 |
|
Message boards : Number crunching : Sulphur (Phase 2) question
Message board moderation
| Author | Message |
|---|---|
|
Send message Joined: 26 Jan 05 Posts: 6 Credit: 158,640 RAC: 0 |
Looking at the results I\'ve been returning to the server, I noticed that the seconds per time slice (sec/TS) keep increasing. Is this normal? If so, does the RAC algorithm take the increased processing time into account? Returned results for WU 463g_b00294796_0 : 04 Oct 2005 4.5437 03 Oct 2005 4.5359 03 Oct 2005 4.5163 02 Oct 2005 4.5008 01 Oct 2005 4.4924 01 Oct 2005 4.4890 30 Sep 2005 4.4885 29 Sep 2005 4.4897 29 Sep 2005 4.4925 28 Sep 2005 4.4944 28 Sep 2005 4.4845 (Phase 1) 27 Sep 2005 4.4755 (Phase 1) 26 Sep 2005 4.4653 (Phase 1) |
|
Send message Joined: 22 May 05 Posts: 7 Credit: 170,326 RAC: 0 |
Looking at the results I\'ve been returning to the server, I noticed that the seconds per time slice (sec/TS) keep increasing. Is this normal? If so, does the RAC algorithm take the increased processing time into account? In my case, at the start of the phase a trickle took from 4.4392 to 4.0798. Ever since then, the time varies between 4.0675 and 4.1156 (mostly around 4.080). So I would say in my case sec/TS is not increasing |
 Send message Joined: 5 Aug 04 Posts: 172 Credit: 4,023,611 RAC: 0 |
It depends a bit on the model being run. I believe that there are some fast processing (ice balls) that get much faster through the run. And there are others that git a bit slower.   BOINC WIKI |
|
Send message Joined: 16 Oct 04 Posts: 692 Credit: 277,679 RAC: 0 |
In the slab model, runs tended to get slower the hotter they got. Phase 2 should be pretty much the same annual mean temperature thoughout (same as phase 1). There usually is a little variation for trickles between summer and winter. Also a little more variation between phases than within phases has also been noticed before. However averaging over 3 trickles I see your variation is from ~48000 sec per trickle to ~54000 sec per trickle which is more than 10%. That is more variation than I would have expected. However, it is not so much more that I would be sure that it is not just working as it should. Other possible causes: Has there been increasing use of the system? Is the disk getting more fragmented? CP gives exactly the same amount of credit per trickle regardless of how much processing it took. The variation is usually small between work units. I suspect that trying to use benchmarks to measure the amount of work done would introduce far larger errors than the current system. Visit BOINC WIKI for help And join BOINC Synergy for all the news in one place. |
©2024 cpdn.org
