Bug in server affecting older BOINC clients with NVIDIA GPUs.


log in

Advanced search

Message boards : News : Bug in server affecting older BOINC clients with NVIDIA GPUs.

Previous · 1 · 2 · 3 · 4 · 5 · 6 · 7 · Next
Author Message
robert spry
Send message
Joined: 24 May 99
Posts: 9
Credit: 6,000,306
RAC: 5,062
United States
Message 1274838 - Posted: 24 Aug 2012, 11:42:35 UTC - in response to Message 1274461.

[quote]I'm looking for this in the logs now. This is probably an artifact of the way are BOINC 6 OpenCL kludge is implemented...

[quote]

Can you suggest how I can fix this?
I've tried reset project. Downloads of the non WUs from SETI work fine.
I have no idea what BOINC 6 OpenCL kludge means. I've been running this for years with no hardware/software changes that I know of.
At this point I just can't seem to receive any WUs from SETI (but do from all the other projects.
Thank You.

____________

ClaggyProject donor
Volunteer tester
Send message
Joined: 5 Jul 99
Posts: 4249
Credit: 34,985,617
RAC: 21,235
United Kingdom
Message 1274860 - Posted: 24 Aug 2012, 12:43:45 UTC - in response to Message 1274838.
Last modified: 24 Aug 2012, 12:52:20 UTC

I'm looking for this in the logs now. This is probably an artifact of the way are BOINC 6 OpenCL kludge is implemented...



Can you suggest how I can fix this?
I've tried reset project. Downloads of the non WUs from SETI work fine.
I have no idea what BOINC 6 OpenCL kludge means. I've been running this for years with no hardware/software changes that I know of.
At this point I just can't seem to receive any WUs from SETI (but do from all the other projects.
Thank You.

See BilBg's post earlier, you need to try installing newer drivers, But Not any newer than Cat12.1 (ATI OpenCL Support is removed from Cat12.2 and later on XP drivers)

Then use GPU-Z or GPUCapsViewer to see if you have OpenCL support.

Only Boinc 7 and later can detect wether a GPU has OpenCL support, The Boinc 6 Kludge is that if you assume a GPU has a particular driver, then OpenCL Support should be there, But there is always a chance some dimwit only installed the driver component, and didn't install the OpenCL runtime (They were available at one point as separate downloads)

The Server Bug is that because Boinc 7 didn't detect OpenCL support (and so wouldn't send opencl_ati_100 Wu's), it also shouldn't use the Boinc 6 Kludge and send ati_opencl_100 Wu's.

Claggy

robert spry
Send message
Joined: 24 May 99
Posts: 9
Credit: 6,000,306
RAC: 5,062
United States
Message 1274883 - Posted: 24 Aug 2012, 15:15:15 UTC - in response to Message 1274860.

Took the Suggestion.
I have re-installed Driver CAT12.1 (Thank You, BilBg)
I ran GPU-Z: GPU-Z Validation k9rpp
I can only assume I do not have OpenCL support.

How do I fix this?
should I Uncheck "Use ATI GPU"?
____________

Horacio
Send message
Joined: 14 Jan 00
Posts: 536
Credit: 75,956,876
RAC: 3,967
Argentina
Message 1274950 - Posted: 24 Aug 2012, 17:45:35 UTC - in response to Message 1274883.

Took the Suggestion.
I have re-installed Driver CAT12.1 (Thank You, BilBg)
I ran GPU-Z: GPU-Z Validation k9rpp
I can only assume I do not have OpenCL support.

How do I fix this?
should I Uncheck "Use ATI GPU"?

If you dont have OpenCl support then yes, you need to uncheck the "Use ATI GPU" but if you have more than one computer you may want to set a different venue for this host so this setting wont keep you from crunching in other GPUs that are OpenCl capable.
____________

taurus
Send message
Joined: 6 Feb 02
Posts: 1
Credit: 1,741,582
RAC: 221
Germany
Message 1274973 - Posted: 24 Aug 2012, 19:08:48 UTC

In new ATI Catalyst Driver is no open CL driver included.
Use Catalyst-Version 11.3 or below.
____________

Sean Farrell
Send message
Joined: 17 Jun 12
Posts: 2
Credit: 17,251
RAC: 0
Canada
Message 1275009 - Posted: 24 Aug 2012, 20:30:56 UTC

Now that I've read the posts on this topic I'm inclined to believe this BUG is the reason my Laptop's onboard video chip no longer works. :-(
Lenovo T61 7664-1ju <--- R.I.P.

ClaggyProject donor
Volunteer tester
Send message
Joined: 5 Jul 99
Posts: 4249
Credit: 34,985,617
RAC: 21,235
United Kingdom
Message 1275017 - Posted: 24 Aug 2012, 20:47:42 UTC - in response to Message 1275009.

Now that I've read the posts on this topic I'm inclined to believe this BUG is the reason my Laptop's onboard video chip no longer works. :-(
Lenovo T61 7664-1ju <--- R.I.P.

Rubbish, Boinc doesn't detect a GPU on any of your hosts, and they have never received any GPU Wu's:

Computers belonging to Sean Farrell

Application details for host 6706265

Application details for host 6714919

Application details for host 6705145

Claggy

OzzFan
Volunteer tester
Avatar
Send message
Joined: 9 Apr 02
Posts: 13706
Credit: 31,749,963
RAC: 13,387
United States
Message 1275018 - Posted: 24 Aug 2012, 20:49:15 UTC - in response to Message 1275009.

Now that I've read the posts on this topic I'm inclined to believe this BUG is the reason my Laptop's onboard video chip no longer works. :-(
Lenovo T61 7664-1ju <--- R.I.P.


Highly unlikely. If your old T61 didn't have an OpenCL compatible driver, likely you were never sent any GPU work. Even if you were sent GPU work by accident, it would have never processed on the GPU and it would have simply erred out.

Your video chip failing had to be for other reasons.

robert spry
Send message
Joined: 24 May 99
Posts: 9
Credit: 6,000,306
RAC: 5,062
United States
Message 1275106 - Posted: 25 Aug 2012, 2:24:15 UTC - in response to Message 1274883.

Fixed by unchecking.
Thank you all
Bob

____________

Profile Interstel
Avatar
Send message
Joined: 29 Nov 01
Posts: 23
Credit: 2,222,625
RAC: 0
United States
Message 1275345 - Posted: 25 Aug 2012, 12:26:05 UTC - in response to Message 1274309.

I find that the dates and times do not match the ones listed in the link.

Normal:
- Your time (file times) is not in UTC
- You finish their download later than server sends (assigns) them to you

OK got it.

But I do find the following 2 entries are not present in any kind of wording....
6.10 (cuda_fermi)

You have NVIDIA GeForce GTS 250 which is not fermi, you have tasks marked "SETI@home Enhanced v6.09 (cuda23)" and "SETI@home Enhanced v6.08 (cuda)":
http://setiathome.berkeley.edu/results.php?hostid=3691074&offset=0&show_names=0&state=2&appid=

They are processed by:
setiathome_6.08_windows_intelx86__cuda.exe
setiathome_6.09_windows_intelx86__cuda23.exe

(and several .dll files)

OK and I got that too now in my head.


6.04 (opencl_nvidia_100)

You still do not have AstroPulse tasks allocated to the GPU, the app will be downloaded with the first AstroPulse task for the GPU:
http://setiathome.berkeley.edu/results.php?hostid=3691074&offset=0&show_names=0&state=0&appid=12

The first ones just downloaded. So apparently I have never been using the GPU for Astropulse? Only for 6.09 cuda WU's?


And I guess I had some stuff not used anymore like astropulse 5.00, 5.03.

Since all the apps (.exe .dll .cl and other types of files) consume very little HDD space - better do not delete things you are not sure what they are (e.g. .DLLs)


And at least twice applications that seemed to have 2 slightly different versions such as

setiathome_6.03_windows_intelx86.exe
setigraphics_6.03_windows_intelx86.exe

Do not delete them - both are standard current apps (setigraphics used for "Show graphics" and screensaver)



Everything there will be left alone then and thank you for all the information it has been permanently filed in the old aging braincase. The upside is that in the next few days I will be getting in a GeForce 295GTX dual GPU 1.7GB memory card. I would get something better but that is the top end card supported by my motherboard.



Final question concerning the operation of BOINC itself. In the last 17 days I have added 6 projects to my primary workstation but left the other machines just running SETI as I figured I had enough machine power to run them with damaging my SETI WU credit score. In fact I tweaked the settings for Resource share so that SETI had 70% of the resources and the other 6 projects had 5% each. But I am finding out that BOINC seems to be completely ignoring that setting and instead allocating time round robin between them all and result is that my average credit for SETI has dropped from nearly 4500 down to 2000 and the Einstein one has jumped to 3900, 3 of the others are together under 1000 and 2 are not generating any work. Was I wrong in believing I had read that you could force BOINC through the resource share option to partition out the work in the percentage that you wanted?


James
____________

Joined SETI@Home in 2001
Online since ArpNET days
First activity on Honeywell 1648
Series Mainframe in 1975 at age 12.

Profile BilBg
Volunteer tester
Avatar
Send message
Joined: 27 May 07
Posts: 2894
Credit: 6,636,821
RAC: 7,881
Bulgaria
Message 1275359 - Posted: 25 Aug 2012, 13:17:59 UTC - in response to Message 1275106.

Fixed by unchecking.

This is not a real fix (to uncheck 'Use ATI GPU')

Your 'AMD ATI Radeon HD 4600 series' is listed as (hardware) supporting OpenCL:
http://developer.amd.com/sdks/AMDAPPSDK/pages/DriverCompatibility.aspx

Listed devices are:
ATI Radeon™ HD - 4600 Series (4670, 4650)

Which is your model?
Can you give GPU-Z screenshot? (click 'Photo-icon' in the top-right of GPU-Z, upload to ..., copy link)

Here is shown 'OpenCL version - OpenCL 1.0' (apparently valid for all 'Radeon R700' chips):
http://en.wikipedia.org/wiki/Radeon_R700


____________



- ALF - "Find out what you don't do well ..... then don't do it!" :)

Profile Sunny129
Avatar
Send message
Joined: 7 Nov 00
Posts: 190
Credit: 3,163,755
RAC: 0
United States
Message 1275374 - Posted: 25 Aug 2012, 13:42:22 UTC - in response to Message 1275345.

Was I wrong in believing I had read that you could force BOINC through the resource share option to partition out the work in the percentage that you wanted?


James

well, that's the idea...but when was the last time BOINC actually did something you wanted it to? that function doesn't work correctly for me either. two of my machines crunch LHC@Home Classic along with a handful of other projects. i used to allocate 90% of the CPU to LHC@H and the remaining 10% to the other projects b/c LHC@H hardly ever sees new work...so i wanted to be sure that my machines would give priority to LHC@H work whenever it is available. so one would think that BOINC would prioritize LHC@H work over all other work whenever its available, but does it? nope. instead, LHC@H doesn't even download new work on the rare occasion that its available, and all the projects that appear to have a minimal CPU resource share continue to grab more work...

...interestingly enough, one day i decided to reallocate the CPU resources such that all projects have an even share...and what do you know, all of the sudden LHC@H starts getting work when its available...go figure.
____________

Profile BilBg
Volunteer tester
Avatar
Send message
Joined: 27 May 07
Posts: 2894
Credit: 6,636,821
RAC: 7,881
Bulgaria
Message 1275383 - Posted: 25 Aug 2012, 13:56:36 UTC - in response to Message 1275345.

6.04 (opencl_nvidia_100)

You still do not have AstroPulse tasks allocated to the GPU, the app will be downloaded with the first AstroPulse task for the GPU:
http://setiathome.berkeley.edu/results.php?hostid=3691074&offset=0&show_names=0&state=0&appid=12

The first ones just downloaded. So apparently I have never been using the GPU for Astropulse? Only for 6.09 cuda WU's?

Yes, of course, the AstroPulse v6 6.04 (opencl_nvidia_100) / Windows app was just released.
(it was available in some 'beta' form only for people/testers using the optimized apps)


... in the next few days I will be getting in a GeForce 295GTX dual GPU 1.7GB memory card.
I would get something better but that is the top end card supported by my motherboard.

Why do you think so?
Don't expect that your motherboard docs (.pdf) will include all the video cards it can accept.
(it will list only those video cards available at the time the motherboard was manufactured)

You can use PCIe 2.0 video card in an PCIe 1.x motherboard.
(my PCIe 1.x motherboard is made 2006, my PCIe 2.0 video card in it is made 2011 (or 2012 ?))


Was I wrong in believing I had read that you could force BOINC through the resource share option to partition out the work in the percentage that you wanted?

BOINC is 'supposed' to respect resource share in a long term (weeks?).


____________



- ALF - "Find out what you don't do well ..... then don't do it!" :)

Profile Keith Jillings
Send message
Joined: 3 Apr 99
Posts: 13
Credit: 1,266,603
RAC: 664
United Kingdom
Message 1275574 - Posted: 26 Aug 2012, 0:11:46 UTC

My PC hasn't downloaded any work units from SETI in the past week. The GPU is Nvidia GTS450 and the drivers are 301.42. BOINC is version 7.0.28 (x64). Is there something I should do?

BOINC is happily beavering away at 4 projects so CPU isn't going to waste. I notice that 3 others - Rosetta, Malariacontrol and Spinhenge also have no WUs. Spinhenge is defunct, but the other two may be suffering from the same problem as SETI.
____________

juan BFBProject donor
Volunteer tester
Avatar
Send message
Joined: 16 Mar 07
Posts: 5489
Credit: 316,485,756
RAC: 134,115
Brazil
Message 1275580 - Posted: 26 Aug 2012, 0:29:30 UTC - in response to Message 1275574.
Last modified: 26 Aug 2012, 0:30:15 UTC

You could try Milkway, at least its always have work to do (GPU+CPU) so you will be sure there is no problem in your host configuration.
____________

Profile Keith Jillings
Send message
Joined: 3 Apr 99
Posts: 13
Credit: 1,266,603
RAC: 664
United Kingdom
Message 1275643 - Posted: 26 Aug 2012, 7:15:00 UTC

Yes, Milkyway I have and it's working like a good 'un. It has 21 tasks in the list at the moment - half complete, half at zero.
____________

Profile Keith Jillings
Send message
Joined: 3 Apr 99
Posts: 13
Credit: 1,266,603
RAC: 664
United Kingdom
Message 1276009 - Posted: 26 Aug 2012, 23:39:50 UTC

I've left it to do its own thing, and there was one attempt to get tasks this morning, with the response "Project has no tasks available". If SETI is out of tasks, that would explain the lack of work being done - but I'm surprised. Has it really run out of work?
____________

rob smithProject donor
Volunteer tester
Send message
Joined: 7 Mar 03
Posts: 8819
Credit: 63,222,170
RAC: 82,501
United Kingdom
Message 1276109 - Posted: 27 Aug 2012, 6:12:06 UTC

Not out of tasks, just no tasks on the download server at the instant your cruncher called for work. Now that one call has been made it should try again fairly soon, and keep going until your cache is full.
____________
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

Profile MikeProject donor
Volunteer tester
Avatar
Send message
Joined: 17 Feb 01
Posts: 25205
Credit: 34,826,002
RAC: 21,243
Germany
Message 1276128 - Posted: 27 Aug 2012, 8:23:01 UTC
Last modified: 27 Aug 2012, 8:23:22 UTC

The feeder can only hold 100 work units at one time.
____________

Grant (SSSF)
Send message
Joined: 19 Aug 99
Posts: 5955
Credit: 62,497,315
RAC: 40,511
Australia
Message 1276130 - Posted: 27 Aug 2012, 8:56:52 UTC - in response to Message 1276128.


One day i'll click on one of these threads, and a find a post that has something to do with the topic & thread.
A crazy dream, i know.
____________
Grant
Darwin NT.

Previous · 1 · 2 · 3 · 4 · 5 · 6 · 7 · Next

Message boards : News : Bug in server affecting older BOINC clients with NVIDIA GPUs.

Copyright © 2014 University of California