1) Message boards : News : New Data from AO (Message 1609847) Posted 6 Dec 2014 by Bjorn Munch Post: I tried again, and this time a got new tasks for both of them :-)

2) Message boards : News : New Data from AO (Message 1609799) Posted 6 Dec 2014 by Bjorn Munch Post: The BOINC's on my two computers hadn't communicated with S@H for 24 and 36 hours, resp. and have been running rosetta@home instead. So now I clicked Update to wake the project up. But still, both of them reported no new tasks available. Have they all been picked up already?

3) Message boards : Number crunching : Completed AstroPulse task has vanished (Message 1528103) Posted 14 Jun 2014 by Bjorn Munch Post: Oops! yes I knew they were cleaned up I just didn't think it happened that fast. Sorry about the false alarm then! Yes I can now see it has one valid AP.

4) Message boards : Number crunching : Completed AstroPulse task has vanished (Message 1528094) Posted 14 Jun 2014 by Bjorn Munch Post: I have just added another computer and have therefore been paying extra attention. I knew it has completed two AstroPulse tasks but when I looked at it on the web page now, to my surprise I only saw one. I know I saw it finished in just over 31 hours, the Event Log in BOINC says it was uploaded on the 12th at 04:19 UTC and it would have been reported shortly after. I am also positive I did see it on the web page, it was waiting for validation. But now it's vanished into thin air! :-O How can this happen? The task identifier was ap_02dc08ad_B0_P1_00220_20140610_30978.wu_0 but I don't know how I can look that up.

5) Message boards : Number crunching : New computer, same CPU but only half the performance? (Message 1526052) Posted 9 Jun 2014 by Bjorn Munch Post: Hmm OK maybe it's not all that different but it did look that way to me at first. And when I now look at more finished rosetta tasks, they look very similar actually. But it's still a fact that the benchmark numbers by BOINC are very different. And even after I compiled by banal square root program with maximum optimization, it still takes 50% longer on the new box.

6) Message boards : Number crunching : New computer, same CPU but only half the performance? (Message 1526045) Posted 9 Jun 2014 by Bjorn Munch Post: Processing times relative to credits do look quite different (also on rosetta) and as I've said, a banal benchmark shows a significant difference. I should mention that the BogoMIPS number reported during boot *is* almost exactly the same for both computers, indicating there is nothing inherently wrong with the CPU. One difference is in the "cpufreq kernel driver" as reported by "cpufreq-info -c 0 -d", which is powernow-k8 on the old and acpi-cpufreq on the new box. No idea what that means but I guess I should dig into that. I don't know how checking the temps would help. I don't have an actual temp, just some number that goes up when I activate all the tasks in BOINC. And this number is much lower that what is reported as critical or even high. For some reason I have lots more on my old desktop, including some genuine temperatures I believe.

7) Message boards : Number crunching : New computer, same CPU but only half the performance? (Message 1526038) Posted 9 Jun 2014 by Bjorn Munch Post: ASUS M5A97 R2.0. Both desktops are using the same. And both shipped *without* OS installed. On the old one, to install Linux I did have to hit some F key during boot in order to get it to boot from DVD but on the second one I just had to wait for DHCP to time out and then hit <enter>.

8) Message boards : Number crunching : New computer, same CPU but only half the performance? (Message 1526035) Posted 9 Jun 2014 by Bjorn Munch Post: I can't get any meaningful temp readings except I can see it goes up when I start running BOINC. As for those bios setting: I have no idea what that means or how to check.

9) Message boards : Number crunching : New computer, same CPU but only half the performance? (Message 1526028) Posted 9 Jun 2014 by Bjorn Munch Post: I have now started running BOINC with SETI/rosetta on my second desktop. It was not intentional, but I ended up bying one with the exact same CPU (6-core AMD FX-6300) and if it matters even the same mother board. But: the new box has only about half the floating point performance! BOINC benchmark shows 2617 MIPS per CPU on the old one, only 1366 on the new. And this can also been seen in the time it's spent on completed tasks. It's not just BOINC or SETI either, I can see it by running a banal test program which essentially adds 1 billion different square roots. The old box runs it in 5.624s, the new takes 12.564 so more than double. Another simple test with integer arithmetic shown similar differences, so it's not just floating point. I have confirmed that all cores are running at top speed 3.50 GHz all the time, so that's not the issue. And I can even hear the fan picking up when I start BOINC so I know it's working harder. :-) Any ideas? In case it matters: old computer is running Linux Mint 14, new is on Linux Mint 17.

10) Message boards : Number crunching : BOINC changed its mind and think I have only one CPU core, not 6 (Message 1525916) Posted 8 Jun 2014 by Bjorn Munch Post: But is says "Click Clear to restore web-based settings" and that's what I did. Also, it says that a setting 0 for this parameter would be ignored but apparently it wasn't. It's also strange that a fresh install of BOINC on the new box would create a config that by default does not utilize the CPU cores I have. But never mind, it works now. :-)

11) Message boards : Number crunching : BOINC changed its mind and think I have only one CPU core, not 6 (Message 1525868) Posted 8 Jun 2014 by Bjorn Munch Post: Thanks but I had actually solved it myself before I saw your reply. :-)

12) Message boards : Number crunching : BOINC changed its mind and think I have only one CPU core, not 6 (Message 1525865) Posted 8 Jun 2014 by Bjorn Munch Post: I just discovered: this was set to 0.00% in the local preferences in each BOINC. I changed that to 100% and now it runs 6 tasks on both computers! :-) But I think this is broken, it should do this by default. Somehow it didn't help to set it to 100% on the web.

13) Message boards : Number crunching : BOINC changed its mind and think I have only one CPU core, not 6 (Message 1525860) Posted 8 Jun 2014 by Bjorn Munch Post: It's also worth noting that the new computer was correctly registered as having 6 CPUs (cores) but then BOINC overrules that and decides it has only 1. The same happens when I run a CPU benchmark: I can observe it only runs it on a single core. Ah well, one good thing about this is that the old computer is much quieter now that the fan has slowed down....

14) Message boards : Number crunching : BOINC changed its mind and think I have only one CPU core, not 6 (Message 1525850) Posted 8 Jun 2014 by Bjorn Munch Post: But it *was* set to 1 even before this. And when I tried manually setting it to 6, BOINC wouldn't take it but reduced it to 1 anyway. Update: I now tried setting it to 0, it had no effect. BOINC still thinks I have only one CPU.

15) Message boards : Number crunching : BOINC changed its mind and think I have only one CPU core, not 6 (Message 1525846) Posted 8 Jun 2014 by Bjorn Munch Post: I have been running SETI@home (and later also rosetta) on my Linux desktop with a 6-core AMD since September. BOINC automatically determined it could run 6 tasks in parallel and has been doing that ever since. Now I've just added another linux box, also with a 6-core AMD, but this time BOINC would only run one task. I wondered why and looked into the Computing Preferences on the web. I noticed "On multiprocessors, use at most 1 processors" which was puzzling. I changed that to 6. But then when I do "Computing Preferences -> Clear" in BOINC on the old desktop, it says in the Event Log "Number of usable CPUs has been changed from 6 to 1" and all of a sudden, that too only runs one task! :-( What's going on here? How can I make it understand it has 6 cores available??? Old box: Linux Mint 14, BOINC Manager 7.2.33 New box: Linux Mint 17, BOINC Manager 7.2.42 Old box: http://setiathome.berkeley.edu/show_host_detail.php?hostid=7105433 - Bjorn in Norway

16) Message boards : News : We are now mostly recovered from a campus wide power outage. (Message 1428019) Posted 13 Oct 2013 by Bjorn Munch Post: FWIW, my BOINC just got a new AstroPulse task for the first time since the outage. :-) Not that it had meny before that either, I just started two weeks ago and am now also running Rosetta@home. Still awaiting validation for two AP's that I finished before the outage.

17) Questions and Answers : Unix/Linux : Trying to move slots dir to /tmp doesn't work (Message 1425012) Posted 6 Oct 2013 by Bjorn Munch Post: And that's exaclty what I did: copied everything over to /tmp/BOINC, corrected the path in the startup script and now it's churning along happily. I sync back to disk *except* the slots directory once every hour. Saves the disk from having to work 24/7 and it might speed things up slightly too. I have 16 Gb memory so no danger of running out of space on /tmp :-)

18) Questions and Answers : Unix/Linux : Trying to move slots dir to /tmp doesn't work (Message 1424578) Posted 5 Oct 2013 by Bjorn Munch Post: Well, a softlink named slots should normally have worked but I guess the system makes some assumptions about relative paths which are not correct with that setup. I had no trouble moving from the SSD to the harddisk. I suspended running tasks, moved everything over, edited a few files then started up again. I could probably copy everything to /tmp/BOINC and run some hourly cron job to rsync stuff back to disk. But I'll let it run as is for now. I joined rosetta@home too so that my CPU can do something useful while waiting for SETI to accept me back. :-)

19) Questions and Answers : Unix/Linux : Trying to move slots dir to /tmp doesn't work (Message 1424507) Posted 5 Oct 2013 by Bjorn Munch Post: I started running SETI@home on my Linux Mint 14 desktop a week ago. Seeing that the 6 concurrent tasks write to files in the slots directory all the time, I moved the stuff over from the SSD to the hard disk. But then I thought maybe it would be better to have it on /tmp to avoid writing to disk at all during processing. I first tried the obvious: mv slots to /tmp/BOINC/slots and replace with a softlink. That caused all tasks to crash after restart, and the error is "execve: no such file or directory". Then I tried creating softlinks *back* to the hard disk from /tmp/BOINC/ except for slots and starting the manager from there. Same result, and now it seems my computer is temporarily banned after reporting 40 errors in a row. :-( Luckily I still have 2 tasks left which I can report in 3+ hours and then hopefully I will be allowed some more. Of course having slots in /tmp means all running tasks are lost if the computer crashes, but that would probably happen anyway. In any case, that happens so rarely it's not a big problem. For a controlled shutdown I could copy over to disk after stopping the tasks. Is there any way to have slots placed under /tmp short of copying *everything* over to /tmp before starting? And yes my /tmp is on tmpfs so it really is in memory, not on disk. :-)

 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.