Join PrimeGrid
Returning Participants
Community
Leader Boards
Results
Other
drummers-lowrise
|
Message boards :
Proth Prime Search :
What happened here?
Author |
Message |
|
My GTX 460 machine has been tearing through the sieve WUs at an alarming rate; I've seen 9:50 to 10:20 for completion times, and tasks never error out.
I'm a bit confused, though, as to why one particular task took so very long. (the one in the middle below)
214779753 150755940 165444 4 Jan 2011 4:01:19 UTC 4 Jan 2011 4:32:45 UTC Completed and validated 617.70 76.56 2,311.00 Proth Prime Search (Sieve) v1.37 (cuda23)
214779630 150755672 165444 4 Jan 2011 4:01:19 UTC 4 Jan 2011 5:39:35 UTC Completed and validated 2,488.47 74.14 2,311.00 Proth Prime Search (Sieve) v1.37 (cuda23)
214779504 150755673 165444 4 Jan 2011 4:01:19 UTC 4 Jan 2011 5:30:38 UTC Completed and validated 612.72 74.48 2,311.00 Proth Prime Search (Sieve) v1.37 (cuda23)
The task in question is this one - completed and validated just fine, but why'd it take over 4 times as long as every other?
All the sieve tasks use 0.36 CPUs, which on my P4 translates to 0.18 of 1 virtual core, and the whole system actually reads as having a bit less memory than the GPU it's running - if anything, I want to be sure I'm not heading down the wrong path by using a brand-new, overclocked Fermi card on a mostly-six-year-old computer.
Note that I have had tasks "hang" in the past, where they'll run for ~20 minutes and hit maybe 5% before I realize there's something wrong and abort them.
It would also appear, from reading some stderrs, that none of the sieve tasks I do ever checkpoint to disk, meaning that I cannot suspend them without having to restart them (and that's not always successful either).
So, is this normal, or should I change something?
As much as I know about GPUs and crunching on 'em, I'm not exactly familiar with how things work "under the hood," so to speak, with regards to apps like PPS.
Any pointers?
____________
| |
|
|
My GTX 460 machine has been tearing through the sieve WUs at an alarming rate; I've seen 9:50 to 10:20 for completion times, and tasks never error out.
I'm a bit confused, though, as to why one particular task took so very long. (the one in the middle below)
214779753 150755940 165444 4 Jan 2011 4:01:19 UTC 4 Jan 2011 4:32:45 UTC Completed and validated 617.70 76.56 2,311.00 Proth Prime Search (Sieve) v1.37 (cuda23)
214779630 150755672 165444 4 Jan 2011 4:01:19 UTC 4 Jan 2011 5:39:35 UTC Completed and validated 2,488.47 74.14 2,311.00 Proth Prime Search (Sieve) v1.37 (cuda23)
214779504 150755673 165444 4 Jan 2011 4:01:19 UTC 4 Jan 2011 5:30:38 UTC Completed and validated 612.72 74.48 2,311.00 Proth Prime Search (Sieve) v1.37 (cuda23)
The task in question is this one - completed and validated just fine, but why'd it take over 4 times as long as every other?
All the sieve tasks use 0.36 CPUs, which on my P4 translates to 0.18 of 1 virtual core, and the whole system actually reads as having a bit less memory than the GPU it's running - if anything, I want to be sure I'm not heading down the wrong path by using a brand-new, overclocked Fermi card on a mostly-six-year-old computer.
Note that I have had tasks "hang" in the past, where they'll run for ~20 minutes and hit maybe 5% before I realize there's something wrong and abort them.
It would also appear, from reading some stderrs, that none of the sieve tasks I do ever checkpoint to disk, meaning that I cannot suspend them without having to restart them (and that's not always successful either).
So, is this normal, or should I change something?
As much as I know about GPUs and crunching on 'em, I'm not exactly familiar with how things work "under the hood," so to speak, with regards to apps like PPS.
Any pointers?
A possible cause for that is when boinc does "benchmarks" of the cpu. It should be possible to see that in the messages tab or in the log files.
____________
Join BOINC United now! | |
|
|
Hm... I though I'm alone. I though smth wrong with hardware or OCing, they are stable for weeks tough.
Same stuff (WU duration went up on somewhat 10%) on both my rigs. Video - GTX 570.
| |
|
|
Same here...
During the day, the duration is about 10-15% longer per WU.
Has somebody an explanation for that?
____________
| |
|
|
http://www.primegrid.com/forum_thread.php?id=3044
____________
35 x 2^3587843+1 is prime! | |
|
|
Yes!
my 470/460/9800 eduration all have about 10% longer than before,except my gt240 have the same time. | |
|
|
I read that topic. Although it's in English I barely understood smth. But to make long story it's OK and nothing wrong from my side :-) That's all I need to know | |
|
Message boards :
Proth Prime Search :
What happened here? |