boinc do not want to use second gpu
1 Jul 2015 4:15:40 UTC
Topic 198139
(moderation:
01.07.2015 7:01:32 | | OpenCL: AMD/ATI GPU 0 (ignored by config): Tahiti (driver version 1642.5 (VM), device version OpenCL 1.2 AMD-APP (1642.5), 3072MB, 3072MB available, 2816 GFLOPS peak)
01.07.2015 7:01:32 | | OpenCL: AMD/ATI GPU 1: Tonga (driver version 1642.5 (VM), device version OpenCL 2.0 AMD-APP (1642.5), 2048MB, 2048MB available, 2180 GFLOPS peak)
ignored by which config? i don't have any cc_config file on my pc
boinc do not want to use second gpu
)
The default configuration is built into the BOINC client so there is not a separate configuration file until you create one. You need to do this any time you wish to over-ride the defaults. Check out BOINC Client Configuration. There is an option to force the use of all GPUs.
Cheers,
Gary.
RE: There is an option to
)
I think BOINC uses all GPUs by default now. I haven't needed that option for a while.
RE: 01.07.2015 7:01:32 | |
)
When I look at your pc it only lists one gpu, did you actually load the driver a 2nd time so the Tahiti gpu can have it's own driver? This is not always needed but in your case it seems to be as Boinc is only seeing one gpu.
For a cc_config.xml file this one works:
1
Just use NOTEPAD in Windows, NOT a word processing program, and copy and paste that in your Boinc directory, or if you ONLY ever want to use both gpu's here you could put it in the Einstein project directory. I personally never put it in the project directories, as there are plenty of options in the link Gary gave to use or not use them at this or that project. Save the file as a 'txt type file' with the xml file extension so it will be saved as cc_config.xml
RE: RE: There is an
)
No, it just uses the 'best' GPU(s).
If you have two identical (or near identical) GPUs, it could well use both of them automatically - there's a 'loose' element to the test which doesn't require a perfect match.
But if the GPUs differ in what is considered a significant feature, then only the 'best' (as defined in code) GPU is used.
And if only one GPU is active locally, then only one GPU will be reported to the server for display.
oh, it works now! but PSU
)
oh, it works now!
but PSU told "i can't" and turned off my PC )
RE: Just use NOTEPAD in
)
If you put the file (cc_config.xml) in a project directory it will not be used. It's a Core Client Configuration file and belongs in the Boinc data directory and nowhere else.
RE: RE: I think BOINC
)
I stand corrected. But after looking at the code, I must say I'm very surprised by what is (not) considered significant. In particular the loose way completely ignoring speed.
RE: RE: Just use NOTEPAD
)
Oops sorry for the misinformation then!
RE: oh, it works now! but
)
Are you trying to use the psu that came with the machine? If so this is pretty normal most psu's that come with the machine are just barely over the line of being able to run the machine itself, smaller is cheaper and over the course of a run of building pc's even 10 cents per pc can add up.
I have noticed a strange
)
I have noticed a strange issue when running 2 identical HD 7950 GPUs.
They are both running 1 BRP6 task, but the run times differ significantly: d0 takes about 6100 s, while d1 ~3100, which is almost double !
I have verified their properties and both have identical constant values - GPU Clock, MemClock, Power limits, Voltages, etc.
Only the GPU Usage on d0 is slightly lower by about 5-6%, which however can't explain the huge difference.
Another difference I noticed is the CPU usage: d0 is ~48%, while d1 ~75%. So it looks like the CPU is saturated, but.. this is a 4C/8T machine running only 2 other Test4Theory (VM) tasks and the overall CPU usage is ~40%. Even when I suspend the T4T tasks, it doesn't change.
I have also tried several different Catalyst drivers including the last Beta, but it's still the same.
Any idea what might be causing such a hige difference? I don't recall seeing this in the past...
It's this host: http://einsteinathome.org/host/7192129/tasks