Advanced search

Message boards : Number crunching : An open letter to the manager of GPUGRID

Author Message
tomba
Send message
Joined: 21 Feb 09
Posts: 497
Credit: 700,690,702
RAC: 0
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 27326 - Posted: 17 Nov 2012 | 17:47:45 UTC
Last modified: 17 Nov 2012 | 17:55:07 UTC

I give you freely my GPU time, and the cost of the electricity to run it, because I believe in GPUGRID. What do you do in return? Send me WUs that make my PC unusable.

Unusable? How about 10 seconds to react to a click on the Start button? How about 18 seconds to open Task Manager? How about 11 seconds to go from one browser tab to another?

I have eight CPU processors, Windows 7 and six Megs of RAM, more than enough to give instantaneous response to any of these actions, under normal circumstances.

Problem? NATHAN WUs. My log of all WUs downloaded these past 30 days shows, for non-benHP NATHANs:

 69 received, of which:
 17 ran with no problems,
 5 ran but with performance problems,
 7 gave ‘error while computing’ and
 40 were aborted by me because they were killing my PC.

I can tell you the WU names of all the 69 received, by category.

A 25% success rate should be a concern for you. It certainly is for me! I have to nurse PC-crippling WUs. In the last hour I’ve aborted six. I signed up for hands-off background processing. That is not what’s happening!

I do hope you will work with me to get to the bottom on this misery…

Tom
____________

Profile AdamYusko
Send message
Joined: 29 Jun 12
Posts: 26
Credit: 21,540,800
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwat
Message 27328 - Posted: 17 Nov 2012 | 21:38:00 UTC

Just curious if you have ever given thought to only crunching these projects when you are not using your computer/ GPU? By the nature of all of these tasks they are there to use up any excess system power you have, but if your need for performance suddenly spikes by suddenly requiring more than you were in processing power for your own use then its only natural for the program to wait until that extra power can be alocated, which can be on the order of several tens of seconds.

While I do not have any projects that I myself administer, I find it funny when people expect their machines to react like they have no extra load put on them when crunching.

Think about it this way if you are employed to be someone's assistant, you can respond to all their requests instantaneously if they haven't assigned you extra projects and you are only focusing on meeting your bosses demands. But if while you are their assistant you are told you need to also sort and file all of the documents in a few hundred person organization, and see your to your bosses needs, when suddenly when your boss asks you to do something, you may want to finishing filing what you were then see to your bosses needs. But, if you were told to only file when your boss was not in the office that would never happen.

So you need to decide are you hunting credits, or are you doing it for the good of the science. If you are doing it for the science, then aborting all these extra work units is detrimental to the science, and you should just set your machine to only crunch when you are not using it.

jlhal
Send message
Joined: 1 Mar 10
Posts: 147
Credit: 1,077,535,540
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 27329 - Posted: 17 Nov 2012 | 23:10:10 UTC - in response to Message 27328.

Just curious if you have ever given thought to only crunching these projects when you are not using your computer/ GPU? By the nature of all of these tasks they are there to use up any excess system power you have
,...
and you should just set your machine to only crunch when you are not using it.


I agree that GPUGRID , as well as other projects using CPU only, is designed for using part or all of the idle time of the system. So it normally sholud let the system responsive if well configured.

But I think that tomba's problem is not with the WUs themselves , a lot of crunchers have no problem with that, except when some WUs are not well designed..

According to me, the cause is either :

- a configuration problem (GPUGRID and/or system)
- a hardware problem (GPU and/or motherboard)
- an inadapted combination of CPU/OS/GPU ...

This seems very fuzzy, but the problem seems not WU related...

@tomba : Please give a COMPLETE description of your system in terms of hardware brands, bios and drivers versions, RAM size and speed etc...
I know we already discussed about your problem in another thread but I'm sure every one here is willing to help solve this "mystery"...

Regards
____________
Lubuntu 16.04.1 LTS x64

Profile Chilean
Avatar
Send message
Joined: 8 Oct 12
Posts: 98
Credit: 385,652,461
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 27330 - Posted: 18 Nov 2012 | 4:12:03 UTC

The only problem I get is laggy graphics when running the long runs. Which is why I'm gonna stick to the normal runs.

werdwerdus
Send message
Joined: 15 Apr 10
Posts: 123
Credit: 1,004,473,861
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 27343 - Posted: 19 Nov 2012 | 8:40:14 UTC
Last modified: 19 Nov 2012 | 8:41:17 UTC

disable "Use GPU while computer is in use", have you even tried?

I have had similar "problems" but this is the way I see the issue: NATHAN has gotten some of his work units SO EFFICIENT that they use up practically ALL of gpu's resources. This is a GOOD THING for the project IMO! There should be no idle resources in a well-coded application. The only work units that have given me "problems" are the ones running at 99% gpu utilization.

If you want to use your system for other tasks then you will have to give up some of those resources, plain and simple (at least I think it's simple).
____________
XtremeSystems.org - #1 Team in GPUGrid

tomba
Send message
Joined: 21 Feb 09
Posts: 497
Credit: 700,690,702
RAC: 0
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 27347 - Posted: 19 Nov 2012 | 13:37:35 UTC - in response to Message 27326.

I give you freely my GPU time, and the cost of the electricity to run it, because I believe in GPUGRID. What do you do in return? Send me WUs that make my PC unusable...


Call of the dogs!! I found the solution. I did not know that, after unchecking "Use GPU while computer is in use" in BOINC / Tools / Computing Preferences / Processor Usage, I had to activate this preference in BOINC / Activity by checking "Use GPU based on preferences".

Now when I wiggle the mouse in BOINC / Tasks, the GPUGRID WU immediately gives "Waiting to run". Perfect!

Apologies to everyone, Tom

____________

Old man
Send message
Joined: 24 Jan 09
Posts: 42
Credit: 16,676,387
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 27348 - Posted: 19 Nov 2012 | 13:38:51 UTC - in response to Message 27329.

Just curious if you have ever given thought to only crunching these projects when you are not using your computer/ GPU? By the nature of all of these tasks they are there to use up any excess system power you have
,...
and you should just set your machine to only crunch when you are not using it.


I agree that GPUGRID , as well as other projects using CPU only, is designed for using part or all of the idle time of the system. So it normally sholud let the system responsive if well configured.

But I think that tomba's problem is not with the WUs themselves , a lot of crunchers have no problem with that, except when some WUs are not well designed..

According to me, the cause is either :

- a configuration problem (GPUGRID and/or system)
- a hardware problem (GPU and/or motherboard)
- an inadapted combination of CPU/OS/GPU ...

This seems very fuzzy, but the problem seems not WU related...

@tomba : Please give a COMPLETE description of your system in terms of hardware brands, bios and drivers versions, RAM size and speed etc...
I know we already discussed about your problem in another thread but I'm sure every one here is willing to help solve this "mystery"...

Regards


Hey. No need write so long because real broblem is here:
I have eight CPU processors, Windows 7 and six Megs of RAM, more than enough to give instantaneous response to any of these actions, under normal circumstances.


6 megabytes of memory + 8 processor cores is not a good combination. Update the machine for at least 6 GB of memory, so we can think about the problem again. Now, the problem is far too small amount of memory.

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 27352 - Posted: 19 Nov 2012 | 15:25:58 UTC - in response to Message 27348.

That was a typo!
____________
FAQ's

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

tomba
Send message
Joined: 21 Feb 09
Posts: 497
Credit: 700,690,702
RAC: 0
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 27353 - Posted: 19 Nov 2012 | 15:33:26 UTC - in response to Message 27352.

That was a typo!


I assume you're referencing my "six megs". Right! Six GIGS, of course...

Sorry.

Tom

____________

tomba
Send message
Joined: 21 Feb 09
Posts: 497
Credit: 700,690,702
RAC: 0
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 27355 - Posted: 19 Nov 2012 | 16:44:22 UTC - in response to Message 27347.

Now when I wiggle the mouse in BOINC / Tasks, the GPUGRID WU immediately gives "Waiting to run". Perfect!


Not quite perfect... When a WU resumes it backs-off a few percentage points of processing so it's repeating what it's already done.

I guess it's resuming at the last checkpoint. I've looked and looked for the parameter that lets me specify the checkpoint period but I can't find it!

Any ideas?

Tom

____________

jlhal
Send message
Joined: 1 Mar 10
Posts: 147
Credit: 1,077,535,540
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 27356 - Posted: 19 Nov 2012 | 16:52:31 UTC - in response to Message 27347.


Now when I wiggle the mouse in BOINC / Tasks, the GPUGRID WU immediately gives "Waiting to run". Perfect!


I think this is not a normal behaviour if crunching has to be suspended every time you move the mouse ?!?!?!?

I never experienced this situation with any GPU !
____________
Lubuntu 16.04.1 LTS x64

tomba
Send message
Joined: 21 Feb 09
Posts: 497
Credit: 700,690,702
RAC: 0
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 27357 - Posted: 19 Nov 2012 | 17:05:47 UTC - in response to Message 27356.

I think this is not a normal behaviour if crunching has to be suspended every time you move the mouse ?!?!?!?!


It is normal if you have a modest GPU - GTX 460 - and many WUs kill your PC performance and you set the GPU to not run when you are using your PC.

____________

Profile Carlesa25
Avatar
Send message
Joined: 13 Nov 10
Posts: 324
Credit: 72,394,453
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 27358 - Posted: 19 Nov 2012 | 18:22:11 UTC - in response to Message 27356.

Hi, The best solution is to have a unique GPU for display and the other GPUs for BOINC, if your requirements allow.

The GPU of the screen is excluded from use BOINC (cc_config.xml) and no problems of delays etc ...

I'm using a GT 620 for the screen (pass a GTS450 coming soon) and a GTX 590 for BOINC without delay and maximum performance. Greetings.

tomba
Send message
Joined: 21 Feb 09
Posts: 497
Credit: 700,690,702
RAC: 0
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 27359 - Posted: 19 Nov 2012 | 18:25:42 UTC - in response to Message 27358.

Hi, The best solution is to have a unique GPU for display and the other GPUs for BOINC.


Lucky to have two GPUs!!

____________

Profile Carlesa25
Avatar
Send message
Joined: 13 Nov 10
Posts: 324
Credit: 72,394,453
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 27360 - Posted: 19 Nov 2012 | 19:19:36 UTC - in response to Message 27359.

Hi, The best solution is to have a unique GPU for display and the other GPUs for BOINC.


Lucky to have two GPUs!!


Hello: A Zotac GeForce GTS 450 ECO Edition with 192 CudaCores no extra power needed is about 75 € and performs perfectly for almost everything (except applications or games very severe).

sardine
Send message
Joined: 17 Sep 08
Posts: 4
Credit: 67,875
RAC: 0
Level

Scientific publications
wat
Message 27369 - Posted: 20 Nov 2012 | 18:10:23 UTC
Last modified: 20 Nov 2012 | 18:13:07 UTC

In my opinion, the problem should be the usage of GDDR memory. When I run this program, the loading of GDDR is almost 100%. But the loading of system memory is very low. These code may claim too large variables on the video card waiting to progressing. I don't know how sources code looks like but this is what i thought.

by the way, the thinking of distributed computing is the same as run on a supercomputer. These program should not be design to run on a powerful computer but an old one ( like the generation of one or two years ago ). Or you will lost a lot of volunteers while more and more people have the same feeling like tomba. Meanwhile, that will be a bad news to you.

werdwerdus
Send message
Joined: 15 Apr 10
Posts: 123
Credit: 1,004,473,861
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 27370 - Posted: 20 Nov 2012 | 18:29:33 UTC - in response to Message 27369.

In my opinion, the problem should be the usage of GDDR memory. When I run this program, the loading of GDDR is almost 100%. But the loading of system memory is very low. These code may claim too large variables on the video card waiting to progressing. I don't know how sources code looks like but this is what i thought.

by the way, the thinking of distributed computing is the same as run on a supercomputer. These program should not be design to run on a powerful computer but an old one ( like the generation of one or two years ago ). Or you will lost a lot of volunteers while more and more people have the same feeling like tomba. Meanwhile, that will be a bad news to you.


the 400 series cards are already 2-2.5 years old and still work on GPUGrid. GPUGrid still works on the 200 series which is in fact 3-4 years old, so what is your point?
____________
XtremeSystems.org - #1 Team in GPUGrid

sardine
Send message
Joined: 17 Sep 08
Posts: 4
Credit: 67,875
RAC: 0
Level

Scientific publications
wat
Message 27371 - Posted: 20 Nov 2012 | 18:42:43 UTC - in response to Message 27370.
Last modified: 20 Nov 2012 | 18:45:25 UTC

In my opinion, the problem should be the usage of GDDR memory. When I run this program, the loading of GDDR is almost 100%. But the loading of system memory is very low. These code may claim too large variables on the video card waiting to progressing. I don't know how sources code looks like but this is what i thought.

by the way, the thinking of distributed computing is the same as run on a supercomputer. These program should not be design to run on a powerful computer but an old one ( like the generation of one or two years ago ). Or you will lost a lot of volunteers while more and more people have the same feeling like tomba. Meanwhile, that will be a bad news to you.


the 400 series cards are already 2-2.5 years old and still work on GPUGrid. GPUGrid still works on the 200 series which is in fact 3-4 years old, so what is your point?


My points is these program are not compatible to these card well. Code could be modified for volunteers. Or GPUGRID think these are not important to this project.

BOINC
Send message
Joined: 26 Jan 13
Posts: 3
Credit: 7,204,825
RAC: 0
Level
Ser
Scientific publications
watwat
Message 28487 - Posted: 13 Feb 2013 | 10:48:02 UTC - in response to Message 27326.

Don't be a barney....it is wise to do a bit of research prior to getting out the blame thrower..just a thought.

Profile MJH
Project administrator
Project developer
Project scientist
Send message
Joined: 12 Nov 07
Posts: 696
Credit: 27,266,655
RAC: 0
Level
Val
Scientific publications
watwat
Message 28495 - Posted: 13 Feb 2013 | 16:50:58 UTC - in response to Message 27355.


I guess it's resuming at the last checkpoint. I've looked and looked for the parameter that lets me specify the checkpoint period but I can't find it!


It's a property of the WU and is not cruncher-configurable.

MJH

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 28558 - Posted: 17 Feb 2013 | 13:01:25 UTC - in response to Message 28495.

I guess it's resuming at the last checkpoint.

There's another setting for this: "leave apps in memory while suspended". Activating this in your BOINC prefs should eleminate this loss. And set your tiem to resume after suspending to a few minutes. Restarting/resuming too often may crash the driver (I had this happen running PrimeGrid, not sure about GPU-Grid).

MrS
____________
Scanning for our furry friends since Jan 2002

Richard Haselgrove
Send message
Joined: 11 Jul 09
Posts: 1451
Credit: 3,575,929,351
RAC: 345,202
Level
Arg
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 28597 - Posted: 18 Feb 2013 | 17:38:20 UTC - in response to Message 28558.

I guess it's resuming at the last checkpoint.

There's another setting for this: "leave apps in memory while suspended". Activating this in your BOINC prefs should eleminate this loss. And set your tiem to resume after suspending to a few minutes. Restarting/resuming too often may crash the driver (I had this happen running PrimeGrid, not sure about GPU-Grid).

MrS

"Leave applications in memory" has (deliberately and by design) no effect for GPU applications. Most consumer GPU cards (perhaps you might come to a different answer for the newer high-power cards people choose here) don't have enough VRAM to be able to hold a whole set of tasks in memory, and to keep them there while a whole different project is switched in for processing. And there's no operating system support for a pagefile or virtual memory, such as has been used to support the over-commitment of main system RAM for many decades.

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 28602 - Posted: 18 Feb 2013 | 21:43:30 UTC - in response to Message 28597.

Thanks Richard, I didn't know this. Makes perfect sense now!

MrS
____________
Scanning for our furry friends since Jan 2002

Betting Slip
Send message
Joined: 5 Jan 09
Posts: 670
Credit: 2,498,095,550
RAC: 0
Level
Phe
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 28757 - Posted: 25 Feb 2013 | 8:42:16 UTC - in response to Message 27326.

I give you freely my GPU time, and the cost of the electricity to run it, because I believe in GPUGRID. What do you do in return? Send me WUs that make my PC unusable.

Unusable? How about 10 seconds to react to a click on the Start button? How about 18 seconds to open Task Manager? How about 11 seconds to go from one browser tab to another?

I have eight CPU processors, Windows 7 and six Megs of RAM, more than enough to give instantaneous response to any of these actions, under normal circumstances.

Problem? NATHAN WUs. My log of all WUs downloaded these past 30 days shows, for non-benHP NATHANs:

 69 received, of which:
 17 ran with no problems,
 5 ran but with performance problems,
 7 gave ‘error while computing’ and
 40 were aborted by me because they were killing my PC.

I can tell you the WU names of all the 69 received, by category.

A 25% success rate should be a concern for you. It certainly is for me! I have to nurse PC-crippling WUs. In the last hour I’ve aborted six. I signed up for hands-off background processing. That is not what’s happening!

I do hope you will work with me to get to the bottom on this misery…

Tom



Tom,

I think I have posted this before to you.

You cannot run a Nathan unit on a GTX460 and use your computer at the same time.

I have the same card and always uncheck the "Use GPU while computer in use" box for Nathan units, All other units can run fine with computer is in use.
____________
Radio Caroline, the world's most famous offshore pirate radio station.
Great music since April 1964. Support Radio Caroline Team -
Radio Caroline

wiyosaya
Send message
Joined: 22 Nov 09
Posts: 114
Credit: 589,114,683
RAC: 0
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 28929 - Posted: 3 Mar 2013 | 4:31:02 UTC - in response to Message 27355.

Now when I wiggle the mouse in BOINC / Tasks, the GPUGRID WU immediately gives "Waiting to run". Perfect!


Not quite perfect... When a WU resumes it backs-off a few percentage points of processing so it's repeating what it's already done.

I guess it's resuming at the last checkpoint. I've looked and looked for the parameter that lets me specify the checkpoint period but I can't find it!

Any ideas?

Tom

I've seen this many times before. I consider it normal behavior.
____________

Post to thread

Message boards : Number crunching : An open letter to the manager of GPUGRID

//