In your first message, you also mentioned trying to use an app_config.xml file. The file you showed doesn't need a <max_concurrent> setting since <gpu_usage> set to 0.5 limits the concurrent FGRPB1G tasks to 2 anyway. AMD GPUs need very little CPU support so setting <cpu_usage> to 2 is way overkill. Since you're not running CPU tasks, that doesn't matter for now.
Yes I was playing a bit with the CPU and GPU usage, because my GPU was not really utilized. And with the app_config.xml it was easy to chnage the setting without ping the servers every time with an update.
Gary Roberts wrote:
Here is an easy way for you to tell if app_config.xml has been recognised. On the assumption that the values you showed earlier are still the same, just open a user terminal session on your Arch Linux and execute the following command:-
where "path/to/boinc" is the path from where you are logged in to where the BOINC state file is stored. You should see at least two <count> values, one for how many GPUs you are running and the other for the fraction of a GPU that a task will use. The <avg_ncpus> value will be 2.000000 if your app_config.xml is in play. Of course, this assumes that Einstein is your only project. You may see multiple sets if there are other projects but the <avg_ncpus> value should still be indicative. You will also see multiple sets for other Einstein science apps that you have run.
Report back on what you find. It's possible that you just happened to get more non-FGRPB1G tasks until the changed settings had been communicated to the client. There doesn't seem to be anything obvious that is allowing more of the wrong type of work. Maybe I'm just blind - or dumb - or both :-).
I reset my BOIN installation and initial it again with the projects, to be sure that I don't change something. The only changes I made is in the cc_config.xml and remote_hosts.cfg with a IP address.
I run some other projects also. Here the grep with a bit more informations. Looks like it takes the settings from the webpage. But I din't get only the selected app FGRPB1G. In most cases I get the O3MDF but also both other GPU apps and I think the scheduler recornices the settings for specific apps.
grep -A 2 -B 3 -E '<count>|<avg_ncpus>' client_state.xml
I installed BOINC on another computer (https://einsteinathome.org/host/13135378) without any changes on this setup and only the Einstein@home project with the generic settings, but it don't work there either...
Gary Roberts wrote: In your
)
Yes I was playing a bit with the CPU and GPU usage, because my GPU was not really utilized. And with the app_config.xml it was easy to chnage the setting without ping the servers every time with an update.
I reset my BOIN installation and initial it again with the projects, to be sure that I don't change something. The only changes I made is in the cc_config.xml and remote_hosts.cfg with a IP address.
cc_config.xml<cc_config>
<options>
<allow_remote_gui_rpc>1</allow_remote_gui_rpc>
</options>
</cc_config>
I run some other projects also. Here the grep with a bit more informations. Looks like it takes the settings from the webpage. But I din't get only the selected app FGRPB1G. In most cases I get the O3MDF but also both other GPU apps and I think the scheduler recornices the settings for specific apps.
grep -A 2 -B 3 -E '<count>|<avg_ncpus>' client_state.xml<wsl_available>0</wsl_available>
<coprocs>
<coproc_ati>
<count>1</count>
<name>AMD Radeon RX 7900 XTX</name>
<available_ram>25753026560.000000</available_ram>
--
<app_name>amicable_10_21</app_name>
<version_num>300</version_num>
<platform>x86_64-pc-linux-gnu</platform>
<avg_ncpus>28.000000</avg_ncpus>
<flops>100709399092.602524</flops>
<plan_class>mt</plan_class>
--
<app_name>einstein_O3MDF</app_name>
<version_num>100</version_num>
<platform>x86_64-pc-linux-gnu</platform>
<avg_ncpus>0.900000</avg_ncpus>
<flops>20000000000.000000</flops>
<plan_class>GW-opencl-ati</plan_class>
--
</file_ref>
<coproc>
<type>ATI</type>
<count>1.000000</count>
</coproc>
<gpu_ram>1837105152.000000</gpu_ram>
--
<app_name>einsteinbinary_BRP4G</app_name>
<version_num>133</version_num>
<platform>x86_64-pc-linux-gnu</platform>
<avg_ncpus>1.000000</avg_ncpus>
<flops>1000000000.000000</flops>
<api_version>7.1.0</api_version>
--
<app_name>einsteinbinary_BRP7</app_name>
<version_num>15</version_num>
<platform>x86_64-pc-linux-gnu</platform>
<avg_ncpus>0.500000</avg_ncpus>
<flops>15000000000.000000</flops>
<plan_class>BRP7-opencl-ati</plan_class>
--
</file_ref>
<coproc>
<type>ATI</type>
<count>1.000000</count>
</coproc>
<gpu_ram>734003200.000000</gpu_ram>
--
<app_name>hsgamma_FGRPB1G</app_name>
<version_num>118</version_num>
<platform>x86_64-pc-linux-gnu</platform>
<avg_ncpus>1.000000</avg_ncpus>
<flops>35000000000.000000</flops>
<plan_class>FGRPopencl1K-ati</plan_class>
--
</file_ref>
<coproc>
<type>ATI</type>
<count>1.000000</count>
</coproc>
<gpu_ram>629145600.000000</gpu_ram>
--
<app_name>mcm1</app_name>
<version_num>761</version_num>
<platform>x86_64-pc-linux-gnu</platform>
<avg_ncpus>1.000000</avg_ncpus>
<flops>1000000000.000000</flops>
<api_version>7.7.0</api_version>
--
<app_name>opn1</app_name>
<version_num>721</version_num>
<platform>x86_64-pc-linux-gnu</platform>
<avg_ncpus>1.000000</avg_ncpus>
<flops>8441446673.755708</flops>
<api_version>7.7.0</api_version>
Hm that's weird. I
)
Hm that's weird.
I installed BOINC on another computer (https://einsteinathome.org/host/13135378) without any changes on this setup and only the Einstein@home project with the generic settings, but it don't work there either...
Can you show the full account
)
Can you show the full account -> prefs -> project page now just to be sure nothing has changed, thanks?
Cheers,
Gary.
Sure
)
Sure :)
https://i.ibb.co/922BN9x/2023-04-14-07-07-16-einsteinathome-org-c5f88bfc1f5b.png
https://i.ibb.co/Kqc429X/2023-04-14-07-09-33-einsteinathome-org-18847384c1dd.png