Message boards :
Number crunching :
Problem on Linux machine
Message board moderation
| Author | Message |
|---|---|
 Mark_SchimmerSend message Joined: 5 Aug 05 Posts: 3 Credit: 37,503 RAC: 0 |
Hi 'ya, first of all: I'm new to boinc, so please be patient ;-) I'm running boinc 4.43 on a Linux machine version SUSe 2.6.11.4-20a on a pentium 4 with 2.6 GHz / 1 Gig RAM. The machine is way too slow, and on the benchmark a mere 737 MFIOps are returned (my other 2.4 GHz ordinateur has 2.11 GFIOps ). I also noticed that on my 2.4 GHz WinXP System there are two concurrent workunits boing :-) calculated, and on the Linux machine only one at a time. Does this sound familiar to anyone? Cheers, Mark |
 Send message Joined: 18 Sep 04 Posts: 163 Credit: 1,682,370 RAC: 0 |
Please unhide your computers, so we do no need to guess too much. Nonetheless. The 2.4 machine is no HT and the 2.6 is a HT enabled processor? If so, please check if you installed the SMP kernel. 2.6.11.4-20a sounds as it would be a non-SMP kernel. AFAIK SUSE names them accordingly. HTH Michael Team Linux Users Everywhere 
|
|
Mark_Schimmer Send message Joined: 5 Aug 05 Posts: 3 Credit: 37,503 RAC: 0 |
Michael, I unhid the computers in my profile, but be aware, it is not up to date. You were right, the kernel was the non-smp. I switched to the smp kernel and it is now the 2.6.11.4-20a-smp . I ran a new benchmark after switching and the FPU Mark is still just 736,7 MFIOps. But at least I run 2 concurrent wus now. tks for your advice, Mark |
littleBouncer Send message Joined: 23 Oct 04 Posts: 358 Credit: 1,439,205 RAC: 0 |
<blockquote>Please unhide your computers, so we do no need to guess too much. Nonetheless. The 2.4 machine is no HT and the 2.6 is a HT enabled processor? If so, please check if you installed the SMP kernel. 2.6.11.4-20a sounds as it would be a non-SMP kernel. AFAIK SUSE names them accordingly. HTH Michael</blockquote> @ Michael, Sorry but it is the other way: the 2.4 GHz machine is a HT and has 2 virtual CPU's, that's why he is chrunching 2 WU's (info for Mark). For the 'linux-Questions': that is not my matter ... greetz littleBouncer   
|
|
Mark_Schimmer Send message Joined: 5 Aug 05 Posts: 3 Credit: 37,503 RAC: 0 |
Both machines are HT P4s. On linux: are there any experiences with priorities? I noticed, that the boinc process, running with prio 0 , submits the wu with prio 19, which is quite low. Since the machine is doing nothing else, I resetted it to prio -11, after which it slowed down otherwise considerably. Now my userload is huge, compared to before. Maybe this will improve performance a little! Cheerz, Mark |
|
Send message Joined: 16 Jul 05 Posts: 84 Credit: 1,875,851 RAC: 0 |
The BOINC benchmark and the sixtrack performance are two different shoes. It's normal, that the Linux BOINC benchmarks are lower. Linux Users Everywhere @ BOINC [url=http://lhcathome.cern.ch/team_display.php?teamid=717] |
|
Send message Joined: 18 Sep 04 Posts: 163 Credit: 1,682,370 RAC: 0 |
<blockquote> Both machines are HT P4s. On linux: are there any experiences with priorities? I noticed, that the boinc process, running with prio 0 , submits the wu with prio 19, which is quite low. Since the machine is doing nothing else, I resetted it to prio -11, after which it slowed down otherwise considerably. Now my userload is huge, compared to before. Maybe this will improve performance a little! Cheerz, Mark </blockquote> nice 19 is the way to go. This ensures that the CPU is used if nothing else must be done, like swapping, IO, user tasks, etc. If you are not using the PC otherwise, you should see no difference in runtimes between nice 0 and 19. Setting values lower than 0 is not advised. They are reserved for system processes. BOINC is resetting prio to 19 at the start of the next WU nonetheless... Michael Team Linux Users Everywhere
|
©2025 CERN