | NVIDIA GPU 1 (not used): GeForce GTX 550 Ti (driver version 306.97, CUDA version 5.0, compute capability 2.1, 5120MB, 687MB available, 745 GFLOPS peak)
22/11/2012 0:20:20 | | OpenCL: NVIDIA GPU 0: GeForce GTX 550 Ti (driver version 306.97, device version OpenCL 1.1 CUDA, 1024MB, 8381103MB available)
22/11/2012 0:20:20 | | OpenCL: NVIDIA GPU 1 (not used): GeForce GTX 550 Ti (driver version 306.97, device version OpenCL 1.1 CUDA, 1024MB, 687MB available)
22/11/2012 0:20:20 | Einstein@Home | URL http://einstein.phys.uwm.edu/; Computer ID 5692767; resource share 100
22/11/2012 0:20:20 | | General prefs: from http://www.primegrid.com/ (last modified 26-Dec-2011 17:17:55)
22/11/2012 0:20:20 | | Host location: none
22/11/2012 0:20:20 | | General prefs: using your defaults
22/11/2012 0:20:20 | | Reading preferences override file
22/11/2012 0:20:20 | | Preferences:
22/11/2012 0:20:20 | | max memory usage when active: 8179.55MB
22/11/2012 0:20:20 | | max memory usage when idle: 14723.19MB
22/11/2012 0:20:20 | | max disk usage: 100.00GB
22/11/2012 0:20:20 | | (to change preferences, visit the web site of an attached project, or select Preferences in the Manager)
22/11/2012 0:20:20 | | Not using a proxy
22/11/2012 0:20:20 | Einstein@Home | Restarting task h1_0343.30_S6GC1__75_S6LV1B_1 using einstein_S6LV1 version 113 (SSE2) in slot 6
22/11/2012 0:20:20 | Einstein@Home | Restarting task h1_0343.30_S6GC1__74_S6LV1B_1 using einstein_S6LV1 version 113 (SSE2) in slot 4
22/11/2012 0:20:20 | Einstein@Home | Restarting task h1_0343.30_S6GC1__76_S6LV1B_1 using einstein_S6LV1 version 113 (SSE2) in slot 8
22/11/2012 0:20:20 | Einstein@Home | Restarting task h1_0343.30_S6GC1__20_S6LV1B_1 using einstein_S6LV1 version 113 (SSE2) in slot 3
22/11/2012 0:20:20 | Einstein@Home | Restarting task h1_0343.30_S6GC1__53_S6LV1B_2 using einstein_S6LV1 version 113 (SSE2) in slot 2
22/11/2012 0:20:20 | Einstein@Home | Restarting task h1_0343.30_S6GC1__55_S6LV1B_2 using einstein_S6LV1 version 113 (SSE2) in slot 1
22/11/2012 0:20:20 | Einstein@Home | Restarting task h1_0343.30_S6GC1__54_S6LV1B_2 using einstein_S6LV1 version 113 (SSE2) in slot 0
22/11/2012 0:20:20 | Einstein@Home | Restarting task h1_0343.30_S6GC1__13_S6LV1B_1 using einstein_S6LV1 version 113 (SSE2) in slot 7
22/11/2012 0:20:20 | Einstein@Home | Restarting task p2030.20111104.G193.32-00.89.C.b1s0g0.00000_1664_1 using einsteinbinary_BRP4 version 132 (BRP4cuda32nv301) in slot 5
Thanks in advance :)
me too astro-marwil, now i am
)
me too astro-marwil, now i am going to sleep that tomorrow I have an exam of Database...
i think that is problem of config file, because my two graphic card are the same, and in the event log, appear one card with:
card0 8381103MB available
card1 1024MB, 687MB available
RE: OpenCL: NVIDIA GPU 0:
)
That's impressive !! :-)
Cheers, Mike.
( edit ) Actually I think the programing error looks to be an integer type/length problem, mixing signed long int versus unsigned int : something like that. If you add 1024 to 8381103 you get real close to a power of two, the deficit being I reckon some MB of used memory.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
Something like this is needed
)
Something like this is needed to use 2 GPUs on Einsten to?
1