Message boards :
Number crunching :
Some changes made to this recent BOINC update BUT
Message board moderation
Author | Message |
---|---|
W-K 666 Send message Joined: 18 May 99 Posts: 19058 Credit: 40,757,560 RAC: 67 |
My details for Application details has changed from "Max tasks per day" of about 120 to: SETI@home Enhanced (anonymous platform, nvidia GPU) Number of tasks completed 1471 Max tasks per day 1567 Number of tasks today 570 Consecutive valid tasks 1467 Average turnaround time 7.50 days And the computer has downloaded about 370 tasks in the last hour or so. The bad news is, I have never has the flops adjustment in my app_info file, but I have now got tasks with expected completion time of 01:23. The fastest VHAR on this slowish computer is ~8:00. |
Helli_retiered Send message Joined: 15 Dec 99 Posts: 707 Credit: 108,785,585 RAC: 0 |
Congratulations. :-) I wish i could see this on my Rigs. Helli |
W-K 666 Send message Joined: 18 May 99 Posts: 19058 Credit: 40,757,560 RAC: 67 |
Additional news is that these tasks with 01:23 completion time are not all VHAR tasks some at least are mid range AR = 0.4xxxx |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
Additional news is that these tasks with 01:23 completion time are not all VHAR tasks some at least are mid range AR = 0.4xxxx Divide the estimate by the Duration Correction Factor, then multiply by 10 to get the allowed amount of crunch time before the task will be aborted by the core client. With my modest hosts, if I saw grossly underestimated work I'd be easily able to modify the <rsc_fpops_bound> for each such task to allow enough time to do the work. For those with much more throughput, there are likely to be too many tasks to handle by manual editing. Replacing all the rsc_fpops_bound values in client_state.xml with 2.5e15 would work[1], though. That's about the unadjusted bound for the longest-running MB work. As that bound is intended simply as protection from corrupted applications or such, making the change does not harm the project in any way. OTOH, if an app did happen to get into a loop, it would be a long time before BOINC would kill the task. [1]When reading xml files, BOINC uses a standard conversion to double for numeric entries. So editing to <rsc_fpops_bound>2.5e15</rsc_fpops_bound> will be read correctly. When BOINC rewrites the file it will become <rsc_fpops_bound>2500000000000000.000000</rsc_fpops_bound> simply because that's the way BOINC formats xml output of such values. Joe |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
My details for Application details has changed from "Max tasks per day" of about 120 to: The reason you are seeing such low estimates is the "DCF squared" transitional problem I wrote about in April, and recently re-posted on these boards. Basically, the server is now applying your DCF value (of about 0.17, I guess) to the tasks before it sends them to you, in anticipation that your local DCF is close to 1.0 But until you work far enough through your cache for the first "short-estimated" task to complete, your local DCF won't adjust. Perhaps the simplest change you could make would be to give it a helping hand, by setting local DCF to 1.0 - that would stop additional downloads in their tracks (BOINC would finally realise that you've got enough work), and let EDF handle any VHARs that needed the hurry-up call. |
W-K 666 Send message Joined: 18 May 99 Posts: 19058 Credit: 40,757,560 RAC: 67 |
Before I went out I did increase the DCF to 0.2 from 0.05 and set to NTT, which was enough to send it into high priority mode for the gpu. It has done 41 tasks since then, and set the DCF to 0.89xxxx. Which has now increased to completion times to about 4 times what they should be for the new MB tasks and about 25 times what it should be for the old MB tasks and the AP tasks. So everything is in panic mode. This does nothing to help the main problem I will run out of AP tasks for the cpu by 12:00 UTC tomorrow. Another thing that is utterly stupid is that on the gpu it is going to have to crunch all the tasks downloaded since this f*** up first. Then it is going to process all the older gpu tasks that is going to reset the DCF back to it's old value. And it is going to . . . . . . . before disappearing up its . . . . |
Rasputin Send message Joined: 13 Jun 02 Posts: 1764 Credit: 6,132,221 RAC: 0 |
This has to be wrong... SETI@home Enhanced (anonymous platform, Unknown) Number of tasks completed 201 Max tasks per day 1110309626 Number of tasks today 0 Consecutive valid tasks 137 Average turnaround time 5.28 days My maximum tasks per day is over 1 billion tasks. LOL! Edit: The rest of you guys and gals can forget about getting any work. I'm about to download one billion one hundred and ten million tasks. http://i289.photobucket.com/albums/ll237/982i/onebilliontasks.jpg |
SciManStev Send message Joined: 20 Jun 99 Posts: 6652 Credit: 121,090,076 RAC: 0 |
This has to be wrong... LOL! May I have some of those please? Oh my goodness! In spite of the strangeness going on, there are some very funny moments! :) Steve Warning, addicted to SETI crunching! Crunching as a member of GPU Users Group. GPUUG Website |
Rasputin Send message Joined: 13 Jun 02 Posts: 1764 Credit: 6,132,221 RAC: 0 |
This has to be wrong... That's fine. By the next day I'll have a quota in the trillions according to this math. |
Rasputin Send message Joined: 13 Jun 02 Posts: 1764 Credit: 6,132,221 RAC: 0 |
This has to be wrong... Yeah, I'll be nice and let you have a couple million. :) Strange it is.. |
zoom3+1=4 Send message Joined: 30 Nov 03 Posts: 65745 Credit: 55,293,173 RAC: 49 |
This has to be wrong... I've reserved zillions upon zillions, So that's that. ;) The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's |
W-K 666 Send message Joined: 18 May 99 Posts: 19058 Credit: 40,757,560 RAC: 67 |
This has to be wrong... And I am going to get the credit for the (21008 - 105 =) 20903 tasks that my Q6600/gt240 computer completed today. SETI@home Enhanced (anonymous platform, nvidia GPU) Number of tasks completed 21008 Max tasks per day 21104 Number of tasks today 676 Consecutive valid tasks 21004 Average turnaround time 7.50 days Who needs an i7 plus multiple Fermi cards to go top of the rankings. |
SciManStev Send message Joined: 20 Jun 99 Posts: 6652 Credit: 121,090,076 RAC: 0 |
Well your kicking the pants off my i7 rig with multiple Fermi cards! :) Steve Warning, addicted to SETI crunching! Crunching as a member of GPU Users Group. GPUUG Website |
soft^spirit Send message Joined: 18 May 99 Posts: 6497 Credit: 34,134,168 RAC: 0 |
This has to be wrong... Well.. now we know where all the credits have been landing... |
Rasputin Send message Joined: 13 Jun 02 Posts: 1764 Credit: 6,132,221 RAC: 0 |
Haha! You'll have to beat me downloading them. |
Rasputin Send message Joined: 13 Jun 02 Posts: 1764 Credit: 6,132,221 RAC: 0 |
Sounds like it. Or maybe someone wanted to take the edge of of our frustration with some humor. |
Rasputin Send message Joined: 13 Jun 02 Posts: 1764 Credit: 6,132,221 RAC: 0 |
That's a boatload of tasks completed! lol Man, things are definitely out of whack at the moment. |
Rasputin Send message Joined: 13 Jun 02 Posts: 1764 Credit: 6,132,221 RAC: 0 |
Don't know about credits but I'm going to be up to my eyeballs in tasks soon. Edit: I'm a spammer tonight. |
zoom3+1=4 Send message Joined: 30 Nov 03 Posts: 65745 Credit: 55,293,173 RAC: 49 |
And their all Ghosts now... [Ghostly haunting]Booooooo![/Ghostly haunting] ;) The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's |
Rasputin Send message Joined: 13 Jun 02 Posts: 1764 Credit: 6,132,221 RAC: 0 |
LOL! ::Runs away screaming:: |
©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.