| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: Suggestions / Feedback
Thread: Request more even CPU usage. Fan speed pumping noise excessive. |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 3
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I have WCG running on two computers, one of which is a powerful one I personally use, while the other is mainly idle but does our NAS to NAS backups.
The other computer has its fans set to run for maximum coolness, which means it runs constantly and noisily, irrespective of CPU usage, but in a room where no one can hear it. However, my personal worstation was built for silent operation (dual boots to a digital audio workstation - DAW), so the fan speeds are sensitive to CPU usage. I like not hearing the fans in normal non-DAW usage as well. When running WCG, my computer fan's speed pumps at about once per second, and is audibly annoying. The CPU Performance graph in Task Manager showed spikes that track the fan pumping. Trying to limit it cannot make it any smoother, but just lowers the amplitude of the spikes (when limiting CPU processor numbers %). Is there a way to actually make the CPU usage even so that the fans are not speed pumping so much? I suspect that such fan usage will not help their long-term survival. If there are no settings that can do this, perhaps some better micromanagement in the design may enable instantaneous usage to have a better match to average use, as the 'bull-at-a-gate --> hit-limit --> stop completely --> rinse-and-repeat' cyclical type of processing is too crude (and annoying). Until something changes in this regard, I have suspended WCG processing on my computer. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
My favored program on Windows to control fanspeed/heat is in fact a cycle controller originally written to manage server loads: ThreadMaster GUI [See Start Here Forum]. The beauty is, it slows down BOINC to give a very even processing load, yet you retain full horse power / turbo boost when needed. The second, even more enjoyable part is, you can set % on a -per-science_app- basis. Some work the CPU harder than others. Anyway, TMG gives a really smooth result which the BOINC internal throttle does not [it alternates processes on and off rounded on whole seconds, so 50% is running 1 second, pausing 1 second, which is the best for laptops BTW, as it stops fan revving up and down]
TThrottle a program written for BOINC does not work optimally, particularly on hyperthreaded systems. Seen dropping computing to 25% because the nature of the HT beast is to allocate spare cycles to running processes i.e. slow BOINC down to 50% and the OS will allocate that back to any running program or service that needs it. In other words, at 50% the other 50% will absorb the spare cycles, so they still run at 100%. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
BTW, there briefly was a BOINC test version that worked at sub-second throttling, so 60% would mean, pause 0.4 seconds, run 0.6. [TMG works at the millisecond level] opposed to pause 2 seconds, run 3 seconds. That feature was pulled as it required the 100s of projects to recompile their science apps as well because they only act at whole second basis [it tested well for WCG though, at least for me]. Since the driving force behind this left WCG [the person pushing on the Berkeley developers], nothing heard of this again.
|
||
|
|
|