Posts by RWIoffice

1) Message boards : Number crunching : Low Scores Anyone? (Message 30652) Posted 5 Nov 2006 by RWIoffice Post: FWIW, this machine gets only occasional foreground interactive use, and crunches for Rosetta in the background 24/7. RAC has dropped from 226 on 10/5 to 207 (11/5). The BOINC manager stats graph shows pretty much linear decline, with the exception of one upward spike around 10/20-21. A quick backward glance through Messages shows nothing in red since the machine's last reboot for updates on 10/25.
2) Message boards : Number crunching : Report Problems with Rosetta Version 5.22 (Message 18465) Posted 11 Jun 2006 by RWIoffice Post: Screensaver or possibly some other flavor of "completion" problem with a t299__CASP7 work unit. I noticed lack of completion progress from home this morning. When I got to the office, the screensaver was sitting on "Model 9, Step 0." Once I got past the screensaver, BOINC Manager was reporting the work unit as "Running" and "100%" for Progress. BOINC Manager would not display the graphics. Shutdown of BOINC Manager seemed to take a long time, but it finally happened. I rebooted the system. Once BOINC Manager launched it reported the status on this work unit as completed, "Ready to Report" and started the next work unit. I forced an update so the result would be available prior to my posting this report. Keeping in mind that I'm a newbie and could easily be misinterpreting, the result seems to be referring to only 8 models, so the screensaver graphic's Model 9 reference doesn't make sense to me.
3) Message boards : Number crunching : Report Problems with Rosetta Version 5.22 (Message 18391) Posted 10 Jun 2006 by RWIoffice Post: Possible problem with a t299_CASP7 work unit (link to WU). Was a happy camper, then at step 370K+ on Model 6 my CPU dropped from 100% to nothing, and graphics display showed no progress. Didn't write down the stuck step number, sorry. Suspend on that task released the waiting next, which drove the CPU back to full load. Suspended task #2 and resumed #1, but it still didn't seem to grab any CPU. For lack of knowing any better (new user), I shutdown BOINC and restarted, which I think I understand to mean that the task resumes at the previous checkpoint (model boundary)? It is now running again. What is accepted practice if it hangs up again, please? Do I wait for some watchdog abort, or manually abort it? I don't really care about credits, I'll take whatever action provides the best feedback about the failure. Thanks!
4) Message boards : Number crunching : A bit of ThreadMaster help? (Message 18323) Posted 9 Jun 2006 by RWIoffice Post: >But on a dual core, you might achieve your objective simply by limiting >BOINC to 1 CPU. This is in your General Preferences. Thanks Feet1st (this is Alan Roberts, I'm over at the account setting for a different box). I was running limited to one CPU, and that was keeping the fan at a tolerable level. I only had a single Rosetta task running in this mode. Watching load in task manager, it was pretty variable (Graphic for each CPU core bouncing around in the 30-60% range ... I'm assuming this was happening as the threads within the process context-switched for various reasons). It certainly looked like both cores were in use for Rosetta (nothing else would have been loading either core). I was assuming threads might get scheduled onto both cores? Switching over to use of ThreadMaster and back to two CPUs in General Preferences was my experiment at squeezing more work out of the box while maintaining a tolerable fan noise level. FWIW, I think I've tracked down the problem in the time since my post. Based on my previous use on a uniprocessor, I had ThreadMaster set to activate when an application crossed at 95% percentage. Then my custom setting for rosetta_..._intelx86.exe would come into play as a limit. Since no other applications (read "foreground" applications) had a custom setting, they remained free to use all of the box without having to keep adding to ThreadMaster's "Exceptions" list. While looking at Task Manager on the dual-core box, I realized that while both cores were running a rosetta...exe process, and the graphical display of CPU usage show 100% on both cores, over on the Processes tab each process was reporting at approximately 50%. I'm guessing that ThreadMaster gets the same information (perhaps out of the WMI API). When I reset ThreadMaster to fire at a threshold percentage of 47%, and drop rosetta to 24%, I end up with two rosetta processes (5 threads each), both running pretty tightly at 23-26%, the box reporting the mid 50% and the fan speed back where it was when on single CPU. I've got one box running with ThreadMaster throttling, an identical box that is running limited to single CPU. I'm assuming that if I compare credits earned over the weekend, it should tell me which approach gets more Rosetta work out of the box at the desired fan noise level. Cheers, Alan