The app_name and plan class info for the different searches can be found on this page: https://einsteinathome.org/apps.php?xml=1
The plan class info can also be found on the Applications page linked at the bottom right here on the site. It's the text within the parentheses.
Just add more <app_version> blocks to control different applications.
Now that I have a working Amd 460 card I will see if I can "scale up" today.
If I am reading things right, you can run 2 or even 3 gpu tasks per card with no slowdowns in the wallclock time?
Tom M
Sort of what you have to do is figure out how long it takes to run one workunit at a time on the gpu, then you try to run two workunits at one time and hope the time does not double, if it doubles you are not saving anything. Running more than one workunit at one time will increase the run time but hopefully not double or triple the run time of a single workunit, or even worse have it take longer than the relative time of you running one workunit at a time. Not all gpu's can do multiple workunits at one time and alot of it depends on your gpu's and the machine it sits in as you will be increasing the heat and pushing the gpu harder than it is right now running one workunit at a time. There are lots of variables and the easiest way to see if it will work is to try it and see.
Now that I have a working Amd 460 card I will see if I can "scale up" today. If I am reading things right, you can run 2 or even 3 gpu tasks per card with no slowdowns in the wallclock time? ---edit--- I am missing some syntax. I want my app_config.xml code to only refer to the AMD gpus not the NVIDIA gpus. ---edit--- In Nvidia there is a gui utility that lets me look at various things the card is reporting is there an equivalent for Amd in Ubuntu/Linux? Tom M
There are a couple of Linux utilities for monitoring AMD cards. Radeontop and Rick's Lab amdgpu-utils.
Tom M wrote:I am missing some
)
First read the documentation: https://boinc.berkeley.edu/wiki/Client_configuration#Project-level_configuration
From that you see that you need to use the <app_version> part of the app_config.xml
Something along the lines of this should get you going:
<app_config> <app_version> <app_name>Application_Name</app_name> <plan_class>mt</plan_class> <avg_ncpus>x</avg_ncpus> <ngpus>x</ngpus> </app_version> </app_config>The app_name and plan class info for the different searches can be found on this page:
https://einsteinathome.org/apps.php?xml=1
The plan class info can also be found on the Applications page linked at the bottom right here on the site. It's the text within the parentheses.
Just add more <app_version> blocks to control different applications.
Tom M wrote:Now that I have a
)
Sort of what you have to do is figure out how long it takes to run one workunit at a time on the gpu, then you try to run two workunits at one time and hope the time does not double, if it doubles you are not saving anything. Running more than one workunit at one time will increase the run time but hopefully not double or triple the run time of a single workunit, or even worse have it take longer than the relative time of you running one workunit at a time. Not all gpu's can do multiple workunits at one time and alot of it depends on your gpu's and the machine it sits in as you will be increasing the heat and pushing the gpu harder than it is right now running one workunit at a time. There are lots of variables and the easiest way to see if it will work is to try it and see.
Tom M wrote:Now that I have a
)
There are a couple of Linux utilities for monitoring AMD cards. Radeontop and Rick's Lab amdgpu-utils.
https://awesomedetect.com/how-to-monitor-amd-ati-or-radeon-gpu-usage-in-linux/
https://github.com/Ricks-Lab/amdgpu-utils