Index  | Recent Threads  | Unanswered Threads  | Who's Active  | Guidelines  | Search
 

Quick Go »
No member browsing this thread
Thread Status: Active
Total posts in this thread: 822
Posts: 822   Pages: 83   [ Previous Page | 22 23 24 25 26 27 28 29 30 31 | Next Page ]
[ Jump to Last Post ]
Post new Thread
Author
Previous Thread This topic has been viewed 958018 times and has 821 replies Next Thread
Ian-n-Steve C.
Senior Cruncher
United States
Joined: May 15, 2020
Post Count: 180
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Work unit availability

Unless perchance there is a newer driver for it?
Meow?

Sorry kittyman, I think this is not an issue of driver support. The Quadro 2000M is a Fermi generation card, which appears to only supports openCL 1.1. Newer generations do support 1.2 as shown by the driver release notes.
driver release notes 377.83


I don't think OpenCL 1.2 is actually required. I just ran several workunits on an openCL 1.1 card (550ti, also Fermi) with 340 and 390 drivers. they all ran fine and validated.

I posted about it in the GPU support forum: https://www.worldcommunitygrid.org/forums/wcg...ad,43352_offset,30#655567
----------------------------------------

EPYC 7V12 / [5] RTX A4000
EPYC 7B12 / [5] RTX 3080Ti + [2] RTX 2080Ti
EPYC 7B12 / [6] RTX 3070Ti + [2] RTX 3060
[2] EPYC 7642 / [2] RTX 2080Ti
----------------------------------------
[Edit 1 times, last edit by Ian-n-Steve C. at Apr 14, 2021 1:35:54 PM]
[Apr 14, 2021 1:34:37 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Grumpy Swede
Master Cruncher
Svíþjóð
Joined: Apr 10, 2020
Post Count: 2195
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Work unit availability

It appears that there could have been a grouping of target/ligands (jobs) that were not viable and thus not good drug candidates. This failed validation, but could still be a valid scientific result. I say this seems like a grouping, because I can see them in the database...previously it was 1 in a batch randomly, which could statistically be unlucky. But, this seems like a different problem and will require more review. You can see the groupings by batch, E and R are basically the same, R stands for rerun, but that means it was marked as error atleast one attempt by a group of members.

Thanks,
-Uplinger

@Uplinger:

Three more of these odd cases, of "failed validation, but could still be a valid scientific result":

https://www.worldcommunitygrid.org/ms/device/...s.do?workunitId=619525017
https://www.worldcommunitygrid.org/ms/device/...s.do?workunitId=619524798
https://www.worldcommunitygrid.org/ms/device/...s.do?workunitId=619524870

Edit: Added one more:
https://www.worldcommunitygrid.org/ms/device/...s.do?workunitId=619524496
----------------------------------------
[Edit 1 times, last edit by Grumpy Swede at Apr 14, 2021 2:06:11 PM]
[Apr 14, 2021 1:36:51 PM]   Link   Report threatening or abusive post: please login first  Go to top 
hnapel
Advanced Cruncher
Netherlands
Joined: Nov 17, 2004
Post Count: 82
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Work unit availability

Wow Grumpy Swede!
What's your secret?
I'm getting between 0 and 2 a day across five 24/7 machines that are various Nvidia, AMD, and Intel iGPU. In general, I noticed that the computers only get GPU tasks during the day.


I don't know of any secrets and am also not gaming the system with scripts, but I get bursts of OPNG jobs now and then, which really helps to gain points (not time as mentioned before).

One thing I like to mention is I configured BOINC with the setting 'save at least x days' amount of work (in my case 0.1 or 0.2), so what happens is if GPU jobs become available, it stores some as backlog which of course then drain quickly but still add a lot of points for me.
[Apr 14, 2021 1:51:52 PM]   Link   Report threatening or abusive post: please login first  Go to top 
hnapel
Advanced Cruncher
Netherlands
Joined: Nov 17, 2004
Post Count: 82
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Work unit availability

Good evening,

I made a mistake on my update of a config file and it caused a bunch of work to be loaded onto the grid. I'm in clean up mode of that issue. It did send a bunch of results out and I'm happy that people got lots, but this was a mistake on my end and I'll make sure the work is distributed evenly going forward.

Thanks for your understanding,
-Uplinger


Mmmh might account for the fact I racked up 2M points instead of (on average) below 1M on the 13th...
[Apr 14, 2021 2:05:04 PM]   Link   Report threatening or abusive post: please login first  Go to top 
goben_2003
Advanced Cruncher
Joined: Jun 16, 2006
Post Count: 145
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Work unit availability

Unless perchance there is a newer driver for it?
Meow?

Sorry kittyman, I think this is not an issue of driver support. The Quadro 2000M is a Fermi generation card, which appears to only supports openCL 1.1. Newer generations do support 1.2 as shown by the driver release notes.
driver release notes 377.83


I don't think OpenCL 1.2 is actually required. I just ran several workunits on an openCL 1.1 card (550ti, also Fermi) with 340 and 390 drivers. they all ran fine and validated.

I posted about it in the GPU support forum: https://www.worldcommunitygrid.org/forums/wcg...ad,43352_offset,30#655567

That is interesting. Does boinc recognize it as as opencl 1.2? Example from when boinc starts up:
OpenCL: NVIDIA GPU 0: GeForce RTX 2060 (driver version 456.71, device version OpenCL 1.2 CUDA, 6144MB, 3550MB available, 5126 GFLOPS peak)

Last time it was mentioned by Uplinger he said opencl 1.2 was required:
What specs do we need to run the units? I didn't get any Beta units.


The minimum requirement is that your computer has a GPU that can run OpenCL 1.2. However, there are some cards that still have issues even though they claim to be able to handle them.

Also, please check your device profile settings to make sure that you allow GPU work on the graphics cards your machines support. You can check the top of your event log to see what graphics cards BOINC recognizes on your machine.

----------------------------------------

[Apr 14, 2021 2:07:08 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Grumpy Swede
Master Cruncher
Svíþjóð
Joined: Apr 10, 2020
Post Count: 2195
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Work unit availability

And just got this one, which probably will fail to validate too in a couple of minutes. Already two invalids for that WU.

https://www.worldcommunitygrid.org/ms/device/...s.do?workunitId=619591423

Edit: I think I stop processing GPU WU's, until that issue is fixed. Just a waste of time, and resources.
----------------------------------------
[Edit 2 times, last edit by Grumpy Swede at Apr 14, 2021 2:19:28 PM]
[Apr 14, 2021 2:17:13 PM]   Link   Report threatening or abusive post: please login first  Go to top 
pututu
Senior Cruncher
United States
Joined: Jan 3, 2016
Post Count: 241
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Work unit availability

I'm also seeing a dozen invalids since early this morning.
https://www.worldcommunitygrid.org/ms/device/...s.do?workunitId=619591608

https://www.worldcommunitygrid.org/ms/device/...s.do?workunitId=619590979

https://www.worldcommunitygrid.org/ms/device/...s.do?workunitId=619563310

https://www.worldcommunitygrid.org/ms/device/...s.do?workunitId=619524387

I've both nvidia and amd cards. Only nvidia has invalid while amd is still running without any invalid. Perhaps start a new thread in gpu support forum?
----------------------------------------

----------------------------------------
[Edit 2 times, last edit by pututu at Apr 14, 2021 2:28:37 PM]
[Apr 14, 2021 2:21:41 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Grumpy Swede
Master Cruncher
Svíþjóð
Joined: Apr 10, 2020
Post Count: 2195
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Work unit availability

I'm also seeing a dozen invalids since early this morning.
https://www.worldcommunitygrid.org/ms/device/...s.do?workunitId=619591608

I've both nvidia and amd cards. Only nvidia has invalid while amd is still running without any invalid. Perhaps start a new thread in gpu support forum?

Uplinger is already working on a fix to this. Check here:
https://www.worldcommunitygrid.org/forums/wcg/viewpostinthread?post=655534
[Apr 14, 2021 2:23:49 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Ian-n-Steve C.
Senior Cruncher
United States
Joined: May 15, 2020
Post Count: 180
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Work unit availability


That is interesting. Does boinc recognize it as as opencl 1.2? Example from when boinc starts up:
OpenCL: NVIDIA GPU 0: GeForce RTX 2060 (driver version 456.71, device version OpenCL 1.2 CUDA, 6144MB, 3550MB available, 5126 GFLOPS peak)

Last time it was mentioned by Uplinger he said opencl 1.2 was required:
What specs do we need to run the units? I didn't get any Beta units.


The minimum requirement is that your computer has a GPU that can run OpenCL 1.2. However, there are some cards that still have issues even though they claim to be able to handle them.

Also, please check your device profile settings to make sure that you allow GPU work on the graphics cards your machines support. You can check the top of your event log to see what graphics cards BOINC recognizes on your machine.


under normal circumstances, BOINC sees it as a OpenCL 1.1 device. though when you understand how BOINC operates and where it gets its information from and the fact that the project gets the card info from BOINC and not directly from your GPU, it's easy to manipulate cool . So I made it look like a 1.2 device for the purposes of getting work sent to me. you could probably get the same effect by installing a newer GPU to get tasks, then shut down the system and install the older card with the older drivers before restarting BOINC. once it's on the system, BOINC will allow it to run on whatever device is compatible, and on the BOINC side it will let an OpenCL 1.1 device run these.

I'm not suggesting anyone realistically try doing either of these methods (and I only tried on Linux). I just wanted to know if the OpenCL 1.2 limit was real or artifical. seems to be the latter, and the project is just gatekeeping the tasks from 1.1 devices.
----------------------------------------

EPYC 7V12 / [5] RTX A4000
EPYC 7B12 / [5] RTX 3080Ti + [2] RTX 2080Ti
EPYC 7B12 / [6] RTX 3070Ti + [2] RTX 3060
[2] EPYC 7642 / [2] RTX 2080Ti
[Apr 14, 2021 2:36:05 PM]   Link   Report threatening or abusive post: please login first  Go to top 
goben_2003
Advanced Cruncher
Joined: Jun 16, 2006
Post Count: 145
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Work unit availability


That is interesting. Does boinc recognize it as as opencl 1.2? Example from when boinc starts up:
OpenCL: NVIDIA GPU 0: GeForce RTX 2060 (driver version 456.71, device version OpenCL 1.2 CUDA, 6144MB, 3550MB available, 5126 GFLOPS peak)

Last time it was mentioned by Uplinger he said opencl 1.2 was required:
What specs do we need to run the units? I didn't get any Beta units.


The minimum requirement is that your computer has a GPU that can run OpenCL 1.2. However, there are some cards that still have issues even though they claim to be able to handle them.

Also, please check your device profile settings to make sure that you allow GPU work on the graphics cards your machines support. You can check the top of your event log to see what graphics cards BOINC recognizes on your machine.


under normal circumstances, BOINC sees it as a OpenCL 1.1 device. though when you understand how BOINC operates and where it gets its information from and the fact that the project gets the card info from BOINC and not directly from your GPU, it's easy to manipulate cool . So I made it look like a 1.2 device for the purposes of getting work sent to me. you could probably get the same effect by installing a newer GPU to get tasks, then shut down the system and install the older card with the older drivers before restarting BOINC. once it's on the system, BOINC will allow it to run on whatever device is compatible, and on the BOINC side it will let an OpenCL 1.1 device run these.

I'm not suggesting anyone realistically try doing either of these methods (and I only tried on Linux). I just wanted to know if the OpenCL 1.2 limit was real or artifical. seems to be the latter, and the project is just gatekeeping the tasks from 1.1 devices.

IIRC, there were a lot of errors from OpenCL 1.1 capable graphics cards that had a driver that reported OpenCL 1.2 during the betas.
----------------------------------------

[Apr 14, 2021 3:59:54 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Posts: 822   Pages: 83   [ Previous Page | 22 23 24 25 26 27 28 29 30 31 | Next Page ]
[ Jump to Last Post ]
Post new Thread