Advanced search

Message boards : Number crunching : Gianni is not short ?!

Author Message
Thorvin
Send message
Joined: 5 Sep 10
Posts: 5
Credit: 92,334,530
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwat
Message 29995 - Posted: 16 May 2013 | 9:23:14 UTC

Hi all,

I just crunched a GIANNI WU which was marked short run, but it took more or less the same time than the NATHAN long runs but giving only half the credit.

that is normal ?

I'm running a HD7950 under Win7 64

Here is the link to the WU http://www.gpugrid.net/workunit.php?wuid=4452168

thanks for comments

BR
Markus

Thorvin
Send message
Joined: 5 Sep 10
Posts: 5
Credit: 92,334,530
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwat
Message 29999 - Posted: 16 May 2013 | 10:59:58 UTC

here is another one, that took even longer...

http://www.gpugrid.net/workunit.php?wuid=4452297

BR
Markus

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30006 - Posted: 16 May 2013 | 13:37:21 UTC - in response to Message 29999.

GIANNI_VIL1 WU's are fairly short (5h23min on a GTX660Ti is short).
Conversely, the issue could be perceived that Nathan's WU's are just as short, and shouldn't really be in the Long queue!

The Long queue is for WU's that take "(8-12 hours on fastest card)"
The Short queue is for WU's that take "(2-3 hours on fastest card)"

On the fastest GPUs (GTX680) both WU's take <4h, which is a lot closer to the Short queue than the Long queue.

While I have suggested that there should be a Medium Sized WU queue, saying as there is a 5h gap between the short and long queue and several different WU's at present, I don't have to maintain the server or the queues.

It would make more sense that these GIANNI_VIL1 WU's granted the same credit as Nate's WU's, one way or the other.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Thorvin
Send message
Joined: 5 Sep 10
Posts: 5
Credit: 92,334,530
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwat
Message 30007 - Posted: 16 May 2013 | 13:52:11 UTC

Thanks for the answer...

I don't know excatly how it works for GPU grid but shouldn't the credits be given by the amount of calculations done (or similar) and not by the queue the WU comes from ? Obviously this is not the case here ?!?

Markus

Profile Mumak
Avatar
Send message
Joined: 7 Dec 12
Posts: 92
Credit: 225,897,225
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 30008 - Posted: 16 May 2013 | 14:01:39 UTC
Last modified: 16 May 2013 | 14:02:18 UTC

The problem is probably because these tasks run at relatively low GPU utilization.
See here: http://www.gpugrid.net/forum_thread.php?id=3365&nowrap=true#29992

Thorvin
Send message
Joined: 5 Sep 10
Posts: 5
Credit: 92,334,530
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwat
Message 30009 - Posted: 16 May 2013 | 14:23:04 UTC

I watched the GPU utilization of the second one because it was my first idea....

I saw a general level of about 80-95% with some drops to about 60% for seconds. Overall stable on a high level.

Earlier reported low utilization WU's had a much lower overall level than these.

Can anybody watch on their GPU's, I'm out of GIANNI right now ;-)

Thorvin
Send message
Joined: 5 Sep 10
Posts: 5
Credit: 92,334,530
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwat
Message 30011 - Posted: 16 May 2013 | 14:41:16 UTC

I just saw that I mentioned the wrong GPU in my first post ...

I'm using a GTX 660 Ti and for sure not the ATI card for GPUGrid

Markus

Profile nate
Send message
Joined: 6 Jun 11
Posts: 124
Credit: 2,928,865
RAC: 0
Level
Ala
Scientific publications
watwatwatwatwat
Message 30017 - Posted: 16 May 2013 | 17:17:34 UTC

There are two main reasons why more credit is awarded on the long queue:
1) Greater risk for the cruncher to crunch (typically) much longer WUs. If a simulation crashes after 18 hours, that's a much bigger loss than a crash after 2 or 6 hours. This is especially true for older/slower cards.
2) To encourage/reward users who dedicate their computers to the long hours crunching that long WUs require. With he short queue, you can run a WU while you sleep or run errands, for example, and by the time you wake up or come home it's finished, and you can use your computer for other things. Dedicating a gpu/cpu/computer to run on long queue means you basically can't use it for other things, such as work, entertainment, etc., and so higher credits reward them for that.

Gianni's WUs may be slightly long for the short queue, but my recent tasks were definitely short for the long queue. The reason was that, at the time, my WUs were the only work for anyone to crunch, so I didn't want to make them too long in case people who typically crunch on short tasks wanted WUs to do, but couldn't sacrifice the time. Basically, it was due to the circumstances at that time. We have had a few weeks recently where I was the only one who had work for you guys, which was never a common situation for us. However, we keep adding more users, and it is becoming harder to come up with ideas fast enough (which is a good problem to have!). We are also trying to bring in new scientists!

Matt
Avatar
Send message
Joined: 11 Jan 13
Posts: 216
Credit: 846,538,252
RAC: 0
Level
Glu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30036 - Posted: 17 May 2013 | 2:47:56 UTC

Also seeing the same issue for the GIANNIs - 19446.58 seconds for the most recent. This is more time than the NATHAN WUs.

I saw the GIANNI WUs using 70 - 75% of my GTX680 on average. The NATHANs and NOELIAs tend to run around 85% on average. (I've never had a problem with NOELIA units.)

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1112
Credit: 6,162,416,256
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30037 - Posted: 17 May 2013 | 3:04:56 UTC - in response to Message 30017.
Last modified: 17 May 2013 | 3:05:42 UTC

There are two main reasons why more credit is awarded on the long queue:
1) Greater risk for the cruncher to crunch (typically) much longer WUs. If a simulation crashes after 18 hours, that's a much bigger loss than a crash after 2 or 6 hours. This is especially true for older/slower cards.
2) To encourage/reward users who dedicate their computers to the long hours crunching that long WUs require. With he short queue, you can run a WU while you sleep or run errands, for example, and by the time you wake up or come home it's finished, and you can use your computer for other things. Dedicating a gpu/cpu/computer to run on long queue means you basically can't use it for other things, such as work, entertainment, etc., and so higher credits reward them for that.

Number 1 has some validity especially in the light of the recent rash of server aborted WUs. But 2x the credit differential seems excessive. Number 2: I'm not so sure about the logic at all. It certainly doesn't factor into my decision making in the least. In fact, the large credit differential causes some of us to move to different projects rather than do the short WUs and basically feel like 2nd class citizens. Four of my GPUs that happily ran long run Nathans and Tonis are unfortunately once again at other projects.

John C MacAlister
Send message
Joined: 17 Feb 13
Posts: 181
Credit: 144,871,276
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwat
Message 30070 - Posted: 18 May 2013 | 16:53:37 UTC - in response to Message 30017.

Hi, Nate:

I particularly like the last two sentences in your post:


However, we keep adding more users, and it is becoming harder to come up with ideas fast enough (which is a good problem to have!). We are also trying to bring in new scientists!


Regards,

John

[AF>France>Aquitaine>Cote...
Avatar
Send message
Joined: 26 Oct 09
Posts: 4
Credit: 2,002,875
RAC: 0
Level
Ala
Scientific publications
watwat
Message 30161 - Posted: 21 May 2013 | 15:47:15 UTC

I'm really disappointed : GIANNI tasks are considered as short tasks when lasting about 3 times more than other "small" tasks, but their deadline is the same. They don't terminate in time if I'm sharing resources on my computer.
As long as the deadline remains the same, I'll have to manually abort all GIANNI tasks.
____________

FoldingNator
Send message
Joined: 1 Dec 12
Posts: 24
Credit: 60,122,950
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwat
Message 30172 - Posted: 21 May 2013 | 22:37:13 UTC
Last modified: 21 May 2013 | 22:55:13 UTC

I'm glad to read this thread! Today I started again with crunching since 2 months ago. So I started with short runs GIANNI. But he's already crunching since 16:00 pm and is almost (00:50 am). So a run took more than 31.500 seconds to complete @ my GTX570!!! I thought I had a hard- or software error... I'm glad it isn't only me.

A screenshot over here (Dutch Boinc): http://prntscr.com/16117w
I've taked that screenshot 15 minutes ago. 8 hours and 15 minutes at that moment! :S :S

One of the workunits: http://www.gpugrid.net/workunit.php?wuid=4468529

Finally it's done... eh, only 1 WU. The other one caused a lockup of my computer in the last 2 minutes. Arghhh -> computation error -> #$%^&* -_-'

nanoprobe
Send message
Joined: 26 Feb 12
Posts: 184
Credit: 222,376,233
RAC: 0
Level
Leu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 30173 - Posted: 21 May 2013 | 23:02:44 UTC - in response to Message 30172.
Last modified: 21 May 2013 | 23:06:44 UTC

I've only had 4 Gianni tasks since returning to GPUGrid. One ended in with an exit status 255 (0xff) Unknown error number.

The other 3 completed successfully in about 5 hours each. 660 TI, Linux Mint.

Ano
Send message
Joined: 15 Feb 10
Posts: 2
Credit: 5,540,501
RAC: 0
Level
Ser
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 30194 - Posted: 22 May 2013 | 11:03:41 UTC

Will real short tasks come back?
I mean I thought those were really abnormally long, but it's worst than that.
In my tasks list, long tasks (I have 2) are all below 68200s of working time, while the recent "short" ones (I have 3) are over 68200s (2 are over 70000s).

I know I don't have what you call a "fastest card", but 19 hours of work for a 2~3 hours task, the gap is a bit overkill.

FoldingNator
Send message
Joined: 1 Dec 12
Posts: 24
Credit: 60,122,950
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwat
Message 30251 - Posted: 23 May 2013 | 9:09:51 UTC - in response to Message 30194.

I know I don't have what you call a "fastest card", but 19 hours of work for a 2~3 hours task, the gap is a bit overkill.

Lol, indeed... :/

Profile Carlesa25
Avatar
Send message
Joined: 13 Nov 10
Posts: 328
Credit: 72,619,453
RAC: 251
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30270 - Posted: 23 May 2013 | 17:19:43 UTC

Hi: Actually these tasks should not be considered SHORT. In my GTX 590 are at about >25,000 sec.

That is by no means a short task, based on the parameters it has set GPUGRID.

As in the past that have also been discussed, the evaluation criteria for this project are very random and little respect for their own guidelines. Greetings.

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30274 - Posted: 23 May 2013 | 19:12:04 UTC - in response to Message 30270.

On the fastest GPU (now a GTX680) Gianni WU's take <4h. That is less than an hour outside the Short queue's approximation of 2-3 hours, and a very long way from typical Long queue WU's, on the GTX680 (8 to 12h).

____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Profile Carlesa25
Avatar
Send message
Joined: 13 Nov 10
Posts: 328
Credit: 72,619,453
RAC: 251
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30275 - Posted: 23 May 2013 | 19:31:57 UTC - in response to Message 30274.
Last modified: 23 May 2013 | 19:33:19 UTC

On the fastest GPU (now a GTX680) Gianni WU's take <4h. That is less than an hour outside the Short queue's approximation of 2-3 hours, and a very long way from typical Long queue WU's, on the GTX680 (8 to 12h).



Hello: ¿ By that criterion within little time calculation base is the GTX TITAN ...?.

Sorry but I think it would have weighed an average / high among the series GTX5xx GTX6xx and not everyone has the GTX680 or TITAN, GTX780 or soon. Greetings.

Vagelis Giannadakis
Send message
Joined: 5 May 13
Posts: 187
Credit: 349,254,454
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 30279 - Posted: 23 May 2013 | 22:16:27 UTC
Last modified: 23 May 2013 | 22:17:15 UTC

I completed a GIANNI (I0R56-GIANNI_VIL1-2-3-RND5973_1) in ~9 hours, receiving ~26K. This means a day of these WUs would get me ~70K. A single Long SDOERR got me 135K in 18 hours. In average, just doing SDOERR Longs will get me ~180K per day.

Sorry, no contest, I'm disabling Short WUs. At least these GIANNIs are seriously wrong-rated as Short and their gain is seriously low. They need to be re-evaluated.

Edit: I have a GTX 650Ti

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1112
Credit: 6,162,416,256
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30281 - Posted: 23 May 2013 | 23:37:45 UTC - in response to Message 30279.

Sorry, no contest, I'm disabling Short WUs. At least these GIANNIs are seriously wrong-rated as Short and their gain is seriously low. They need to be re-evaluated.

Edit: I have a GTX 650Ti

Personally I find the whole short queue long queue issue with the huge credit differential irritating. I understand the rational, I buy part of it. Reeks of elitism though IMO. People want to help out here because they like the goals of the project. Sometimes the project makes it hard to do.

Profile dskagcommunity
Avatar
Send message
Joined: 28 Apr 11
Posts: 456
Credit: 817,865,789
RAC: 0
Level
Glu
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30288 - Posted: 24 May 2013 | 11:42:34 UTC

I was interesting about gianni and from my results short giani (as example 23k secs) runs longer! then a long old nathan (as example 21k on the same card). But giani pays 31k credits while nathan paid 70k. So the argument it cant be run in long queue is not right. Or the old nathans are not right in the long queue ;) from the creditside it is right in the long queue. I wanted to mix one of the machines with short and long to get a better place on the next publication but i deactivaded short too at the moment ^^
____________
DSKAG Austria Research Team: http://www.research.dskag.at



Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30296 - Posted: 24 May 2013 | 15:04:50 UTC - in response to Message 30288.
Last modified: 24 May 2013 | 15:05:36 UTC

The previous Nathan's were put in the Long queue as there was no other work available, and most people are attached to the long queue. So it was just to accommodate crunchers that don't check the forums frequently. They were also made to a length to accommodate everyone. Also, some testing was being done around that time, and I think they might have been using the short queue for that on occasions, as well as the beta queue.

Nate has already explained his reasons, and in this thread,
http://www.gpugrid.net/forum_thread.php?id=3370&nowrap=true#30017
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Vagelis Giannadakis
Send message
Joined: 5 May 13
Posts: 187
Credit: 349,254,454
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 30316 - Posted: 24 May 2013 | 20:48:52 UTC - in response to Message 30281.

Sorry, no contest, I'm disabling Short WUs. At least these GIANNIs are seriously wrong-rated as Short and their gain is seriously low. They need to be re-evaluated.

Edit: I have a GTX 650Ti

Personally I find the whole short queue long queue issue with the huge credit differential irritating. I understand the rational, I buy part of it. Reeks of elitism though IMO. People want to help out here because they like the goals of the project. Sometimes the project makes it hard to do.

I am interested in the goals of GPUG and the science behind it. If I wouldn't - and were only interested in the credits - I would be running DistrRTgen, which "pays" good and like clockwork. I'm also a long-time WCG cruncher.

I guess credit and the associated rankings is the vanity-satisfying "fix" we get that keeps us going! :)

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30320 - Posted: 24 May 2013 | 21:47:03 UTC - in response to Message 30275.

Sorry but I think it would have weighed an average / high among the series GTX5xx GTX6xx and not everyone has the GTX680 or TITAN, GTX780 or soon.

Well.. giving the expected run time for the fastest cards is very different from expecting everyone to have the fastest card. As cards get faster GPU-Grid has to adjust the WU durations in order not to overwhelm the server.

You can argue that you want shorter WUs or a different time estimate (your average.. which would require more work to calculate), but don't blame them for sticking to their word.

[I think the current Giannis are too long, too.. but that's not a fundamental issue and could easily be fixed, at least when the next WUs are submitted]

MrS
____________
Scanning for our furry friends since Jan 2002

Stefan
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 5 Mar 13
Posts: 348
Credit: 0
RAC: 0
Level

Scientific publications
wat
Message 30325 - Posted: 25 May 2013 | 8:34:20 UTC - in response to Message 30316.

@Vagelis Hey, a fellow Greek! :D Χαίρομαι που βλέπω και Έλληνες στο πρότζεκτ μας! Ο πρώτος είσαι που έχω δεί ως τώρα. Ευχαριστούμε πολύ για την συνεισφορά!

Vagelis Giannadakis
Send message
Joined: 5 May 13
Posts: 187
Credit: 349,254,454
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 30349 - Posted: 25 May 2013 | 14:08:29 UTC - in response to Message 30325.

@Vagelis Hey, a fellow Greek! :D Χαίρομαι που βλέπω και Έλληνες στο πρότζεκτ μας! Ο πρώτος είσαι που έχω δεί ως τώρα. Ευχαριστούμε πολύ για την συνεισφορά!

Γεια σου φίλε Stefan :)

I'm glad to be a part of GPUG! Let's hope we'll see more Greeks joining!

John C MacAlister
Send message
Joined: 17 Feb 13
Posts: 181
Credit: 144,871,276
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwat
Message 30358 - Posted: 25 May 2013 | 16:02:59 UTC

Welcome Vagelis!!

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30381 - Posted: 26 May 2013 | 8:54:30 UTC

That's weird, the numbers and operators are missing between the greek letters in your formulas.. ;)

MrS
____________
Scanning for our furry friends since Jan 2002

Vagelis Giannadakis
Send message
Joined: 5 May 13
Posts: 187
Credit: 349,254,454
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 30424 - Posted: 26 May 2013 | 18:04:11 UTC - in response to Message 30358.
Last modified: 26 May 2013 | 18:25:56 UTC

Welcome Vagelis!!

Thanks! :)

That's weird, the numbers and operators are missing between the greek letters in your formulas.. ;)

MrS

:)

[AF>France>Aquitaine>Cote...
Avatar
Send message
Joined: 26 Oct 09
Posts: 4
Credit: 2,002,875
RAC: 0
Level
Ala
Scientific publications
watwat
Message 30436 - Posted: 27 May 2013 | 0:18:36 UTC
Last modified: 27 May 2013 | 0:33:17 UTC

My concern is not the credit, but the deadline.

PLEASE think for a while that all computers running GPUGRID don't have the fastest GPU.

My home computer has a GT 540 M GPU, and GIANNI WUs last from 48 to 72 hours. If I have 2 projects sharing GPU resources at 100-100, it takes 4 to 6 days to complete. And the deadline remains desesperately set to ... 5 days, the same than shortest WUs.

I16R3-GIANNI_VIL1-0-3-RND7631_3 has used 242,046.30 seconds (67 hours) of my GPU time and was out of time when completed (without error). BINGO ! You've got the result and I've got 0 credit.

I1R8-GIANNI_VIL1-1-3-RND1955_1 has been loaded during the execution of the previous one. When starting, there were only 2 days remaining to complete, so I had to abort it. I've got 0 credit, but at least I have spared my GPU time.

I've a very small chance to see the current WU (I16R89-GIANNI_VIL1-2-3-RND3587_1) complete in time because it will only last about 52 hours.

Please increase the deadline for Gianni tasks.
____________

[AF>France>Aquitaine>Cote...
Avatar
Send message
Joined: 26 Oct 09
Posts: 4
Credit: 2,002,875
RAC: 0
Level
Ala
Scientific publications
watwat
Message 30440 - Posted: 27 May 2013 | 9:34:43 UTC - in response to Message 30436.
Last modified: 27 May 2013 | 9:53:34 UTC

WAOUH.
I16R89-GIANNI_VIL1-2-3-RND3587_1 passed OK whilst completed half an hour after the official deadline.

I tailored my resource sharing between projects with "normal" WUs in order to get them in time. If you suddenly increase the workload by 2 or 3, you should increase the deadline by the same factor.

But nobody here seems to take my remarks into account. It's a shame.
So I repeat my demand :

Please increase the deadline for Gianni tasks.
____________

Profile Carlesa25
Avatar
Send message
Joined: 13 Nov 10
Posts: 328
Credit: 72,619,453
RAC: 251
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30442 - Posted: 27 May 2013 | 12:00:35 UTC - in response to Message 30440.

Hello: A card that does not allow complete a task (short or long) in GPUGRID on a GPU 24H is not suitable for this project, as designed, it's that simple. Greetings.

Vagelis Giannadakis
Send message
Joined: 5 May 13
Posts: 187
Credit: 349,254,454
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 30446 - Posted: 27 May 2013 | 14:52:00 UTC - in response to Message 30440.

WAOUH.
I16R89-GIANNI_VIL1-2-3-RND3587_1 passed OK whilst completed half an hour after the official deadline.

I tailored my resource sharing between projects with "normal" WUs in order to get them in time. If you suddenly increase the workload by 2 or 3, you should increase the deadline by the same factor.

But nobody here seems to take my remarks into account. It's a shame.
So I repeat my demand :

Please increase the deadline for Gianni tasks.

You seem to be using two laptops. While their GPUs may be powerful enough for general laptop use and gaming, even for some BOINC GPU projects, unfortunately they are not suitable for GPUGRID.

I had a GT 440, close in specs with one of your laptop cards, and it only managed to complete work within the deadlines, forget bonuses! I upgraded to a GTX 650TI, which manages to crunch the longest WUs in just under 24h.

A sad fact perhaps, but GPUGRID seems to be a demanding, high-maintenance project: not only does it need powerful GPUs, you also have to keep an eye on what WUs you get, how they progress, abort them quickly (so somebody else gets them quickly) if you spot something going wrong, etc.

To be honest, I am enjoying this! At least for the time being. And of course, I didn't have a problem to upgrade my GPU to accommodate the project's crunching appetite! That may not be the case with everybody of course and that's perfectly fine! As long as you wish to contribute with number crunching, there are many projects out there that would appreciate your contribution!

[AF>France>Aquitaine>Cote...
Avatar
Send message
Joined: 26 Oct 09
Posts: 4
Credit: 2,002,875
RAC: 0
Level
Ala
Scientific publications
watwat
Message 30459 - Posted: 27 May 2013 | 20:51:23 UTC - in response to Message 30446.
Last modified: 27 May 2013 | 21:33:10 UTC

Usual GPUGRID "short tasks" last around 24 hours on my 2 fastest hometoys. I'm not competing for bonuses, so I tried to share their limited GPU resources between 3 BOINC projects at 100-100-100, and it worked.

Maybe it's a sucker's idea to do so with my laptoys, but they're good enough for my needs: no games, no video, just little home office, mail access, surf on internet and a CUDA GPU board for BOINC. One of them has a very comfortable backlighted keyboard, and this is quite more important than to have the fastest GPU board. You know what? I'm happy. I'm an old timer. I still have a 12-years old Core 2 Duo crunching Enigma 24/7.

My concern is that Gianni WUs are not short tasks: they last 2 times more than usual ones but they have the same deadline.

To solve this issue, I'll try to dedicate one of my ridiculous laptoys at 100% to GPUGRID.
____________

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30498 - Posted: 28 May 2013 | 18:48:46 UTC

While I agree that those Giannis are too long for the short queue, you should be aware of how GPU-Grid works: WUs are generated based on the results of returning WUs. They can launch parameter sweeps etc. in parallel, but at some point they need entire strings of WUs finished to evaluate the results and decide what's next run. Simply extending the deadline would make the results take longer to be returned and hence slow the project down. Even if overall throughput was increased because more slower GPUs would participate.

MrS
____________
Scanning for our furry friends since Jan 2002

Post to thread

Message boards : Number crunching : Gianni is not short ?!

//