GPU work unit's fail to download
18 Oct 2022 19:30:02 UTC
Topic 228402
(moderation:
I'm new here but not new to distributed computing as I used to run SETI...
I just got my pc setup but I cant get any work units to download for my gpu when I set GPU utilization to 0.5 to run 2 wu's at a time. If I set it to anything but 1.0 I will get a "download failed". Also the screensaver for this project will load, stay on for 5 secs and then it'll go back to the desktop?
My GPU in this machine is a Vega 64 8gb.
I am so confused right now...
You need to have the correct
)
You need to have the correct video drivers that have OpenCL support loaded for your AMD gpu.
If BOINC does not detect your card and recognize it as capable of sending work to, you won't get any gpu work.
You need to have specified in your project preferences that you want to use the gpu also and then select sub-projects that actually have science gpu applications.
First thing to check is the beginning of the Event Log of BOINC at startup. Does it detect and show your gpu?
You should see the gpu listed along with the OpenCL drivers detected that Einstein needs to use for the AMD gpu tasks. If you see no gpu listed or the correct drivers, you will get no work.
You need to switch to Advanced mode for the BOINC Manager. SHIFT-CTRL-A in the Simple View screen or use the Simple View menus to switch to Advanced mode.
The Event Log is available from the Tools menu.
[Edit] I should have looked at your profile first. You are already processing gpu tasks fine with no issues.
It was working perfect
)
It was working perfect earlier but only 1 wu at a time... I changed it to run 2 in my preferences and now all gpu downloads fail. I set it back to 1.0 and still they wont download.
Everything else is working fine with the gpu...
I am in advanced mode
I was processing them
)
I was processing them earlier.... not now. It says Download failed and they're at the bottom of the list.
You must have crafted your
)
You must have crafted your app_config.xml incorrectly and that is causing BOINC to fail the download.
Or simply the project is having issues providing adequate work for the past few days. There might be no work available.
This is my app_config.xml file for GR#1 and BRP7 tasks. Yours should look similar:
<app_config>
<app>
<name>hsgamma_FGRPB1G</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.9</cpu_usage>
</gpu_versions>
</app>
<app>
<name>einsteinbinary_BRP7</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.9</cpu_usage>
</gpu_versions>
</app>
</app_config>
You can turn on http_debug
)
You can turn on http_debug and http_xfer_debug logging in the Event Log and see what that tells you about your work requests and why they are failing.
Here's a pic showing what's
)
Here's a pic showing what's happening... it will go to download them and then it will fail
I will try that
)
I will try that
Ok I used your app_config and
)
Ok I used your app_config and it's downloading gpu wu's and crunching two at a time although I get a Notice saying my app_config.xml file refers to an unknown application 'hs_gamma_FGRPB1G"
Then either change your
)
Then either change your project preferences to accept the FGRPB1G gpu tasks or edit the app_config.xml file and remove the section for the FGRPB1G task.
Your app_config.xml would simply be:
<app_config>
<app>
<name>einsteinbinary_BRP7</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.9</cpu_usage>
</gpu_versions>
</app>
</app_config>
The FGRPB1G tasks reward more credit and crunch faster than the BRP7 but they are in short supplly and eventually the campaign for those tasks is ending. Meaning the only gpu tasks available in the future will be the BRP7.
Keith Myers wrote: You must
)
Keith,
I too tried your <app_config.xml> file and I also get a Notice saying my "app_config.xml file refers to an unknown application hs_gamma_FGRPB1G".
So I replaced it with what I had originally.
<app_config>
<app>
<name>hsgamma_FGRPB1G</name>
<gpu_versions>
<cpu_usage>0.9</cpu_usage>
<gpu_usage>0.5</gpu_usage>
</gpu_versions>
<name>einsteinbinary_BRP7</name>
<user_friendly_name>Binary Radio Pulsar Search (MeerKAT)</user_friendly_name>
<non_cpu_intensive>0</non_cpu_intensive>
</app>
<project_max_concurrent>4</project_max_concurrent>
</app_config>
...and now it is working again.
I don't quite understand the differences. Yes, I know there are some, but the what and why I don't know.
Proud member of the Old Farts Association