| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Got a problem |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 22
|
|
| Author |
|
|
Movieman
Veteran Cruncher Joined: Sep 9, 2006 Post Count: 1042 Status: Offline |
One of the machines I run is having issues with the awarded points.
----------------------------------------The machine is running BOINC 5.10.13 An 8 clore clovertown machine running at 3157mhz Here's the benchmark: Boinc 5.10.13 at 3157mhz: 8/16/2007 6:02:16 PM||Benchmark results: 8/16/2007 6:02:16 PM|| Number of CPUs: 8 8/16/2007 6:02:16 PM|| 3011 floating point MIPS (Whetstone) per CPU 8/16/2007 6:02:16 PM|| 9356 integer MIPS (Dhrystone) per CPU I happened to wonder why it wasn't receiving what I thought it would and looked at a few finished WU today. It does just FAAH WU. This is a few of the ones I saw: Result Name Status Sent Time Time Due / Return Time CPU Time (hours) Claimed/ Granted BOINC Credit faah2098_ d087n758_ x2AZC_ 04_ 1-- Valid 08/16/2007 23:00:06 08/17/2007 08:30:06 1.89 26.9 / 26.9 faah2098_ d087n758_ x2AZC_ 04_ 0-- Valid 08/16/2007 23:00:01 08/19/2007 00:55:27 1.89 43.4 / 26.9 <-me ----------------------------------------------------------------------------------------------------------- Result Name Status Sent Time Time Due / Return Time CPU Time (hours) Claimed/ Granted BOINC Credit faah2098_ d087n711_ x2AZC_ 01_ 0-- Valid 08/16/2007 23:02:28 08/17/2007 21:52:48 13.11 77.0 / 77.0 faah2098_ d087n711_ x2AZC_ 01_ 1-- Valid 08/16/2007 23:00:01 08/19/2007 00:55:27 4.13 94.5 / 77.0<-me --------------------------------------------------------------------------------------------------------- Result Name Status Sent Time Time Due / Return Time CPU Time (hours) Claimed/ Granted BOINC Credit faah2098_ d087n734_ x2AZC_ 01_ 1-- Valid 08/16/2007 23:02:56 08/17/2007 06:08:52 2.28 40.5 / 40.5 faah2098_ d087n734_ x2AZC_ 01_ 0-- Valid 08/16/2007 23:00:01 08/19/2007 00:55:46 2.39 54.9 / 40.5,<-me ---------------------------------------------------------------------------------------------------------- Result Name Status Sent Time Time Due / Return Time CPU Time (hours) Claimed/ Granted BOINC Credit faah2098_ d087n752_ x2AZC_ 00_ 0-- Valid 08/16/2007 23:03:22 08/18/2007 02:25:12 6.00 79.2 / 83.2 faah2098_ d087n752_ x2AZC_ 00_ 1-- Valid 08/16/2007 23:00:01 08/19/2007 00:55:27 3.81 87.2 / 83.2<-me --------------------------------------------------------------------------------------------------------- Result Name Status Sent Time Time Due / Return Time CPU Time (hours) Claimed/ Granted BOINC Credit faah2098_ d087n752_ x2AZC_ 01_ 1-- Valid 08/16/2007 23:02:25 08/17/2007 17:54:19 10.65 69.3 / 69.3 faah2098_ d087n752_ x2AZC_ 01_ 0-- Valid 08/16/2007 23:00:01 08/19/2007 00:55:27 3.86 88.4 / 69.3<-me --------------------------------------------------------------------------------------------------------- Result Name Status Sent Time Time Due / Return Time CPU Time (hours) Claimed/ Granted BOINC Credit faah2098_ d087n757_ x2AZC_ 03_ 0-- Valid 08/16/2007 23:03:17 08/18/2007 06:24:27 2.75 35.1 / 37.7 faah2098_ d087n757_ x2AZC_ 03_ 1-- Valid 08/16/2007 23:00:01 08/19/2007 00:55:27 1.76 40.3 / 37.7<-me --------------------------------------------------------------------------------------------------------- Result Name Status Sent Time Time Due / Return Time CPU Time (hours) Claimed/ Granted BOINC Credit faah2098_ d087n758_ x2AZC_ 03_ 1-- Valid 08/16/2007 23:00:01 08/19/2007 00:55:27 1.89 43.3 / 29.5 faah2098_ d087n758_ x2AZC_ 03_ 0-- Valid 08/16/2007 22:59:49 08/18/2007 22:05:34 1.63 29.5 / 29.5<-me ------------------------------------------------------------------------------------ In all cases except WU #3+5 I'm being awarded the points of the lower/slower machine. I thought that an average of the 2 machines "claimed" points was used as it was with the 3 member quorum? That's what I see on result #3+5 down on the list but not the others. What's changed or what am I missing here? Then on #6 the other gentleman claimed more than I did and WCG awarded him what I claimed. My gut tells me that the servers are seeing my claims as outliars and therefore not averaging the 2 scores as they should but giving me the lower claimed score. This machine is rock solid. ALL it does 24/7 since last December is WCG. BOINC 5.10.13 is installed and is exactly as downloaded from Berkley's servers. OS is Win2K3 Ent 64 bit with all the updates done. Thank you for reading.  [Edit 4 times, last edit by Movieman at Aug 19, 2007 3:00:09 AM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I thought that an average of the 2 machines "claimed" points was used as it was with the 3 member quorum? With a quorum of 2, BOINC normally takes the lower of the two claims rather than an average. On some WUs that I've looked at, if it's been sent out 3 times (because a host errored) then the granted credit seems to be the average of the two good claims. |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
knreed published a description of the rule that was also added to the FAQ's: http://www.worldcommunitygrid.org/forums/wcg/printpost?post=120888
----------------------------------------If the deviations are small, the average is taken, if the deviation is large it looks at which machine is closer to it's latest claim average per second. The one in the first sample must have been closer to his/her average claim. That said, who ever it is, is running with a dubiously low benchmark.... some just crunch for the crunch. Added: The Missed link
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 2 times, last edit by Sekerob at Aug 19, 2007 10:53:06 AM] |
||
|
|
Movieman
Veteran Cruncher Joined: Sep 9, 2006 Post Count: 1042 Status: Offline |
knreed published a description of the rule that was also added to the FAQ's: http://www.worldcommunitygrid.org/forums/wcg/printpost?post=120888 If the deviations are small, the average is taken, if the deviation is large it looks at which machine is closer to it's latest claim average per second. The one in the first sample must have been closer to his/her average claim. That said, who ever it is, is running with a dubiously low benchmark.... some just crunch for the crunch. Added: The Missed link Thanks for the reply Sekerob. I hadn't seen that post of Kreeds and it helps explain things. Now what I am still unclear on is whether Boinc is awarding based on MY average claim per sec in my history on that machine or on the average of that "type" of cpu's average history claim. If it is just looking at mine then the issue will take care of itself over time as this benchmak is higher than what was being used with the previously installed version of BOINC 5.4.11 With 5.4.11 I was getting a bench of 2968/6204 and that would be the factor that is skewing things now. If that is the case, then as I complete more WU the "history"that BOINC sees should gradually see a higher claimed points per second? Is my logic corrent on this? Thanks for reading. |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
If you look in other BOINC points announcement, you'll see that the system keeps track of the claims for the last 15-20 results. If the latest claim deviates from YOUR claim average, the claim also gets normalized. You can see that happening with FAAH, but I've not seen it with HPF2 which uses a different rule given the quorum is 15. After, the personal normalizing, it looks at who's closest to the personal average IF the 2 claims are far apart. Because of these normalizations, the deviations considered are actually very small.
----------------------------------------If your claim would have been the closer of the 2, you'd gotten your claim and the lower one, who seemingly is not interested, would get your credit value. At any rate, I see this frequently. You are correct in your logic in that the machine average is a sliding average, so eventually it finds its balance again, were it not that the benchmark is varying somewhat when it runs every 5 days. And to answer those whose wheels started spinning between the ears: No that re-benchmarking cannot be stopped. BOINC checks this as some have tried to fudge the number and set the benchmark date ever shifting forward. There's complex hashing that would force an immediate test. Edit: 15-20 not 150-20 in first line.
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 1 times, last edit by Sekerob at Aug 19, 2007 7:18:16 PM] |
||
|
|
olympic
Senior Cruncher Joined: Jun 12, 2005 Post Count: 156 Status: Offline |
Your Dhrystone is way too high for the speed you are running, atleast compared to the 8000 I get with my Quad at 3.5GHz on BOINC 5.10.13. Hopefully it will normalize out like Sek suggested, if not it may be best to downgrade to an older BOINC version. You'll get a lower benchmark but atleast you'll get the quorum averaged instead of the lowest claim.
---------------------------------------- [Edit 1 times, last edit by olympic at Aug 19, 2007 11:58:40 PM] |
||
|
|
Movieman
Veteran Cruncher Joined: Sep 9, 2006 Post Count: 1042 Status: Offline |
If you look in other BOINC points announcement, you'll see that the system keeps track of the claims for the last 15-20 results. If the latest claim deviates from YOUR claim average, the claim also gets normalized. You can see that happening with FAAH, but I've not seen it with HPF2 which uses a different rule given the quorum is 15. After, the personal normalizing, it looks at who's closest to the personal average IF the 2 claims are far apart. Because of these normalizations, the deviations considered are actually very small. If your claim would have been the closer of the 2, you'd gotten your claim and the lower one, who seemingly is not interested, would get your credit value. At any rate, I see this frequently. You are correct in your logic in that the machine average is a sliding average, so eventually it finds its balance again, were it not that the benchmark is varying somewhat when it runs every 5 days. And to answer those whose wheels started spinning between the ears: No that re-benchmarking cannot be stopped. BOINC checks this as some have tried to fudge the number and set the benchmark date ever shifting forward. There's complex hashing that would force an immediate test. Edit: 15-20 not 150-20 in first line. Thanks Sekerob, As always you give good clear info that answers my questions. Very much appreciated. If I understand all of this correctly assuming my benches don't vary greatly and they shouldn't as all that machine does is WCG then I should be out of this mess in a day as boinc recognises my claim as the norm and goes back to hopefully averaging the 2 submitted WU's. I never manually bench the machine anymore since the Intel multi core glitch was fixed so I'll leave it alone and check in a few days. Again my thanks for your help and time. Now Olympic: Perhaps yours isn't claiming high enough!  Go check the thread on Xs showing the benches on highly clocked quads and you'll see that mine is lower than those. The dual socket machines take a hit due to memory and buss speed. This is probably one of the 10 fastest clovers in the world but on a core to core basis those quads kill me easily. I'm limited to DDR2-667FBDimms at 5-5-5-15-19 and the quads can use up to DDR2-1200 with cas4 latency's...MUCH faster and FSB close to 2000 vs my 1400.. All my best pal. |
||
|
|
retsof
Former Community Advisor USA Joined: Jul 31, 2005 Post Count: 6824 Status: Offline Project Badges: 
|
Now Olympic: Perhaps yours isn't claiming high enough! Go check the thread on Xs showing the benches on highly clocked quads and you'll see that mine is lower than those. The dual socket machines take a hit due to memory and buss speed. This is probably one of the 10 fastest clovers in the world but on a core to core basis those quads kill me easily. I'm limited to DDR2-667FBDimms at 5-5-5-15-19 and the quads can use up to DDR2-1200 with cas4 latency's...MUCH faster and FSB close to 2000 vs my 1400.. All my best pal. So my AMD steam engines aren't doing too badly compared to those Intel strap-on diesels, eh? I have two 2.4GHz with DDR1-400-3200 and one newer 2.0GHz DDR1-<400?-1800 OC nicely to 2.4GHz running cooler than the others. Benchmarks for EACH are very close with BOINC 5.10.7. Here's a sample. 8/19/2007 9:48:43 PM||Suspending computation - running CPU benchmarks 8/19/2007 9:49:14 PM||Benchmark results: 8/19/2007 9:49:14 PM|| Number of CPUs: 1 8/19/2007 9:49:14 PM|| 2392 floating point MIPS (Whetstone) per CPU 8/19/2007 9:49:14 PM|| 4468 integer MIPS (Dhrystone) per CPU 8/19/2007 9:49:16 PM||Resuming computation So that's something like 13200 Drystone for 3 AMD cores???
SUPPORT ADVISOR
----------------------------------------Work+GPU i7 8700 12threads School i7 4770 8threads Default+GPU Ryzen 7 3700X 16threads Ryzen 7 3800X 16 threads Ryzen 9 3900X 24threads Home i7 3540M 4threads50% [Edit 1 times, last edit by retsof at Aug 20, 2007 3:07:56 AM] |
||
|
|
Movieman
Veteran Cruncher Joined: Sep 9, 2006 Post Count: 1042 Status: Offline |
Now Olympic: Perhaps yours isn't claiming high enough! Go check the thread on Xs showing the benches on highly clocked quads and you'll see that mine is lower than those. The dual socket machines take a hit due to memory and buss speed. This is probably one of the 10 fastest clovers in the world but on a core to core basis those quads kill me easily. I'm limited to DDR2-667FBDimms at 5-5-5-15-19 and the quads can use up to DDR2-1200 with cas4 latency's...MUCH faster and FSB close to 2000 vs my 1400.. All my best pal. So my AMD steam engines aren't doing too badly compared to those Intel strap-on diesels, eh? I have two 2.4GHz with DDR1-400-3200 and one newer 2.0GHz DDR1-<400?-1800 OC nicely to 2.4GHz running cooler than the others. Benchmarks for EACH are very close with BOINC 5.10.7. Here's a sample. 8/19/2007 9:48:43 PM||Suspending computation - running CPU benchmarks 8/19/2007 9:49:14 PM||Benchmark results: 8/19/2007 9:49:14 PM|| Number of CPUs: 1 8/19/2007 9:49:14 PM|| 2392 floating point MIPS (Whetstone) per CPU 8/19/2007 9:49:14 PM|| 4468 integer MIPS (Dhrystone) per CPU 8/19/2007 9:49:16 PM||Resuming computation So that's something like 13200 Drystone for 3 AMD cores??? Hi: The AMD's are good and I like them BUT you can't add benchmarks of multiple machines and compare to a single machine that way. Sort of apples and oranges in that one machine is sharing so many parts among the cores while the 3 machines each have all of a boards resources to them self. Then you mention"So that's something like 13200 Drystone for 3 AMD cores??? Yes, it is, but your adding and comparing the total to my single core. Add the 8 of mine and my drystone is 74,848 App 5.5 times the total of your 3 machines combined. Scary huh? Now you get an idea of why I sat here in wonder when I saw what this thing was capable of.Then toss in a cost factor and especially today the electrical cost of running the machines. Let me give an example. We've worked out that the most cost effective machine today is a Intel Q6600 G0 clocked to app 3500mhz. At that speed it pulls a constant 180w from the wall with a 83% effective PSU. It will make app 15,000 points a day at that speed. cost to build is app $300 for cpu,$200 for board,$100 for 2-1 gig of DDR2-6400 ram,$50 for high end heatsink,$3.00 for a cheap used pci vid. $653.00 total The other parts are static(almost) to all builds Now lets look at the clover. I'll use my "faster" one as an example as I know it's numbers well. Pulls a constant 450w from the wall and makes 26,000 points a day(claims over 30,000, can you match me up with some fast machines please! LOLOLOLOL) Cost to build is app $1500.00 for the cpu's,$450.00 for the board,$360.00 for 4x1 gig DDR2-667FBDimms,$100.00 for heatsinks and the same $3.00 for a cheapo pci vid( I use a $50.00 ATI X550, wasn't tossing a PCI card in that machine Total cost:$2413.00 So even though the clover is the king of the single machines it's far from the best combo for WCG work. You can essentially build 3 Quads and make 45,000 points a day for the same money that the clover makes 26,000 and at cheaper electrical cost. Downside of the quads is space need for the 3 systems vs the one but thats the only downside and a very small one at that. Now when your ready to upgrade those AMD's to some heavyweight crunchers come over and visit at XS and we'll set you up with a shopping cart thats guaranteed to work and work well together. We may be xtremely competitive but we don't keep secrets as the project finding the answers to these diseases is our and everyones #1 priority.The more and better machines we can get on this the sooner the quest is done and I can go from seeing a $300.00/month elec bill back to seeing a $100.00 one. As always, a pleasure to speak with you. [Edit 2 times, last edit by Movieman at Aug 20, 2007 7:20:24 AM] |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Just a small interjection from a non-tech, but if the RAM cant keep up with the CPU, you are actually running longer than the CPU could do a job in. The benchmark, flawed as it is, only tests the CPU!
----------------------------------------
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
|