| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: Suggestions / Feedback
Thread: Speeding up BOINC - how to find out whether it works |
| 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 |
Hello erveryone,
As far as I know, for now, there are four possibilities to speed up BOINC: 1. Use Dotsch operating system instead of your normal one 2. Change the priority of the BOINC process 3. Attach each WCG process (workunit) to a single core 4. Use your graphic processor (GPU) and your CPU (normal processor) instead of using only the CPU. But: How can I find out, whteher this speeding up methods work? How can I find out, how much they speed BOINC up. I thought that running Benchmarks could help, but in the end I came to the conslusion, that the runnig benchmarks dose not realy help, as the processor s capabilities stay the same, no matter whether I run normal Ubuntu or Dotsch. So how can I compare BOINCs speed under Dotsch operating system to its speed under normal Ubuntu or Windows? Thank you very much for any comments and answers. Martin |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello Martin.Schnellinger,
Comparing speed is frustrating since the only precise comparison is comparing the exact same work unit for a project in different situations, and this is not possible. Your method 1 is . . . uncertain. Linux 64 seems more efficient than Windows, but most of the time is spent running the compiled project code, not the OS. Method 2 does not really work, because the priority is set in the code. People who think that they are overriding the coded priority using OS commands are fooling themselves. Method 3 is very OS-dependent. It could speed things up slightly, but that depends on many factors Method 4 is not possible with the current compiled projects. We hope to experiment with this later in 2011, but it will be a while before many projects are coded to run on GPUs. Currently, the most hopeful speedup for any particular hardware setup seems to be running Linux 64. Lawrence |
||
|
|
Crystal Pellet
Veteran Cruncher Joined: May 21, 2008 Post Count: 1320 Status: Offline Project Badges: 
|
But: How can I find out, whteher this speeding up methods work? How can I find out, how much they speed BOINC up. . . So how can I compare BOINCs speed under Dotsch operating system to its speed under normal Ubuntu or Windows? Thank you very much for any comments and answers. Martin Hello Martin, As Lawrence said, you only can test this properly with the same task. To do this: 1. Finish running tasks and suspend the others. After finish report them. 2. Stop the Boinc client. 3. Make a copy of the whole Boinc data root to e.g. usb-stick. 4. Start the Boinc client for your proper test1. 5. Start 1 task for your measurement. Up to you if you want to report it, when finished. 6. Stop BOINC and restart your system with your second choice OS. 7. Make a backup of BOINC-data for restoring later. 8. Overwrite the BOINC-data directory with your copy. 9. Start the Boinc client for your proper test2. 10. Start the same task for your measurement. If reported earlier, it's useless to report it once again. During all the tests do nothing else with the machine of course and perhaps pull even your internet-connection. In principle it's even possible to do this comparing between Linux and Windows, but then you've to hack several xml-files .... Good luck with that  The difference is in general is so big, that you can read it here on the forum.For comparing tests with renicing tasks, fixing tasks to cores, etc. the procedure is similar, but then on the same machine/OS-combination. CP |
||
|
|
|