Message boards :
Number crunching :
Enhanced version 10x longer!
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 · 6 · 7 . . . 9 · Next
Author | Message |
---|---|
Scary Capitalist Send message Joined: 21 May 01 Posts: 7404 Credit: 97,085 RAC: 0 |
Great, thanks. So, basically if seti knows someone's machine is dug out of an archaelogical pit it will attempt to send units to it accordingly? And no manual handling of seti will be necessary so that all of the "d/l and forget" computers will keep crunching---excellent. (except possibly for optimized clients---I don't think anyone in my group use any and the people that do are savy enough to make the necessary adjustments if needed--also excellent) If I'm wrong here in my interpretation then let me know. Thanks again for all of the responses. It's a relief to know that in my small group of 38 (14 active boicers) I won't see their numbers decrease to 4 or 5. Founder of BOINC team Objectivists. Oh the humanity! Rational people crunching data! I did NOT authorize this belly writing! |
Paul D. Buck Send message Joined: 19 Jul 00 Posts: 3898 Credit: 1,158,042 RAC: 0 |
Interesting discussion as usual. Just to reinforce a couple points ... :) Though processing times are increasing we are continuing on the trend to increase capability in the application. Since I started I think this is the 4th overhaul. Each increased processing times by about 2-4x and was always accompanied by howls of anguish about the end of the world and life as we know it. Not suprising me at all, within a short period of time, sanity returned and we have gone from 100's of hours (my starting average was ~32 hours) to process one work unit to ~30 minutes with optimizaition. Which only means that with optimization we are going to increase back to, on average, 10's of hours. Also, if I have been understanding the work going on over on the Beta project, the new application will be much more optimized, and correct, from all the testing and tuning going on. Also, this brings SETI@Home more in line with Einstein@Home and Rosetta@Home for processing times ... :) |
TPR_Mojo Send message Joined: 18 Apr 00 Posts: 323 Credit: 7,001,052 RAC: 0 |
On the 500MHz point, I have two PIII Katmai 500MHz attached to the test project, and with the app in its current (which I don't believe will be final) form runtimes are of the order of 4-5 days. So for a machine of this spec attached only to SETI I can't see deadlines will be a big issue. For a machine of this spec running multiple projects - BOINC will still (try to) hit deadline but overloading the machine with several "long" projects, say a mix of CPDN, SETI, Rosetta and LHC, might cause BOINC to be chasing said deadlines more. |
W-K 666 Send message Joined: 18 May 99 Posts: 19122 Credit: 40,757,560 RAC: 67 |
Looks like your worries about crunch times may be over, see Seti Beta or Enhanced 4.11 |
J D K Send message Joined: 26 May 04 Posts: 1295 Credit: 311,371 RAC: 0 |
"For a machine of this spec running multiple projects - BOINC will still (try to) hit deadline but overloading the machine with several "long" projects, say a mix of CPDN, SETI, Rosetta and LHC, might cause BOINC to be chasing said deadlines more." Lowfield You can have a faster machine get to chasing deadlines also, running 840ee 3.2 ht, running, Seti, Einstein, Rosetta, Predictor, LHC and CPDN. Started with CPDN, Regular and a Sulphur, about 20 Rosetta, due 12/31, about 30 Seti, Due 12/17, added LHC, about 40, due 12/06, Einstein about 20, due 12/6, and Predictor would not down load. Connect to set for 2...... About 8 days before 12/6 Boinc goes into deadline mode and processes LHC all four CPUs for almost 2 days, it gets that cleared up and then goes to Einstein for almost 2 days, I have 3 1/2 to go, OK now predictor downloads about 60, due on 12/06, it has been round robining like there is no problem..... I just sit and watch it, tis interesting to see it work....... And the beat goes on Sonny and Cher BOINC Wiki |
The Colourful jester Send message Joined: 18 Oct 01 Posts: 35 Credit: 2,680,511 RAC: 0 |
What was the solution to stopping enhanced from reverting back to 0.00% again? Have to ask here. Not enough credit for the actual enhanced board yet. Hullo there. |
W-K 666 Send message Joined: 18 May 99 Posts: 19122 Credit: 40,757,560 RAC: 67 |
What was the solution to stopping enhanced from reverting back to 0.00% again? Have to ask here. Not enough credit for the actual enhanced board yet. As far as I can see the recommendation is to keep in memory when paused. What OS are you using with SetiB, it seems to be hitting Win98 most. To highlight another possible problem. I spotted someone claiming a weird no of credits, TMR says it is a graphics driver problem. The odd claim, the host claimed -10,675,189.39 |
Tetsuji Maverick Rai Send message Joined: 25 Apr 99 Posts: 518 Credit: 90,863 RAC: 0 |
As for this one, I guess the problem is not in the s@h_enhanced application. That computer sometimes gets "Incorrect function" error at beta project like this. I wrote it should be a graphics driver's problem because I checked Wiki and found it says "Incorrect function" error is because boinc manager fails to find old graphics driver. see here (though I saw another page on this matter.... I lost the link!) However I received email from the owner of the computer that he was using the latest driver with ATI Radeon 9250, and am at a loss. I also found he got the same error with this public s@h (see this one). So this problem isn't specific to setiathome_enhanced. He may be getting this error with other projects. So I wrote to him that he'd better ask this on this message board because I don't know much about Windows... He will appear in this message board or somewhere around here. Please help him when he comes.... With "boinc incorrect function" on google, I got several results about this error with other projects also. |
TPR_Mojo Send message Joined: 18 Apr 00 Posts: 323 Credit: 7,001,052 RAC: 0 |
Latest runtimes with SETI 4.11 (standard 5.2.x client and vanilla S@H 4.11): PIII 500 Linux 57 hours AMD XP2400+ Linux 18 hours Intel P4 Northwood 2.8 Windows 21 hours AMD XP2800+ Windows 23 hours AMD 64 3200+ Windows (32-bit) 23.5 hours AMD 64 3200+ Linux (64-bit) 9.9 hours Now obviously angle ranges and the general flux of units and applications within Beta would mean it would be unwise to draw too detailed a conclusion from these runtimes. The difference in processing times between the Windows and Linux installations on AMD 3200+, for example, looks like a glitch to me not a pattern - although both results are valid. It does show, I believe, that runtimes will not be stupidly long and that older machines will still be able to play an active part. Also I just love the way credit claims are so level, this run of 5 results demonstrates: 11/12 AMD 64 3200 Linux 53,244.36 seconds 191.56 claimed 12/12 AMD MP2400+ Linux 66,843.11 seconds 191.55 claimed 12/12 AMD MP2400+ Linux 66,729.97 seconds 191.54 claimed 13/12 AMD XP2400+ Linux 67,686.82 seconds 191.56 claimed 13/12 AMD 64 3200 Linux 51,855.00 seconds 191.54 claimed Now thats what I call good :) |
The Colourful jester Send message Joined: 18 Oct 01 Posts: 35 Credit: 2,680,511 RAC: 0 |
What was the solution to stopping enhanced from reverting back to 0.00% again? Have to ask here. Not enough credit for the actual enhanced board yet. My OS is WinXP, and I've got it set to keep the apps in memory when paused. So it's abit weird and annoying. I do have version 4.09 of it though. Hullo there. |
W-K 666 Send message Joined: 18 May 99 Posts: 19122 Credit: 40,757,560 RAC: 67 |
Lowfield, Thanks for the information, but without the angle_range they don't really mean that much. So it might be an idea to give the claimed credits numbers as that would indicate approximately how many Flops were used. My results so far on Pent M (dothan) 1.86Ghz are; CPU time(secs) Claimed Credit 66,995.36 (18.6hr)147.85 104,213.73 (29hr) 191.62 105,022.66 (29hr) 191.59 101,736.08 (28hr) 191.65 |
W-K 666 Send message Joined: 18 May 99 Posts: 19122 Credit: 40,757,560 RAC: 67 |
My OS is WinXP, and I've got it set to keep the apps in memory when paused. So it's abit weird and annoying. I do have version 4.09 of it though. If you are still running 4.09 you should abort that unit, and any other 4.09's, and it will download new unit and 4.11. |
Paul D. Buck Send message Joined: 19 Jul 00 Posts: 3898 Credit: 1,158,042 RAC: 0 |
Maybe we can get them to store and display the number of FLOPS for the work unit/result. In the long run that may also be interesting information for the developers to monitor regarding the behavior of the application ... I know that it would be a metric *I* would save ... |
TPR_Mojo Send message Joined: 18 Apr 00 Posts: 323 Credit: 7,001,052 RAC: 0 |
Lowfield, Your wish is my command :) PIII 500 Linux 57 hours 147.81 claimed AMD XP2400+ Linux 18 hours 191.56 claimed Intel P4 Northwood 2.8 Windows 21 hours 147.80 claimed AMD XP2800+ Windows 23 hours 147.16 claimed AMD 64 3200+ Windows (32-bit) 23.5 hours 147.17 claimed AMD 64 3200+ Linux (64-bit) 9.9 hours 147.80 claimed AMD 64 3200+ Linux (64-bit) 14.7 hours 191.56 claimed The last two are from the same machine, they look too quick, but this is my only Fedora Core 4 64 bit machine and the operating system seems much much quicker in use, I believe due to the later version of GCC it was compiled with. |
TPR_Mojo Send message Joined: 18 Apr 00 Posts: 323 Credit: 7,001,052 RAC: 0 |
Maybe we can get them to store and display the number of FLOPS for the work unit/result. I agree, as a metric for both end user and project team this would be a nice piece of information to have available. It would make true comparisons between machines possible, for example, in a way benchmarks do not. If I have two identical servers running, say, Fedora Core 2 and Fedora Core 4 and the Core 2 machine takes significantly longer to process the same number of flops, then I know the operating system version is having a big impact. The same logic can be applied to memory, disk, .......... |
Ingleside Send message Joined: 4 Feb 03 Posts: 1546 Credit: 15,832,022 RAC: 13 |
Maybe we can get them to store and display the number of FLOPS for the work unit/result. If the claimed credits is based on the "flops-count" in Seti_Enhanced, just calculate: Cobblestone * 8.64e11 = fpops_cumulative |
Tom Send message Joined: 17 Apr 00 Posts: 24 Credit: 283,013 RAC: 0 |
I agree. My thought is this. A cure for AIDS is found. The company uses information from one of the projects and makes $100 million. Saving 40 million peoples lives. That is OK with me. The only thing I hope is that they pump some money into these projects so they can continue to enrich man kind. |
Tom Send message Joined: 17 Apr 00 Posts: 24 Credit: 283,013 RAC: 0 |
Then don't crunch. Either you understand the math and science behind the project or you don't. I have some PII running and they are also cruching on 2 other projects. I probably never get credit for any chrunging they do on seti. I am not worried about it. If you are here for credits you need to get a life, really. The credits are fun and all that, but try to figure out what is really important about these projects. It ain't you, and it ain't me, and it definitly isn't the credits. Get over it. |
Mr.Pernod Send message Joined: 8 Feb 04 Posts: 350 Credit: 1,015,988 RAC: 0 |
Maybe we can get them to store and display the number of FLOPS for the work unit/result. from the client_state.xml for a workunit: <rsc_fpops_est>414215375930942.000000</rsc_fpops_est> (Enhanced Beta) <rsc_fpops_est>27924800000000.000000</rsc_fpops_est> (Regular) these numbers seem independant of angle-range |
zoom3+1=4 Send message Joined: 30 Nov 03 Posts: 65815 Credit: 55,293,173 RAC: 49 |
Though I applaud the idea of the improved sensitivity, I am disturbed by the increased work unit times. When my "new" computer is completed I'm anticipating the completion of 16-18 work units a day between the two dual processor computers using optimized clients. I'll admit that they're not fast when compared to many new computers but I think 8 to 10 work units a day from a couple of P3 computers is pretty good. If It takes 9hrs or more for to do a WU(To say nothing of 30-51hr or longer WUs), Then I may as well do EinsteinBoinc instead as SetiBoinc will be taking too much time for Me to be comfortable with. As I won't feel like I'm helping any, But then I don't have a cray Super computer. Unless someone makes an Optimized app for the New WUs of course. The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's |
©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.