I tried to create an app_config.xml file, but I got an error message saying that it couldn't find the program when I reloaded the config.
I guess this is a server setting that is attached to the task itself.
It used to be 0.2 CPU I think. I have just done a quick check of what it is running on my computer.
Each task uses 3-5% of my total CPU or 25-40% per core, I7 920 with HT enabled. Based on this I guess a CPU usage of say 0.33 cores per task should work fine.
If you also leave at least one core free for normal computer usage there is a little spare room should they all run at 5% usage.
Next time you want to create an app_config.xml check in the event log in Boinc for the line where one of the task you wish to configure starts, the short name is printed in that line. Something like:
01/08/2013 21:03:18 | Einstein@Home | Starting task PA0012_01151_8_1 using einsteinbinary_BRP5 version 139 (BRP5-cuda32-nv301) in slot 7
I just got some 1.39 workunits. They are obviously crunched by my HD4000. For my NVIDIA 610M i have a GPU utilization factor of .5 and crunch two workunits at a time. For now this factor did not apply for the BRP4 1.34 tasks that my HD4000 crunched. This is totally ok, because I do not really want to do more than two tasks at a time on the HD4000, since it is not so fast.
Now comes the problem: Obviously the factor is taken into account for BRP5 1.39 workunits but not for BRP4 1.34 workunits. I don't think this is a good solution and also it is quite confusing. I think it would be best to handle Intel-BRP5s just like Intel-BRP4s in terms of the utilization factor or at least to have the opportunity to allocate seperate utilization factors for both GPUs.
Obviously the factor is taken into account for BRP5 1.39 workunits but not for BRP4 1.34 workunits. I don't think this is a good solution and also it is quite confusing. I think it would be best to handle Intel-BRP5s just like Intel-BRP4s in terms of the utilization factor or at least to have the opportunity to allocate seperate utilization factors for both GPUs.
Today I saw that my INTEL HD4000 seems actually to be faster than my NVIDIA 610M at BRP5 tasks. I guess that is because of some improvement in the calculations in version 1.39?
And another observation: Something still seems to be wrong with the utilizaion factor of INTEL-GPUs. My Boinc currently runs one BRP4 (0.5 cpus + 1 INTEL-gpu) and one BRP4 (0.5 cpus + 0.5 INTEL-gpu).
edit: Whatever caused the one BRP4 to run with 1 INTEL-gpu is gone. I don't know what caused that though. Now BRP4s are running with 0.5 INTEL-gpus. Wouldn't it be better to force the gpu utilization value for INTEL-gpus to 1 for all apps? Otherwise I think it is somewhat confusing to have 1 for BRP5 and the custom set (in my case .5) for BRP4.
RE: As far as I can find
)
App_config.xml is a Boinc feature and can be used to configure the apps of any attached project.
To set it up for the BRP5 run read the instructions at the bottom of this page and then us "einsteinbinary_BRP5" (without the quotes) in the tag.
Are they still the new v1.39
)
Are they still the new v1.39 or did they revert to v1.33 ?
I tried to create an
)
I tried to create an app_config.xml file, but I got an error message saying that it couldn't find the program when I reloaded the config.
I guess this is a server setting that is attached to the task itself.
It used to be 0.2 CPU I think. I have just done a quick check of what it is running on my computer.
Each task uses 3-5% of my total CPU or 25-40% per core, I7 920 with HT enabled. Based on this I guess a CPU usage of say 0.33 cores per task should work fine.
If you also leave at least one core free for normal computer usage there is a little spare room should they all run at 5% usage.
Many thanks
Arvid
Arvid Almstrom
RE: Are they still the new
)
Mine are all the new 1.39 app.
Arvid
Arvid Almstrom
RE: RE: As far as I can
)
Thanks Holmis
I used the full name I found in the client_state.xml file, but using the shorter version worked.
The task still tells me that it is using 1 CPU, but I now have all the other CPU tasks running.
Here is my app_config.xml file
Arvid
Arvid Almstrom
Next time you want to create
)
Next time you want to create an app_config.xml check in the event log in Boinc for the line where one of the task you wish to configure starts, the short name is printed in that line. Something like:
01/08/2013 21:03:18 | Einstein@Home | Starting task PA0012_01151_8_1 using einsteinbinary_BRP5 version 139 (BRP5-cuda32-nv301) in slot 7
The bold part (my doing) is the name your after.
I just got some 1.39
)
I just got some 1.39 workunits. They are obviously crunched by my HD4000. For my NVIDIA 610M i have a GPU utilization factor of .5 and crunch two workunits at a time. For now this factor did not apply for the BRP4 1.34 tasks that my HD4000 crunched. This is totally ok, because I do not really want to do more than two tasks at a time on the HD4000, since it is not so fast.
Now comes the problem: Obviously the factor is taken into account for BRP5 1.39 workunits but not for BRP4 1.34 workunits. I don't think this is a good solution and also it is quite confusing. I think it would be best to handle Intel-BRP5s just like Intel-BRP4s in terms of the utilization factor or at least to have the opportunity to allocate seperate utilization factors for both GPUs.
RE: Obviously the factor is
)
Agreed. Done.
BM
BM
Thanks!
)
Thanks!
Today I saw that my INTEL
)
Today I saw that my INTEL HD4000 seems actually to be faster than my NVIDIA 610M at BRP5 tasks. I guess that is because of some improvement in the calculations in version 1.39?
And another observation: Something still seems to be wrong with the utilizaion factor of INTEL-GPUs. My Boinc currently runs one BRP4 (0.5 cpus + 1 INTEL-gpu) and one BRP4 (0.5 cpus + 0.5 INTEL-gpu).
edit: Whatever caused the one BRP4 to run with 1 INTEL-gpu is gone. I don't know what caused that though. Now BRP4s are running with 0.5 INTEL-gpus. Wouldn't it be better to force the gpu utilization value for INTEL-gpus to 1 for all apps? Otherwise I think it is somewhat confusing to have 1 for BRP5 and the custom set (in my case .5) for BRP4.