Originally Posted by mmonnin
Originally Posted by fragamemnon
You can create an app_config.xml file in \BOINCdata\projects\www.cosmologyathome.org\
and limit the concurrently running tasks.
would limit these two apps (lsplitsims and camb_boinc2docker) to 6 CPU cores per task, and a maximum of one task of each running at a single point in time. You can take only the concurrency flags from this config.
With a little bit of estimation, you should be able to alleviate the memory problem.
lsplitsims is not an app, at least not now. The legacy app is single threaded and can eat up the mem. I have some using 750mb as they seem to go up in memory usage as they complete.
Originally Posted by Diablosbud
Going to put the 480 to work bunkering for Einstein. I figure it's a lot more effective than running Cosmology on my 6600K.
Edit: Have 2 cores running Cosmology and the GPU running 2 Einstein tasks.
RX 480s are awesome at E@ for upfront cost and power usage.
This is the #2 computer. All the ones above have 4 more GPUs in their system.
And I noticed @4thKor
has the #7 computer in E@H.
At first I though the same, but when I add planck_param_sims
to the app_config and reload, I get this displayed in the log:
05/05/2017 14:38:31 | Cosmology@Home | Your app_config.xml file refers to an unknown application 'planck_param_sims'. Known applications: 'lsplitsims', 'camb_boinc2docker', 'camb'
So I figured the client must be recognizing planck_param_sims under the name of lsplitsims.
Edit: as for the memory issue, for me a balance of Resource share WCG 200:100 Cosmo satisfies 48 cores with 16GB RAM at their disposal (limited to 90% and 25% of page file, which is 1/2RAM). I now changed C@H to 50 to see whether something will change.Edited by fragamemnon - 5/5/17 at 4:43am