Message boards : Graphics cards (GPUs) : Not gettting work, how to see LTD?
| Author | Message |
|---|---|
|
Upgraded to BOINC 6.6.11 this morning, seems to be working well. However, I'm getting errors when GPUgrid requests work. | |
ID: 7033 | Rating: 0 | rate:
 /   Reply
Quote
| |
 K1atOdessaSend message Joined: 25 Feb 08 Posts: 249 Credit: 370,320,941 RAC: 0 Level  Scientific publications      | |
|
Try BOINC 6.5.0 -- that's what I am running and many other users have reported it works the best for cpu and gpu projects right now. I checked the Server and there are 100+ GPU WU's ready to be sent out. | |
| ID: 7034 | Rating: 0 | rate:
/ Reply
Quote
| |
Dieter Matuschek Send message Joined: 28 Dec 08 Posts: 58 Credit: 231,884,297 RAC: 0 Level  Scientific publications   | |
... is there a way to see the "debt" of various projects? Find file 'client_state.xml' and search therein the tag <long_term_debt>. ____________ | |
| ID: 7035 | Rating: 0 | rate:
/ Reply
Quote
| |
|
[B^S] Gamma^Ray Send message Joined: 24 Aug 07 Posts: 7 Credit: 1,354,047 RAC: 0 Level  Scientific publications | |
|
Well Im having the same issue as of lately too. Keep getting the same message listed by Jeremy. Was running 6.6.11 64 bit so dropped down to 6.4.6 64 bit to see if that would help but it didnt. Even tried detaching and reattaching but still getting the same message. As of right now there are 233 work units ready to be sent and yet I havent run one (Or gotten one) in the past day or so. | |
| ID: 7039 | Rating: 0 | rate:
/ Reply
Quote
| |
|
Michael Doerner Send message Joined: 28 Feb 09 Posts: 37 Credit: 666,889 RAC: 0 Level  Scientific publications | |
|
Hello, | |
| ID: 7040 | Rating: 0 | rate:
/ Reply
Quote
| |
|
Michael Doerner Send message Joined: 28 Feb 09 Posts: 37 Credit: 666,889 RAC: 0 Level Scientific publications | |
|
OK, got it....Upgraded to the true 6.6.11 client (not the 6.6.11 compat version, which I pulled down somehow.) Anyhoo, I have WU's to work with now. L8r. | |
| ID: 7041 | Rating: 0 | rate:
/ Reply
Quote
| |
|
Paul D. Buck Send message Joined: 9 Jun 08 Posts: 1050 Credit: 37,321,185 RAC: 0 Level  Scientific publications | |
|
With 6.5.0 and later (maybe some version earlier) you can look at the internals of a project by clicking on the projects tab, then the project name, then the "Properties" button. That gives a pop-up window with most of the data. | |
| ID: 7042 | Rating: 0 | rate:
/ Reply
Quote
| |
|
MarkJ  Volunteer moderator Volunteer tester Send message Joined: 24 Dec 08 Posts: 738 Credit: 200,909,904 RAC: 0 Level Scientific publications | |
With 6.5.0 and later (maybe some version earlier) you can look at the internals of a project by clicking on the projects tab, then the project name, then the "Properties" button. That gives a pop-up window with most of the data. BOINCview can also display STD (short term debt) and LTD (long term debt) on the projects tab (yes even for 6.6.11 clients). They could do something similar in BOINCmgr. In BOINCmgr 6.6.11 in Advanced view you can go to the Projects tab, click on a project and click the Properties button. It displays the scheduling priority, work fetch priority, the DCF and some other stats. STD seems to = scheduling priority. LTD seems to = work fetch priority. ____________ BOINC blog | |
| ID: 7045 | Rating: 0 | rate:
/ Reply
Quote
| |
|
Installed 6.5.0 today. All SEEMED to be going well until I checked the message log just now. | |
| ID: 7061 | Rating: 0 | rate:
/ Reply
Quote
| |
|
[B^S] Gamma^Ray Send message Joined: 24 Aug 07 Posts: 7 Credit: 1,354,047 RAC: 0 Level Scientific publications | |
|
I'm at a loss whats causing the error or message, I've dropped down to a earlier version of the client, back up to the newest, detached-reattached, even deleted boinc completely and reinstalled fresh and yet just get this message... | |
| ID: 7062 | Rating: 0 | rate:
/ Reply
Quote
| |
|
a fast look at the server status page tells theres no work avilable for nvidia gpu | |
| ID: 7067 | Rating: 0 | rate:
/ Reply
Quote
| |
|
Paul D. Buck Send message Joined: 9 Jun 08 Posts: 1050 Credit: 37,321,185 RAC: 0 Level Scientific publications | |
|
@Jeremy | |
| ID: 7071 | Rating: 0 | rate:
/ Reply
Quote
| |
|
This not answering my question why I not getting more then one new wu when requesting work for 8 days buffer. | |
| ID: 7078 | Rating: 0 | rate:
/ Reply
Quote
| |
|
ExtraTerrestrial Apes Volunteer moderator Volunteer tester  Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level  Scientific publications | |
This not answering my question why I not getting more then one new wu when requesting work for 8 days buffer. You've got a dual core CPU and the server tells you that there is a limit for the number of concurrent WUs, 1 per CPU (core). You've reached that with 2 WUs, so you've got to send one result back to get more WUs. This behaviour is not ideal, but it's the setting the project team choose. MrS ____________ Scanning for our furry friends since Jan 2002 | |
| ID: 7082 | Rating: 0 | rate:
/ Reply
Quote
| |
|
so with the ongoing errors with the boinc client i have to pause 2 other projects wich are running on pure cpu-core everytime i send request to fetch new work for gpugrid, this is not a ideal situation for me with CUDA running on at most 0,2 CPU there would be no problem to fetch work to fill for at least 5 days workbuffer | |
| ID: 7083 | Rating: 0 | rate:
/ Reply
Quote
| |
|
ExtraTerrestrial Apes Volunteer moderator Volunteer tester Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level Scientific publications | |
|
Not sure what you mean, but it sounds like you got something wrong. | |
| ID: 7085 | Rating: 0 | rate:
/ Reply
Quote
| |
|
Scott Brown Send message Joined: 21 Oct 08 Posts: 144 Credit: 2,973,555 RAC: 0 Level Scientific publications | |
|
This is the key line for understanding why you are not getting work
The number of tasks is tied to the number of CPU's (or more specifically number of cores), which in your case is two. Unlike CPU-based projects, increasing cache, increasing resource share, etc. will have no effect on number of workunits received at GPUGRID. This is set at the server. Thus, if you are crunching one task with another waiting, when you complete the first task you will only be able to get a single replacement. If you were to reset the project, you would download two workunits. Hopefully, this will change in the future to an allocation of work based on number and speed of GPU's, but my guess is that we are some ways off (at least a few months) from this being a stable reality. | |
| ID: 7087 | Rating: 0 | rate:
/ Reply
Quote
| |
|
Paul D. Buck Send message Joined: 9 Jun 08 Posts: 1050 Credit: 37,321,185 RAC: 0 Level Scientific publications | |
so with the ongoing errors with the boinc client i have to pause 2 other projects wich are running on pure cpu-core everytime i send request to fetch new work for gpugrid, this is not a ideal situation for me with CUDA running on at most 0,2 CPU there would be no problem to fetch work to fill for at least 5 days workbuffer Now I have the feeling that there are actually two questions. The first has been answered by Scott and ETA that you will only have a limited number of tasks on hand regardless of the settings of BOINC on your client. On my 8 CPU system I have 4 in flight and 4 pending most of the time. On occasion, I will turn in a couple and be down to one or two pending until I get another request to fill my queue. Again, this is a server side setting that is not likely to change (as previously stated) anytime soon. And yes, the project is well aware that we the people are unhappy with it ... :) BUt, the task you complete is used to generate another task, that when complete is used to generate another task ... on and on ... The second question it seems to me is that you are having to "fiddle" with the system to even keep this happy situation. This *IS* an artifact of the BOINC Manager and there are a couple things you can try. In your case with a dual CPU system you need to set the resource share of GPU Grid so that it is above 50% ... this will make BOINC Manager think that GPU Grid is "starved" and, with version 6.5.0 will tend to automatically fetch work for you. Since the BOINC Manager works on a proportional basis, all you have to do is make suer that the shares for your other projects are set accordingly ... if you want 50/50 on two other projects, then set a share for them to be 25% each ... and so on ... I use a share to set my i7 so that it is 20% of my resources ... This unhappy situation is, or should be, fixed in the upcoming 6.7.x release of BOINC manager where the CUDA resources should be "properly" managed separately from the CPU Projects. | |
| ID: 7091 | Rating: 0 | rate:
/ Reply
Quote
| |
|
[B^S] Gamma^Ray Send message Joined: 24 Aug 07 Posts: 7 Credit: 1,354,047 RAC: 0 Level Scientific publications | |
|
Well, I finially tried removing the Cuda installation I had and reinstalled it from scratch, once done that seem to fix my error as I picked up a work unit right after. :) | |
| ID: 7094 | Rating: 0 | rate:
/ Reply
Quote
| |
Message boards : Graphics cards (GPUs) : Not gettting work, how to see LTD?
