Advanced search

Message boards : Wish list : Deadline

Author Message
Profile rebirther
Avatar
Send message
Joined: 7 Jul 07
Posts: 53
Credit: 3,048,781
RAC: 0
Level
Ala
Scientific publications
watwatwatwatwat
Message 6341 - Posted: 3 Feb 2009 | 9:25:09 UTC

- a one week deadline
- larger cache as set in preferences (not only 4 WUs)

Profile Paul D. Buck
Send message
Joined: 9 Jun 08
Posts: 1050
Credit: 37,321,185
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 6345 - Posted: 3 Feb 2009 | 11:42:55 UTC - in response to Message 6341.

- a one week deadline
- larger cache as set in preferences (not only 4 WUs)


With my cards I am not sure I need a longer deadline ...

But a larger cache is certainly something I would wish for the simple reason I want to keep the hopper full and the turning burning ...

I know this project has not yet had an outage in the last month or so I have been on it ... out of work because of the feeder bug, but not really down ... but that is the point, with such a low queue allowed many ran dry.

When there are other projects to pick up the slack this is less of an issue, but, right now there are only two places to go and that is not enough for the addicts (like me).

And when you have a set of 4 GPUs with the ability to do a task in 4-5 hours, even a short outage is "expensive" to us who once started, never want the iron to cool ...

Profile Bender10
Avatar
Send message
Joined: 3 Dec 07
Posts: 167
Credit: 8,368,897
RAC: 0
Level
Ser
Scientific publications
watwatwatwatwatwatwat
Message 6350 - Posted: 3 Feb 2009 | 12:37:46 UTC - in response to Message 6345.

- a one week deadline
- larger cache as set in preferences (not only 4 WUs)


Deadline...Hmmm, that's been beat to death...

Larger Cache...Well, this project is still in Beta I believe. So this would be driven by what results the developers want to see to solve problems, not by how much the 'testers' want to crunch (but you knew that). But if those 2 purposes happen to coincide...good for everyone, if not...

____________


Consciousness: That annoying time between naps......

Experience is a wonderful thing: it enables you to recognize a mistake every time you repeat it.

Profile Paul D. Buck
Send message
Joined: 9 Jun 08
Posts: 1050
Credit: 37,321,185
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 6353 - Posted: 3 Feb 2009 | 16:57:10 UTC - in response to Message 6350.

- a one week deadline
- larger cache as set in preferences (not only 4 WUs)


Deadline...Hmmm, that's been beat to death...

Larger Cache...Well, this project is still in Beta I believe. So this would be driven by what results the developers want to see to solve problems, not by how much the 'testers' want to crunch (but you knew that). But if those 2 purposes happen to coincide...good for everyone, if not...


After saying that I had not seen an outage on GPU GRid, we just had one ...

Thankfully it seems to have only been for a few moments ...

But, when you can plow through the quota of work in just a few hours it is bothersome to many of us ...

But I agree that it is up to the project, we are just agitating to have the policy re-examined. :)

Kaylie
Send message
Joined: 14 Feb 09
Posts: 2
Credit: 2,479
RAC: 0
Level

Scientific publications
wat
Message 6984 - Posted: 25 Feb 2009 | 16:22:41 UTC - in response to Message 6350.

Maybe the deadline issue has been a bit hammered, but it is extremely frustrating for those of us who don't have blazing fast GPU's. It is maddening to get close to finishing in time, only to have things slow up at the end.

I would leave this project to the elite group with the over-capable machines, but SETI@HOME is the only other project I can find that runs CUDA tasks and they often run out of work.

I would like to put my machinery to good use, but I can't do it here because it seems to have just a tiny bit too little speed to beat the seemingly unrealistic deadline. (Another couple miles an hour...)

Does anyone know of any other CUDA enabled BOINC projects to participate in?

Thanks. (I better quit typing, the deadline may have passed.)

Kaylie
Send message
Joined: 14 Feb 09
Posts: 2
Credit: 2,479
RAC: 0
Level

Scientific publications
wat
Message 7032 - Posted: 27 Feb 2009 | 18:11:10 UTC - in response to Message 6984.

Okay, I should join “Underclockers Anonymous”. But I would still like to know if anyone out there knows of projects to donate the unused (inadequate) clock cycles of my GPU’s to.

Profile Paul D. Buck
Send message
Joined: 9 Jun 08
Posts: 1050
Credit: 37,321,185
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 7043 - Posted: 28 Feb 2009 | 7:12:30 UTC - in response to Message 7032.

Okay, I should join “Underclockers Anonymous”. But I would still like to know if anyone out there knows of projects to donate the unused (inadequate) clock cycles of my GPU’s to.


Because it is early days there are only four projects that are using GPUs at this time. All of these projects have "issues" so to speak.

GPU Grid, Nvidia, needs higher end cards because of the work
SaH, Nvidia, does not always play well with others and some tasks take forever to run
SaH Beta, Nvidia, often no work and lockups
Milky Way, ATI, also requires higher end cards that have double precision capability

The only other project that is even rumored to be working on GPU version is Einstein@Home and that is a rumor only ... no word from project people ...

Scott Brown
Send message
Joined: 21 Oct 08
Posts: 144
Credit: 2,973,555
RAC: 0
Level
Ala
Scientific publications
watwatwatwatwatwat
Message 7052 - Posted: 28 Feb 2009 | 15:33:40 UTC - in response to Message 7043.


The only other project that is even rumored to be working on GPU version is Einstein@Home and that is a rumor only ... no word from project people ...


PrimeGrid is also investigating the use of GPU's for their sieving tasks, though any actual application is likely months away...


Profile Paul D. Buck
Send message
Joined: 9 Jun 08
Posts: 1050
Credit: 37,321,185
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 7055 - Posted: 28 Feb 2009 | 16:20:47 UTC - in response to Message 7052.


The only other project that is even rumored to be working on GPU version is Einstein@Home and that is a rumor only ... no word from project people ...


PrimeGrid is also investigating the use of GPU's for their sieving tasks, though any actual application is likely months away...


I had not heard that before... Thanks ...

I don't do as much Prime Grid as the longer running versions seem way too unstable to me. Perhaps with GPU processing I might put more time on them ... In the mean time ... we wait ... :)

Profile Bender10
Avatar
Send message
Joined: 3 Dec 07
Posts: 167
Credit: 8,368,897
RAC: 0
Level
Ser
Scientific publications
watwatwatwatwatwatwat
Message 7059 - Posted: 28 Feb 2009 | 21:04:10 UTC - in response to Message 7032.
Last modified: 28 Feb 2009 | 21:04:52 UTC

Okay, I should join “Underclockers Anonymous”. But I would still like to know if anyone out there knows of projects to donate the unused (inadequate) clock cycles of my GPU’s to.

And the OTHER GPU projects available are Folding@Home and Distributed.net. Both have non-Boinc clients, and will run fine on your GPU while your run non-GPU projects on Boinc.

And these projects work with (most) any CUDA GPU's.

On Folding, you can do a Wu with an 8800GT in ~4-8 hours (depending on which wu you get).

The distributed.net CUDA clients are still in Beta. All GPU's work on the same Wu's. How fast you finish depends on the speed of your GPU. But Wu completion times are measured at (5-9) seconds for GTX260/280, and (20-25) seconds for an 8800GT (last I looked).
____________


Consciousness: That annoying time between naps......

Experience is a wonderful thing: it enables you to recognize a mistake every time you repeat it.

PunkSkeleton
Send message
Joined: 1 Mar 09
Posts: 1
Credit: 0
RAC: 0
Level

Scientific publications
wat
Message 7089 - Posted: 1 Mar 2009 | 22:38:47 UTC - in response to Message 7059.

I'm sorry but with current deadlines (5 days) and the impact on the computer's performance gpugrid is suitable only for fanatics with high-end video cards. On my 9600 I completed < 4% in an hour so it would take well over 25 hours to complete a WU. So you basically want me to do 2 things (if I want to participate):
1. Suffer unpleasantly slow interactive mode. I could survive with that but not 100% of the time my CPU is on.
2. Leave my CPU on all the time.

I won't do any of those 2 things unless to compute something like Clean Energy Project (and I'm still not leaving my CPU on for it).

ussr
Send message
Joined: 16 Jul 09
Posts: 1
Credit: 850,236
RAC: 0
Level
Gly
Scientific publications
watwatwatwat
Message 11503 - Posted: 30 Jul 2009 | 15:16:26 UTC

The deadline is too small!
My nVidia GTX260 just cannot complete a Wu in time if my computer is in use...
With BOINC 6.6.36 I always see status as a 'Waiting (0.21 CPU, 1 CUDA)'

I have to abort most of my tasks at 75-90%:
http://www.gpugrid.net/results.php?userid=30685

The deadline HAS TO be increased.

Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Send message
Joined: 14 Mar 07
Posts: 1914
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 11524 - Posted: 30 Jul 2009 | 22:14:02 UTC - in response to Message 11503.

That's because it's in use, so it's not actually doing anything. You can change that in the preferences. Your card should take just 6 hours to finish.

gdf

Betting Slip
Send message
Joined: 5 Jan 09
Posts: 666
Credit: 2,498,095,550
RAC: 1,931
Level
Phe
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 11536 - Posted: 31 Jul 2009 | 9:46:16 UTC - in response to Message 11503.

I for one like the deadlines, 25% bonus for returning in 48hrs and the small caches.
If I understand correctly each WU is part of a chain and the next one can't be sent until the previous has been completed.
____________
Radio Caroline, the world's most famous offshore pirate radio station.
Great music since April 1964. Support Radio Caroline Team -
Radio Caroline

Profile robertmiles
Send message
Joined: 16 Apr 09
Posts: 409
Credit: 280,744,946
RAC: 507,501
Level
Asn
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 12042 - Posted: 22 Aug 2009 | 20:10:58 UTC - in response to Message 7089.

I'm sorry but with current deadlines (5 days) and the impact on the computer's performance gpugrid is suitable only for fanatics with high-end video cards. On my 9600 I completed < 4% in an hour so it would take well over 25 hours to complete a WU. So you basically want me to do 2 things (if I want to participate):
1. Suffer unpleasantly slow interactive mode. I could survive with that but not 100% of the time my CPU is on.
2. Leave my CPU on all the time.

I won't do any of those 2 things unless to compute something like Clean Energy Project (and I'm still not leaving my CPU on for it).


There's a project related to clean energy among those offered by World Community Grid, but I've seen no sign it can use a GPU. Do you have enough memory that you can leave BOINC projects in memory while the computer is in use for something else, as I currently do on both of my desktops?

HerakulmanRuhtinas
Send message
Joined: 2 May 09
Posts: 1
Credit: 934,301
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwatwatwatwat
Message 13174 - Posted: 14 Oct 2009 | 18:24:22 UTC - in response to Message 6341.

- a one week deadline


I second to that, my comp is not on 24/7 and with my GTX260 maybe every third WU goes overtime. I'd like to see the deadline increased to 7 to 14 days.

Profile Damaraland
Send message
Joined: 7 Nov 09
Posts: 152
Credit: 16,181,924
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwatwat
Message 13425 - Posted: 9 Nov 2009 | 14:59:30 UTC

I agree deadlines are too short!!. There are many computers that are not used everyday, they process units and when they are finish... too late (waste of electricity power and time).

I think the servers could make distintions between "fast computers" and "slow computers" this way people that process results with slow hardware or that are not 24/7 could help to the proyect even if their results come out one week or 2 weeks later.

Profile Damaraland
Send message
Joined: 7 Nov 09
Posts: 152
Credit: 16,181,924
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwatwat
Message 13482 - Posted: 11 Nov 2009 | 10:37:11 UTC

I think proyect you should consider more deeply about this matter. I have been colaborating with boinc for years. People get very upset when they see that after hours and hours of computing they don't get on time. I have the point of view most people can't afford/don't want to by a big computer or run it 24/7.
I have the oppinion that a lot of littles make a big. But if people feel they have worked for nothing just quit the proyect and join another one.
What is the longuer time the proyect can wait to get a result? Can the proyect wait for slowers computers/GPU to get a result?

Profile Damaraland
Send message
Joined: 7 Nov 09
Posts: 152
Credit: 16,181,924
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwatwat
Message 13537 - Posted: 13 Nov 2009 | 16:13:57 UTC - in response to Message 13482.
Last modified: 13 Nov 2009 | 16:18:02 UTC

Here I put an example. You can see that even that my computer got the result, it was sent to another computer. I know now that the problem is that my videocard "GeForce 9500 GT" is not fast enought for this proyect. My solution is buying a new graphic card. But anyway I see this as a waste of CPU cicles, and I guess many users just don't consider this as an option.

02 Nov 2009 20:01:49 UTC 07 Nov 2009 20:01:49 UTC Timed out - no response 0.00 0.00 --- ---
07 Nov 2009 20:06:45 UTC 13 Nov 2009 09:36:51 UTC Completed and validated 271,836.89 2,789.05 4,716.44 4,716.44
12 Nov 2009 20:20:40 UTC 17 Nov 2009 20:20:40 UTC In progress --- --- --- ---

PingWim
Send message
Joined: 10 Oct 09
Posts: 1
Credit: 83,082
RAC: 0
Level

Scientific publications
watwat
Message 15207 - Posted: 13 Feb 2010 | 11:45:54 UTC

I'm sure the GPUGRID administrators can see how many work units are processed before the deadline expires. So if only 40% is processed in time the deadline has to go up, when over 90% the deadline is fine ?

Would it be an idea to make deadline dependant on the processing power of the GPU ? I allow the GPU to be used to when my computer is idle and besides that my system is mostly on during the weekends and nights. Thus it's hard to make the deadline but I do want to help !!!

Profile liveonc
Avatar
Send message
Joined: 1 Jan 10
Posts: 292
Credit: 41,567,650
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwat
Message 15208 - Posted: 13 Feb 2010 | 11:56:03 UTC
Last modified: 13 Feb 2010 | 12:17:15 UTC

I'd argue that having 2 different priorities could hand over all the ASAP WU to those who want priority 1 & longer deadlines for all who want to be in priority 2. It shouldn't have to be a them or us, it should be them & us.

It would be nice too if Priority 1 got a 24 or 12 hour deadline. Just today I cancelled 2 WU's that had been running SUPER SLOW & still had 24-48 hours left to go. When Beta WU's don't run well, & take up all the space & time, other WU's might be delayed. So maybe a shorter deadline & smaller cache would ensure that at least BETA WU's would run out of time, & hopefully could automatically cancel when the deadline has passed, so that the PC could automatically pick up new WU's (if Daddy isn't home).

It's not segregation I'm talking about, because even on highways there is a fast & slow lane, even though a certain minimum speed is required as to not break the law, otherwise it's like Dennis Leary's song "@sshole", where he talks about driving really slow, on the ultra fast lane, while people behind him are going insane!
____________

Profile Michael Goetz
Avatar
Send message
Joined: 2 Mar 09
Posts: 124
Credit: 2,974,144
RAC: 0
Level
Ala
Scientific publications
watwatwatwatwatwatwat
Message 15210 - Posted: 13 Feb 2010 | 12:11:34 UTC - in response to Message 15208.

FYI -- you two are responding to a thread from February 2009.

In the intervening year, deadlines HAVE been increased -- I think they used to be 3 days back then, IIRC.

Profile robertmiles
Send message
Joined: 16 Apr 09
Posts: 409
Credit: 280,744,946
RAC: 507,501
Level
Asn
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 15447 - Posted: 25 Feb 2010 | 22:33:13 UTC - in response to Message 7052.


The only other project that is even rumored to be working on GPU version is Einstein@Home and that is a rumor only ... no word from project people ...


PrimeGrid is also investigating the use of GPU's for their sieving tasks, though any actual application is likely months away...




I'm already running the Primegrid GPU application, the Collatz GPU application, and the Einstein GPU application to keep the GPU on one of my other computers busy. It won't handle the GPUGRID application or the Milkyway application, and I'm not sure if I want to run anything from the SETI project again.

The last I noticed, the Einstein GPU application ran at a speed not much different from their CPU application - maybe 30% faster on some computers, and 30% slower on others. Still an early version, though.

I've heard of a GPU application called rc5, not BOINC-compatible. Close enough to the Yoyo@home BOINC project that they might possibly be getting close to having a wrapper that will make it BOINC-compatible. May be associated with this site:

http://www.distributed.net/

http://www.distributed.net/download/prerelease.php

Profile Michael Goetz
Avatar
Send message
Joined: 2 Mar 09
Posts: 124
Credit: 2,974,144
RAC: 0
Level
Ala
Scientific publications
watwatwatwatwatwatwat
Message 15449 - Posted: 25 Feb 2010 | 22:48:04 UTC - in response to Message 15447.

PrimeGrid has one CUDA application running. It's for their AP26 project, which is searching for an arithmetic progression of 26 primes. It's not quite as GPU intensive as some other projects, but it's not terrible either -- it runs a few times faster than the normal CPU version.

Post to thread

Message boards : Wish list : Deadline