CuFFT Plan Failure

Message boards : Number crunching : CuFFT Plan Failure
Message board moderation

To post messages, you must log in.

AuthorMessage
EdMacsc

Send message
Joined: 18 Apr 03
Posts: 17
Credit: 98,900
RAC: 0
United States
Message 1384247 - Posted: 24 Jun 2013, 16:31:50 UTC

On any of the CUDA(Any Number) tasks, they run for 12-31 seconds, then I get the message: Scheduler Wait: CuFFT Plan Failure.

I had a gentlemen help me this past Friday, and everything worked until Sunday evening! I back to the CuFFT Plan Failures!!!!

The person did the following:

Your thread 'Scheduler Wait: CuFFT Plan Failure'
http://setiathome.berkeley.edu/forum_thread.php?id=72078
has been hidden by moderator Zapped Sparky (ID 8935686).
Reason: other
Moderator comment: Since you have a duplicate thread in number crunching this one is being closed.

For assistance with SETI@home go to http://setiathome.berkeley.edu/


Today, when I went to the ".edu/forum_thread.php?id=72708", it states that the file was closed by a moderator.

I have suspended ALL work, instead of aborting them all like I did on Friday, until I get an answer, that will work 24/7.

Thanks so much for your help!
Ed Lawrence
Aberdeen, MD[/quote]
Actually, my thread http://setiathome.berkeley.edu/forum_thread.php?id=71953,
only applies to specific problems with cuda22 and cuda23 tasks, where the v7 app is actually picking up a couple v6 dll's and then failing with ERR_TOO_MANY_EXITS. To get around that specific problem, you need to manually delete the cudart.dll and cudafft.dll files in the S@H project directory. The caveat is that if you reboot your machine, or restart the BOINC Manager, BOINC will notice that those files are missing and download them again, thus requiring you to manually delete them again.

As far as CuFFT Plan Failure errors are concerned, I think you'll have to post again in your own thread to get additional advice on those, as I haven't been experiencing that problem.[/quote]


I deleted the files that you asked me to, and when I went and started all the suspended tasks, I got the same message! I just went back and checked to see if those specific files re-downloaded, and they haven't; since I haven't re-booted, and my tasks were in suspension.

ALL my work is now in Suspension, until someone can help me with this issue... Anyone? :o)
ID: 1384247 · Report as offensive
EdMacsc

Send message
Joined: 18 Apr 03
Posts: 17
Credit: 98,900
RAC: 0
United States
Message 1384271 - Posted: 24 Jun 2013, 17:04:45 UTC - in response to Message 1384247.  

I should remind you all that these are ONLY the SETI@HomeV7 7.00 CUDA(ANY Number) that this happens on! Any or All other tasks run just fine. Should I just "Uncheck" the SETI@home V7, and delete ANOTHER 30 or so tasks.... I deleted 23 this past Friday until a work around was found, then Sunday evening for some reason, I started getting these error messages AGAIN; and it processed at least 23 or 24 of them BEFORE these messages started up again.

I hope someone can help, because banging your head on the desk repeatedly HURTS AFTER AWHILE!!! LOL

Again, Thanks to EVERYONE who has tried to resolve this issue!!!

Ed

ID: 1384271 · Report as offensive
Tom*

Send message
Joined: 12 Aug 11
Posts: 127
Credit: 20,769,223
RAC: 9
United States
Message 1384277 - Posted: 24 Jun 2013, 17:16:10 UTC
Last modified: 24 Jun 2013, 17:24:34 UTC

You might want to check out

http://setiathome.berkeley.edu/forum_thread.php?id=72103&postid=1383956#1383956

256 megs of GPU memory may be too small for either 4.2 or 5.0 MB tasks

3.2 may be your best choice however you will need to download Lunatics

optimized apps to just select this version.

- Cuda 3.2 (min driver 263.06 ), also to be distributed as 'stock' for widest support of Video drivers, and recommended for use on older GPUs prior to Fermi & Kepler, such as 8xxx series, 9xxx series, GTX 200 series.

Make sure to download the 64 bit Lunatics version if you decide to use this fix


Lunatics_Win64_v0.41_setup.exe

http://www.arkayn.us/forum/index.php?action=tpmod;dl=cat3

May need to downgrade nvidia to less than 301 not 294 or 295 to freeup memory
and or try disabling aero if 3.2 by itself is not enough

2. upgrade to a GPU with more than 256 Megs
3. de-select SETIatHOME v7

Your latest error was on MB version 4.2
ID: 1384277 · Report as offensive
EdMacsc

Send message
Joined: 18 Apr 03
Posts: 17
Credit: 98,900
RAC: 0
United States
Message 1384668 - Posted: 25 Jun 2013, 23:20:39 UTC - in response to Message 1384247.  

Well, Well, Well....

Wouldn't you know it...

AFTER dumping everything ONCE again, (Thank God it was ONLY two tasks!)

I re-installed BONIC, and just for the HECK of it, went to the NIVIDA web site, and LOW & BEHOLD, there was an UPDATE just for CUDA, and, damn if after I installed it... EVERYTHING WORKS!!!

So, to ANYONE who has NIVIDA Graphics, and you run into "Scheduler Wait: CUDA task failure", Please go to the web-site, and on the right hand column, it has the update for CUDA!!!

I'll tell you, it's easier than banging your head repeatedly against the desk! LMAO! (BTW... still hurts!) :o)

Thank You to Each and Everyone one of you, who helped me!

Ed
ID: 1384668 · Report as offensive

Message boards : Number crunching : CuFFT Plan Failure


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.