How hard does PG work / stress a GPU??
Do people run one wu per GPU or more ??
PG runs the CPU and GPU pretty hard. In the preferences (https://www.primegrid.com/prefs.php?subset=project) you can state weather you want to run multi-threaded where more than one CPU core is run per task, and they finish quicker and more chance of being the first to return if it is Prime Number.
I also give permission and give my name in the "Reporting primes to the Prime Pages" in the default preferences so I am credited with any primes I find.
I run 2 x wu's at a time on my cards. The Radeon VII's can handle it but it's barely worth running 2 on my Vega 56 & 280X.... I just can't be bothered changing the app config :)
Running 2 wu's on my Radeon VII still doesn't put the GPU under full load and it runs MUCH cooler than say running Milkyway or Einstein.
CPU is a different story. Quite hard on the temps although multi threading isn't used quite as much now with LLR2.
http://www.primegrid.com/forum_thread.php?id=9303
Have found plenty of info on how to setup PG CPU side of things, but nothing on how to run multiple wu's on a GPU.
Can i simply tell the project to run multiple wu's on every project with a simple change or do i have to setup for each project i choose to run ?
@chooka how have you gone about setting up the config file to run multiple wu's on your GPU's ?
I use this -
<app_config>
<app>
<name>pps_sr2sieve</name>
<report_results_immediately/>
<fraction_done_exact/>
</app>
<app_version>
<app_name>pps_sr2sieve</app_name>
<plan_class>openclatiPPSsieve</plan_class>
<avg_ncpus>0.0001</avg_ncpus>
<ngpus>0.5</ngpus>
</app_version>
</app_config>
Happy crunching!
@chooka So i have to do those settings for every project i crunch on that particular GPU ?