Thanks Mikey! It did really help. It's sad that I have to retire my top rig of Einstein, but it yet can help with Milky Way!
No problem...I use this line when I have multiple gpu's in a machine and don't want both crunching for the same project. I sometimes have both an AMD and an Nvidia gpu in the same machine and some projects like one much better than the other. When I do that I have to have two exclude sections, one for each card, one for device zero and another for device one.
Looks like cc_config.xml doesn't work as it should. I still have to manually enable tasks download for a minute, disable it after that and delete BRP6 CUDA tasks manually. Any ideas on that?
MilkyWay OpenCL tasks completed successfully. So I have to set BOINC to use GPU only for MilkyWay and to use CPU for Einstein. Any ideas how to do that?
P.S. all venues are busy for other hosts...
You can use an exclude line in your cc_config.xml file like this:
This way you can have your pc on a venue that allows both cpu and gpu units, but the exclude section will prevent you from getting any gpu units from Einstein. You can have multiple exclude sections if needed.
That cc_config option will stop received tasks being run on GPU device 0, allowing any received work to be run GPU devices 1, 2, 3 etc,
It will not stop you getting those GPU tasks in the first place, to do that you have to deselect either the 'Use **** GPU' setting, or the GPU work type in the project preferences.
There are four locations/venues available, so you can do four different sets of preferences.
64-bit NVidia driver for YAU (Yet Another Ubuntu) releases completely broke with a recent update. Even with nvidia-modprobe installed from the backport no GPUs will be found, at least under Cinnamon. I have a fresh install of Mint17 64-bit KDE which seems to have found the CUDA using nvidia-modprobe and the standard 331 driver.
Thank you for information. But I'm currently concerning on Nvidia GPU under Win7 x86 and don't have any experience on Ubuntu yet.
For me I've found a suitable solution to exclude GPU from crunching Einstein tasks, current and future ones, modifying cc_config.xml file in BOINC data folder. I've added following lines to .
The history. May be useful in investigation.
I've remembered, this host some month ago refused to support NVidia GeForce 750.
That time I found that internal power of PCI-E x16 and x8 lines died.
Fortunately I have a friend with GeForce 660Ti. So I made an exchange with him. Now he plays on my silent GeForce 750, but I can not use in on his host for E@H because of very thin and expensive 4G internet connection.
But his GeForce 660Ti since then worked on E@H for several months.
Now I've added to my host GeForce 250 GTS with two 6-pin external power lanes.
So overall consumption of 250 and 660Ti under load is about 250 Watts.
It works suitably but sometimes (once per 10-20 WUs) 660Ti brakes video driver under MilkyWay. If I try to compute 2 MW tasks simultaneously it falls on each task. This was never happen when I used only one 660Ti.
So I made a conclusion that 12-v lanes may be overloaded and my 585W PSU (in fact 460W) doesn't produce enough power. I have to try a more powerful PSU then.
I'll try to find one later on next week.
RE: Thanks Mikey! It did
)
No problem...I use this line when I have multiple gpu's in a machine and don't want both crunching for the same project. I sometimes have both an AMD and an Nvidia gpu in the same machine and some projects like one much better than the other. When I do that I have to have two exclude sections, one for each card, one for device zero and another for device one.
But I still don't understand
)
But I still don't understand why I can't crunch CPU versions of GPU applications. I have only Gamma Ray WU's now.
RE: But I still don't
)
Because none of the GPU work types have CPU applications at present?
Einsteinathome applications
Claggy
That clarifies this a bit.
)
That clarifies this a bit. Thank you.
Looks like cc_config.xml
)
Looks like cc_config.xml doesn't work as it should. I still have to manually enable tasks download for a minute, disable it after that and delete BRP6 CUDA tasks manually. Any ideas on that?
RE: RE: MilkyWay OpenCL
)
That cc_config option will stop received tasks being run on GPU device 0, allowing any received work to be run GPU devices 1, 2, 3 etc,
It will not stop you getting those GPU tasks in the first place, to do that you have to deselect either the 'Use **** GPU' setting, or the GPU work type in the project preferences.
There are four locations/venues available, so you can do four different sets of preferences.
Claggy
RE: There are four
)
Unfortunately all venues in my computer park are already used for different host types, so I have to find a way to configure this host individually.
64-bit NVidia driver for YAU
)
64-bit NVidia driver for YAU (Yet Another Ubuntu) releases completely broke with a recent update. Even with nvidia-modprobe installed from the backport no GPUs will be found, at least under Cinnamon. I have a fresh install of Mint17 64-bit KDE which seems to have found the CUDA using nvidia-modprobe and the standard 331 driver.
Roland Hughes, President
Logikal Solutions
http://www.theminimumyouneedtoknow.com
http://www.infiniteexposure.net
http://www.johnsmith-book.com
Thank you for information.
)
Thank you for information. But I'm currently concerning on Nvidia GPU under Win7 x86 and don't have any experience on Ubuntu yet.
For me I've found a suitable solution to exclude GPU from crunching Einstein tasks, current and future ones, modifying cc_config.xml file in BOINC data folder. I've added following lines to .
http://einstein.phys.uwm.edu/ NVIDIAFor now this GPU is suitably working on MilkyWay OpenCL tasks without any error for about 2 weeks already.
I yet don't know how could it be.
Found out some new
)
Found out some new things.
The history. May be useful in investigation.
I've remembered, this host some month ago refused to support NVidia GeForce 750.
That time I found that internal power of PCI-E x16 and x8 lines died.
Fortunately I have a friend with GeForce 660Ti. So I made an exchange with him. Now he plays on my silent GeForce 750, but I can not use in on his host for E@H because of very thin and expensive 4G internet connection.
But his GeForce 660Ti since then worked on E@H for several months.
Now I've added to my host GeForce 250 GTS with two 6-pin external power lanes.
So overall consumption of 250 and 660Ti under load is about 250 Watts.
It works suitably but sometimes (once per 10-20 WUs) 660Ti brakes video driver under MilkyWay. If I try to compute 2 MW tasks simultaneously it falls on each task. This was never happen when I used only one 660Ti.
So I made a conclusion that 12-v lanes may be overloaded and my 585W PSU (in fact 460W) doesn't produce enough power. I have to try a more powerful PSU then.
I'll try to find one later on next week.