Message boards :
Number crunching :
Long WU's
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 . . . 6 · Next
Author | Message |
---|---|
Send message Joined: 17 Sep 04 Posts: 26 Credit: 319,239 RAC: 0 |
Would this not have been what the old Aplha project was for ? |
Send message Joined: 12 Jul 11 Posts: 857 Credit: 1,619,050 RAC: 0 |
Good point; too late now I think but I shall try and remember. Also change user side to make deadline proportional to number of turns or max CPU estimate. Eric. |
Send message Joined: 12 Jul 11 Posts: 857 Credit: 1,619,050 RAC: 0 |
Sorry don't understand the credit system. I'll review with Igor when possible. |
Send message Joined: 12 Jul 11 Posts: 857 Credit: 1,619,050 RAC: 0 |
We cannot predict the unpredictable. We are studying chaos. The stimate is only an estimate and of the Maximum time required. BUT I agree the deadline should have been extended. Eric. |
Send message Joined: 1 Sep 04 Posts: 57 Credit: 2,835,005 RAC: 0 |
"floating-point error accumulation" Be very carefull and watch out for those, they can give wild errored results. Had to deal with them in my laborary all the time. ( Can vary from type cpu's, OS) |
Send message Joined: 26 Oct 04 Posts: 3 Credit: 814,721 RAC: 0 |
Seems i win 1st prize :) 37hrs and no bang to wall :) 10m turns wooo yay ! science in my computer :D http://lhcathomeclassic.cern.ch/sixtrack/workunit.php?wuid=2479609 |
Send message Joined: 12 Dec 05 Posts: 31 Credit: 9,709,398 RAC: 0 |
i'm going on 33 hours myself w/ a task. it's ~75% complete though, so it should finish with a total run time well within the 48-hour upper bound that Eric mentioned several posts up. this is on an X6 1090T simulating 7 cores - 5 for LHC@H, and 2 for T4T@H dual-threaded tasks. |
Send message Joined: 24 Apr 11 Posts: 37 Credit: 1,295,012 RAC: 0 |
Wow, one of my 3.75GHz AMD boxes finished one that took 24 hours... and the wingman is a 1.6GHz Pentium-M that will probably take 3 times longer... Does this mean the equations are turning out more and more relevant data??? Many longer tasks lately... but they seem to finish just fine... I have one task at 29 hrs now with 14 more to go on the same box... I have another on the 950 box at 39hrs elapsed with 11 more to go as well... LOL! :) http://lhcathomeclassic.cern.ch/sixtrack/workunit.php?wuid=2493244 |
Send message Joined: 9 Oct 10 Posts: 77 Credit: 3,671,357 RAC: 0 |
I'm afraid my old P4-m 1.4 GHz is not gonna make the deadline on these WUs ... :( |
Send message Joined: 29 Sep 04 Posts: 281 Credit: 11,866,264 RAC: 0 |
35 hrs in with 30 hrs estimated left on a 2.5Ghz i3. Others completed at 20hrs and 26 hrs with a couple of 10 minute ones thrown in just to confuse the estimator. Good to see longer WUs here as the longer the run, the more stable the beam. Not sure which is the preferred project between here and T4T as they're not really "rival" projects, just opposite sides of the detector, although I suppose a more stable beam should give better results so I've stopped getting T4T work to concentrate here. |
Send message Joined: 1 Sep 04 Posts: 57 Credit: 2,835,005 RAC: 0 |
Some of these are extra long 72% @ 41 hours ( i7, 3.33 ghz) T4T has 10 times less users, around 500 active in the last 24hrs. I'm sticking with them. |
Send message Joined: 12 Jul 11 Posts: 857 Credit: 1,619,050 RAC: 0 |
Well I should have said "floating-point rounding error" accumulation. Every WU is run twice (at least) and is validated if we have two identical results. t appears that we have only some 50 CPUs out of tense of thousand givibg erroneous results. Eric. |
Send message Joined: 12 Jul 11 Posts: 857 Credit: 1,619,050 RAC: 0 |
Well I should have said "floating-point rounding error" accumulation. Every WU is run twice (at least) and is validated if we have two identical results. t appears that we have only some 50 CPUs out of tense of thousand givibg erroneous results. Eric. |
Send message Joined: 27 Sep 08 Posts: 847 Credit: 691,851,631 RAC: 115,797 |
This just finished http://lhcathomeclassic.cern.ch/sixtrack/workunit.php?wuid=2467282 almost 200,000 sec's of processing time |
Send message Joined: 28 Sep 04 Posts: 9 Credit: 11,354,557 RAC: 0 |
Long long WU's..... my monster one http://lhcathomeclassic.cern.ch/sixtrack/workunit.php?wuid=2467332 155,297.90 crunching sec... 43h and something... o.O |
Send message Joined: 15 Jul 05 Posts: 8 Credit: 1,036,470 RAC: 0 |
i have long WU's too, some of them are finished but the credits are so poor. long WU ~8 credits/hour normal WU ~30 credits/hour there is a correction necessary |
Send message Joined: 12 Jul 11 Posts: 857 Credit: 1,619,050 RAC: 0 |
Right I went a bit over the top here BUT...... As this is an experiment we are exploring new ground. It certainly seems there is a problem with credits. I don't think BOINC should send WUs if the Estimate divided by the speed of the CPU exceeds the deadline or is more than 50%/40%/10%??? of the deadline. Maybe that is OK. All to be looked at soonest. Eric. |
Send message Joined: 12 Jul 11 Posts: 857 Credit: 1,619,050 RAC: 0 |
Right I went a bit over the top here BUT...... As this is an experiment we are exploring new ground. It certainly seems there is a problem with credits. I don't think BOINC should send WUs if the Estimate divided by the speed of the CPU exceeds the deadline or is more than 50%/40%/10%??? of the deadline. Maybe that is OK. All to be looked at soonest. Eric. |
Send message Joined: 12 Jul 11 Posts: 857 Credit: 1,619,050 RAC: 0 |
Right I went a bit over the top here BUT...... As this is an experiment we are exploring new ground. It certainly seems there is a problem with credits. I don't think BOINC should send WUs if the Estimate divided by the speed of the CPU exceeds the deadline or is more than 50%/40%/10%??? of the deadline. Maybe that is OK. All to be looked at soonest. Eric. |
Send message Joined: 15 Jul 05 Posts: 8 Credit: 1,036,470 RAC: 0 |
thx Eric, maybe this wu's are helpful to locate the problem http://lhcathomeclassic.cern.ch/sixtrack/workunit.php?wuid=2481201 http://lhcathomeclassic.cern.ch/sixtrack/workunit.php?wuid=2481003 http://lhcathomeclassic.cern.ch/sixtrack/workunit.php?wuid=2509899 they get all the same credit (281.02) but have realy different runtimes |
©2024 CERN