Message boards :
Number crunching :
All I get are Cuda WU's...
Message board moderation
Previous · 1 · 2
Author | Message |
---|---|
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13751 Credit: 208,696,464 RAC: 304 |
I've found when requesting work, and it doesn't get any, the Manager tends to be somewhat aggressive in it's backoff for asking for work. The "work fetch deferred for" & "interval" start off at about 1 for the first miss on a request for work, a 2nd miss and it's 2 to 5 min. 3rd miss it's 10-30 min. A 4th straight miss & i've seen it backoff for several hours. As long as some work gets allocated, everything is OK. As long as something gets allocatd on a second request after missing out the first time around it's all good. But if you don't get anything for 2 consecutive requests, it all turns ugly very quickly. Grant Darwin NT |
hiamps Send message Joined: 23 May 99 Posts: 4292 Credit: 72,971,319 RAC: 0 |
Well have received 1000's of GPU units since my first post and still no CPU units. Could reschedule but will just let it be. Bought a new "old" stereo and play mostly from hard drive so I guess it all works out. Was just curious if mine was the only one and it's not so all is fine in seti land.... Official Abuser of Boinc Buttons... And no good credit hound! |
hiamps Send message Joined: 23 May 99 Posts: 4292 Credit: 72,971,319 RAC: 0 |
Well have received 1000's of GPU units since my first post and still no CPU units. Could reschedule but will just let it be. Bought a new "old" stereo and play mostly from hard drive so I guess it all works out. Was just curious if mine was the only one and it's not so all is fine in seti land.... WooHoo first time CPU has had work that I have seen...not really worried was just checking...No longer in the race. I usually have no idea what I am talking about either..............ROFL Official Abuser of Boinc Buttons... And no good credit hound! |
Joe Januzzi Send message Joined: 13 Apr 03 Posts: 54 Credit: 307,134,110 RAC: 492 |
Thanks so much for the on/off ticking trick, it works great. For way toooo long one of my PCs cache for CPU tasks would always be low or none (it would always ask for both, but mostly got GPU down loads) , while my other PCs ran okay for both the CPU & GPU (cache = 10 Days +/- 40). I did way to much rescheduling to keep my Q9550 happy. This trick works fast, all my PC's are topped off, while writing this comment. I'm set for the 4th of July because of your Great Wisdom. Thanks again Real Join Date: Joe Januzzi (ID 253343) 29 Sep 1999, 22:30:36 UTC Try to learn something new everyday. |
BMH Send message Joined: 27 May 99 Posts: 419 Credit: 166,294,083 RAC: 125 |
Thanks from me too. I had the opposite problem - my CPU was loaded up but GPUs running cold, affecting my RAC :( Brian. |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13751 Credit: 208,696,464 RAC: 304 |
Been keeping an eye on things since this last outage & yeah, so far all the work has been going to the GPU, even when there is no deferral for the CPU. Requests are made for both CPU & GPU, but it's the GPU that gets the lot each time. Grant Darwin NT |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
The server-side Scheduler process which handles a work request looks through the tasks which the Feeder has told it about. When it finds one which matches a kind of work (setiathome_enhanced or astropulse_v505) the host is set to accept, and the request is for both CPU and GPU, it looks to see which of the available applications should do the task quickest. Given that the Scheduler code needs to be fast in order to handle a large number of requests in minimum time, and it's clearly best for the project to send work to the most efficient application, that logic is unlikely to be modified. The preference settings recommended earlier are certainly one way to get a mix, but I think there's another - don't ask for so much work. For example, if you already have GPU work which will last a few hours but no CPU work, asking for 3 days of work will just get more GPU work. Asking for 2 hours of work ought to produce a request for CPU work only since the idle CPU state will still trigger a request. I don't have a suitable setup to test that theory, but I'd simply use the local prefence settings for "Connect about every" and "Additional work buffer" to reduce the request size. Joe |
kittyman Send message Joined: 9 Jul 00 Posts: 51469 Credit: 1,018,363,574 RAC: 1,004 |
My computer went to Seti and all I got was this Cuda workunit....... "Freedom is just Chaos, with better lighting." Alan Dean Foster |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.