only GPU calculates units, not multicore CPU

Questions and Answers : Windows : only GPU calculates units, not multicore CPU

Author	Message
Gregory Wahl Send message Joined: 31 May 99 Posts: 2 Credit: 5,587,559 RAC: 5,329	Message 1309838 - Posted: 24 Nov 2012, 17:11:06 UTC
	hi, I have seen similar threads, but no satisfying answer. This started happening a couple of months ago with no configuration changes made. For some reason, my i5 quad core will stop computing, but the Nvidia GPU always continues. I have checked many times the configuration, and everything is wide open, use 100% of every core 100% of the time. No memory or hd restrictions. Then for unknown reasons the i5 will start computing, often for several days before suddenly stopping for between several days to a week, before starting again. During the i5 inactive period there are dozens of 'ready to start' wu in my local que. Any idea as to what is causing this. Thanks.
	ID: 1309838 ·

Gundolf Jahn Send message Joined: 19 Sep 00 Posts: 3078 Credit: 341,914 RAC: 96	Message 1309875 - Posted: 24 Nov 2012, 18:39:28 UTC - in response to Message 1309838.
	During the i5 inactive period there are dozens of 'ready to start' wu in my local queue. Any idea as to what is causing this. Thanks. Probably all those WUs are designed to run on the GPU. Lower your cache settings until only CPU tasks are requested because there already are enough GPU tasks downloaded. Gruß, Gundolf ____________ Computer sind nicht alles im Leben. (Kleiner Scherz) SETI@home classic workunits 3,758 SETI@home classic CPU time 66,520 hours
	ID: 1309875 ·

rob smith Volunteer moderator Send message Joined: 7 Mar 03 Posts: 5473 Credit: 21,219,737 RAC: 135,467	Message 1309912 - Posted: 24 Nov 2012, 20:38:35 UTC
	BOINC will always load the most capable processor first. A point of clarification, all work-units are created equal, however not every processor is capable, or permitted, to receive particular types of work At the time of distribution the server checks to see if the requesting processor is allowed to process the work and if it is, then the work unit is sent; if it isn't then either another work unit of a different type is selected for dispatch, or if there are no suitable work units you get the "no work available" message. ____________ Bob Smith Member of Seti PIPPS (Pluto is a Planet Protest Society) Somewhere in the (un)known Universe?
	ID: 1309912 ·

Gregory Wahl Send message Joined: 31 May 99 Posts: 2 Credit: 5,587,559 RAC: 5,329	Message 1310635 - Posted: 27 Nov 2012, 0:26:49 UTC - in response to Message 1309912.
	thanks for both responses. When I expand the Application tab to show everything that was previously truncated, I now see where the WU for the GPU is defined as a cuda unit. I didn't know that there were 2 types of WUs. I modified my buffer and it seems to be more balanced now between both types of WU instead of mostly GPU WUs. Thanks again.
	ID: 1310635 ·

rob smith Volunteer moderator Send message Joined: 7 Mar 03 Posts: 5473 Credit: 21,219,737 RAC: 135,467	Message 1310757 - Posted: 27 Nov 2012, 18:30:20 UTC
	You cannot modify your buffer in the way you have described - work is allocated by the servers preferentially to the GPU. ____________ Bob Smith Member of Seti PIPPS (Pluto is a Planet Protest Society) Somewhere in the (un)known Universe?
	ID: 1310757 ·

John McLeod VII Volunteer developer Volunteer tester Send message Joined: 15 Jul 99 Posts: 22343 Credit: 439,209 RAC: 396	Message 1310889 - Posted: 28 Nov 2012, 2:00:47 UTC - in response to Message 1310757.
	You cannot modify your buffer in the way you have described - work is allocated by the servers preferentially to the GPU. If you shrink your buffer, it may start giving you tasks for the under utilized processor. ____________ BOINC WIKI
	ID: 1310889 ·

Questions and Answers : Windows : only GPU calculates units, not multicore CPU