Message boards :
Number crunching :
Bugs and Credits
Message board moderation
Author | Message |
---|---|
Eric Korpela ![]() Send message Joined: 3 Apr 99 Posts: 1382 Credit: 54,506,847 RAC: 60 ![]() ![]() |
Hello Everybody, Sorry for my long absence from the discussion. I picked up a stomach bug that had me pretty well incapacitated for the last week. I'm still down several kilos of water weight, so I'm not yet back up to my usual work hours. Regarding bugs in the enhanced client, there are still some that we are working on. The fix for the jpeg loading bug didn't cure the entire problem, but it did make crashes less common. A possible fix for the "hang on overflow" bug is going to go out to beta soon. I see there are still a lot of discussions about the new credit system and that a lot of these have degenerated into flame wars. I don't think there is any way to avoid flame wars in any public forum. I would encourage everyone to be civil and to withold accusations. We don't intend to police these discussions or get into the middle of them. For now the credit system is what it is. We are investigating the cross-project calibration to see if the credit claims need adjusting. This process will be ongoing rather than a one time deal as new computer architectures become available and new projects arrive. Don't expect weekly or monthly adjustment. Annually or semi-annually is more likely. We don't intend to police the credit claims of optimized versions unless problems with exaggerated claims become severe. That doesn't mean people should feel free to develop versions of the application that claim outrageous amounts of credit. If such versions become widespread, we would likely be forced to remove all credit claim calculation from the application and turn it into a server side determination based upon an estimation of the work a result would require. I would prefer that we remain with the current method of measuring the actual work done. My hope is that porters/optimizers will self-police. I assume we're mostly honorable people here. If someone is doing something wrong they know it and my hope is they would stop without being requested or forced to. Accidental problems should be fixed without anyone needing to intervene. There will always be a few people with malicious intent. Most will disappear if ignored. @SETIEric@qoto.org (Mastodon) ![]() |
![]() ![]() Send message Joined: 5 Jul 99 Posts: 4548 Credit: 35,667,570 RAC: 4 ![]() |
Hi Eric , I'm sorry to hear about your illness I hope you get better soon try to get some rest this 3 day Memorial Holiday weekend Thank you very much for all your hard your hard work and long hours. for --- SETI@home ---- I would like to congratulate you and your colleagues on perhaps the most significant computing effort and perhaps the most significant cooperative effort in human history. --- SETI@home --- Best Wishes Byron Vancouver Canada |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 ![]() |
Eric: Applause! Treating all participants as adults even when they seem to be acting infantile is very sensible. Joe |
John McLeod VII Send message Joined: 15 Jul 99 Posts: 24806 Credit: 790,712 RAC: 0 ![]() |
The one claim that I have seen that needs investigation is that long running results claim less than they should per hour and/or short running results claim more than they should per hour. ![]() ![]() BOINC WIKI |
![]() Send message Joined: 2 Aug 00 Posts: 1851 Credit: 5,955,047 RAC: 0 ![]() |
Yes, I've, by quick inspection, seen a considerable disparity of points awarded with different units on my two P950s, from eight to 24 credits per hour per core. I don't think all that variation is due to the two machines or to conditions. And yes, shorter units seem to yield more cobblestones per hour of computation. So this seems to indicate that the true work done by the computer might not best be measured by the number of flops (or fpops) done. But maybe fpops are the real thing. Memory transfers, even though necessary, maybe shouldn't be counted as work. Maybe fpops correspond to getting that box from point A to point B but memory transfers only correspond to moving that box to the side to allow another box to fit into the plane or truck. Maybe something else oughta be counted as work, too, but I wouldn't know what that would be. |
![]() Send message Joined: 21 Apr 00 Posts: 1459 Credit: 58,485 RAC: 0 ![]() |
Memory transfers, even though necessary, maybe shouldn't be counted as work.faster memory = more work shortly after boinc was launched, there was dicussion of credit, at some point, including things like memory bandwidth, network throughput, hard drive space donated, etc. Want to search the BOINC Wiki, BOINCstats, or various BOINC forums from within firefox? Try the BOINC related Firefox Search Engines |
W-K 666 ![]() Send message Joined: 18 May 99 Posts: 19480 Credit: 40,757,560 RAC: 67 ![]() ![]() |
Yes, I've, by quick inspection, seen a considerable disparity of points awarded with different units on my two P950s, from eight to 24 credits per hour per core. I don't think all that variation is due to the two machines or to conditions. And yes, shorter units seem to yield more cobblestones per hour of computation. So this seems to indicate that the true work done by the computer might not best be measured by the number of flops (or fpops) done. But maybe fpops are the real thing. Memory transfers, even though necessary, maybe shouldn't be counted as work. Maybe fpops correspond to getting that box from point A to point B but memory transfers only correspond to moving that box to the side to allow another box to fit into the plane or truck. Maybe something else oughta be counted as work, too, but I wouldn't know what that would be. As Lee mentioned there have been many discussions, not only on Seti but other projects as well, when two cpu's share memory it always seems one cpu will do worse than the other if they are doing the same of very similar tasks. Even on My Dual P3 which is two separate cpu's there is an average slowdown if I do two units from the same project at the same time. So here on my Dual P3 and my son's P4 HT we arrange tha the cpu's are doing different projects, and the efficiency is increased ~15%. Not quite the times we get by only employing only one cpu but better that doing two from the same project at once. Moving boxes takes time, if it is the cpu cache memory maybe not much, but if it requires a move to or from main memory quite a lot in relative terms. Even if your cpu is operating at 3 GHz or higher the true speed of your memory is probably 200 MHz, 15 times slower, not counting the delays as it switches from read to write and from columns to rows etc. Those numbers for your memory like 4,4,4,12 are delays at memory bus speeds. Andy |
![]() ![]() Send message Joined: 26 Jun 99 Posts: 139 Credit: 9,063,168 RAC: 15 ![]() ![]() |
Speaking of bugs, i have ran into one with the installer. I have been upgrading the farm and one box will just not finish up the install process with the new boinc client. I get this message that the msi file is invalid after it has almost finished the entire install. :( I have tried re-downloading (obvious) but did not help. I have tried the different options within the installer but that did not help either. I have tried moving the msi file to the root of C thinking that maybe the path was getting to long. did not help. And no... boinc is not running while i am trying this. Ideas? -searcher ![]() |
Eric Korpela ![]() Send message Joined: 3 Apr 99 Posts: 1382 Credit: 54,506,847 RAC: 60 ![]() ![]() |
This is the BOINC installer? Which version are you trying to install? Eric Speaking of bugs, i have ran into one with the installer. @SETIEric@qoto.org (Mastodon) ![]() |
Berserker Send message Joined: 2 Jun 99 Posts: 105 Credit: 5,440,087 RAC: 0 ![]() |
Speaking of bugs, i have ran into one with the installer. I have come across cases where msi can get confused if a previous version of software did not uninstall properly for some reason (and you may not even be aware that this has happened). This was not with BOINC - it happened with software I work on. Have you had a previous version of BOINC on this machine? Stats site - http://www.teamocuk.co.uk - still alive and (just about) kicking. |
![]() Send message Joined: 2 Aug 00 Posts: 1851 Credit: 5,955,047 RAC: 0 ![]() |
Memory transfers, even though necessary, maybe shouldn't be counted as work.faster memory = more work Lee and Andy: Thanks for the info. Just before Enhanced came out I studied crunchtimes of the top computers accessible from the frint page and concluded that the throughput of four-thread computers was less than that of two-thread ones. That's the reason I bought two separate boxes, each with a Pentium D950. To economize I used a 'Net configurator (CyberPower). I had to rely on a salesman who may have had some technical knowledge. I made several goofs, one of which might be the memory. Corsair Value-Select DDR2 may have a CAS-latency of a lousy 5 (even though it is a fast 667) but that was not apparent to me at the time. It seems like the hard drive, cable modem, network interface like a router and monitor switch would affect crunchtimes little but I could be wrong. I had to be on my toes to get two sticks of memory for each box to afford dual-channel transfer. I had to catch myself in order to get a digital monitor, but, the guys at Best Buy never addressed the fact (I believe) that digital monitors need a special 8X3 wire interface and the monitor switch Best Buy sold me has no such thing, just the two-wire-rows thing. So I still get an analog display - and - the thing - or Windows- always pops out of the resolution I set (native 1440 x 900). |
![]() ![]() Send message Joined: 26 Jun 99 Posts: 139 Credit: 9,063,168 RAC: 15 ![]() ![]() |
yes it has been crunching for a long time. current version is 5.2.13 I was trying to install the latest version (5.4.9) of the boinc software for windows. Maybe I can have it complete all the work currently underway then un-install it fully and delete any directory leftovers. This is not a very fast cruncher so no big loss if it eventually is phased out. :) -searcher Speaking of bugs, i have ran into one with the installer. ![]() |
W-K 666 ![]() Send message Joined: 18 May 99 Posts: 19480 Credit: 40,757,560 RAC: 67 ![]() ![]() |
[quote]Lee and Andy: Thanks for the info. Just before Enhanced came out I studied crunchtimes of the top computers accessible from the frint page and concluded that the throughput of four-thread computers was less than that of two-thread ones. That's the reason I bought two separate boxes, each with a Pentium D950. To economize I used a 'Net configurator (CyberPower). I had to rely on a salesman who may have had some technical knowledge. I made several goofs, one of which might be the memory. Corsair Value-Select DDR2 may have a CAS-latency of a lousy 5 (even though it is a fast 667) but that was not apparent to me at the time. It seems like the hard drive, cable modem, network interface like a router and monitor switch would affect crunchtimes little but I could be wrong. I had to be on my toes to get two sticks of memory for each box to afford dual-channel transfer. I had to catch myself in order to get a digital monitor, but, the guys at Best Buy never addressed the fact (I believe) that digital monitors need a special 8X3 wire interface and the monitor switch Best Buy sold me has no such thing, just the two-wire-rows thing. So I still get an analog display - and - the thing - or Windows- always pops out of the resolution I set (native 1440 x 900). If you need info on how memory works, and you have some time, see corsair memory_basics Andy |
Berserker Send message Joined: 2 Jun 99 Posts: 105 Credit: 5,440,087 RAC: 0 ![]() |
yes it has been crunching for a long time. current version is 5.2.13 Not a bad idea. If that doesn't work, there's a Microsoft tool that can clean up broken MSI data, but it's best kept for use as a last resort (it just deletes the broken stuff, so it's more like a sledgehammer than a scalpel). Stats site - http://www.teamocuk.co.uk - still alive and (just about) kicking. |
Bob Guy Send message Joined: 7 Sep 00 Posts: 126 Credit: 213,429 RAC: 0 ![]() |
I was trying to install the latest version (5.4.9) of the boinc software for windows. Please read this thread, see my response there. I had the same trouble - easy to fix. |
Berserker Send message Joined: 2 Jun 99 Posts: 105 Credit: 5,440,087 RAC: 0 ![]() |
I was trying to install the latest version (5.4.9) of the boinc software for windows. Thanks. Just spotted that thread and was about to come here and post the same as you did. The WICU was the utility I was referring to. Steps are:
Stats site - http://www.teamocuk.co.uk - still alive and (just about) kicking. |
![]() ![]() Send message Joined: 26 Jun 99 Posts: 139 Credit: 9,063,168 RAC: 15 ![]() ![]() |
Thanks all! Esp the link to the other thread. I did the uninstall with the old installer then ran the new installer. Worked like a charm. I hope they fixed the installer in this release not to look for the temp msi files. Oh well, gave me the opportunity to move the app off my C drive and onto my bigger E disk. :) Cheers!! -searcher I was trying to install the latest version (5.4.9) of the boinc software for windows. ![]() |
![]() ![]() Send message Joined: 30 Dec 04 Posts: 29 Credit: 700,896 RAC: 0 ![]() |
-I’ am pissed behave of Crunch3r's good work not getting the credit he supposed to. Next week all my seti computers (20) are on NO NEW WORK , until he gets a APOLOGY from all people on this forum that got him pissed of and seti administrators. -A second reason is I got 5 work units yesterday, about 45 hours work/units on my AMD Opteron 165 computer. I have got only 5/20 optimized computers. Sucks. -Until we get back our 4 hour Crunch units! cya2 : Turku, Finland Comments to: thomas.bymark@pp.inet.fi http://www.boincstats.com/ : SETI@HomeCredit/day 2,833 Position based on Total Credit 1484, and going down next week (seti). Until we meet again................... ( SETI@home enhanced, sucks, as now) ![]() |
![]() Send message Joined: 2 Aug 00 Posts: 1851 Credit: 5,955,047 RAC: 0 ![]() |
[quote]Lee and Andy: Thanks for the info. Just before Enhanced came out I studied crunchtimes of the top computers accessible from the frint page and concluded that the throughput of four-thread computers was less than that of two-thread ones. That's the reason I bought two separate boxes, each with a Pentium D950. To economize I used a 'Net configurator (CyberPower). I had to rely on a salesman who may have had some technical knowledge. I made several goofs, one of which might be the memory. Corsair Value-Select DDR2 may have a CAS-latency of a lousy 5 (even though it is a fast 667) but that was not apparent to me at the time. It seems like the hard drive, cable modem, network interface like a router and monitor switch would affect crunchtimes little but I could be wrong. I had to be on my toes to get two sticks of memory for each box to afford dual-channel transfer. I had to catch myself in order to get a digital monitor, but, the guys at Best Buy never addressed the fact (I believe) that digital monitors need a special 8X3 wire interface and the monitor switch Best Buy sold me has no such thing, just the two-wire-rows thing. So I still get an analog display - and - the thing - or Windows- always pops out of the resolution I set (native 1440 x 900). Thanks, Andy. When I can get to it...... |
©2025 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.