PrimeGrid
Please visit donation page to help the project cover running costs for this month

Toggle Menu

Join PrimeGrid

Returning Participants

Community

Leader Boards

Results

Other

drummers-lowrise
1) Message boards : Sieving : ppsieve/tpsieve CUDA testing (Message 29328)
Posted 3602 days ago by aaronhaviland
Ken_g6,

Latest sources seem to have eliminated the problem on my GTS 250 as well.

Do you still think there is a bug in the newer NVidia drivers, or was it the tpsieve code?
2) Message boards : Proth Prime Search : Computation Errors - XML Error (Message 28654)
Posted 3616 days ago by aaronhaviland
How do you "lock" a version? The only method I know (apt package pinning, read the apt_preferences manpage) is confusing at best, for most people. Half the time I feel I've only gotten it to work out of pure luck.
3) Message boards : Proth Prime Search : Computation Errors - XML Error (Message 28636)
Posted 3616 days ago by aaronhaviland
sceptre0, correct me if I'm wrong here. Are you using the 64-bit version of Ubuntu Maverick? You'll have to work outside the restricted drivers manager to deal with this.

If you are, you should be able to install this version of nvidia drivers and be on your merry way:

http://launchpadlibrarian.net/56414517/nvidia-current_256.53-0ubuntu3_amd64.deb

Just keep in mind that every time you run the Update Manager, you'll have to deselect nvidia so it doesn't re-upgrade them (which it will try to do, because 260 is newer than 256)
4) Message boards : Proth Prime Search : Computation Errors - XML Error (Message 28606)
Posted 3617 days ago by aaronhaviland
Indeed the validator does not like the ampersand in that line (should probably be escaped), but that's not the error he's getting from BOINC.

The error is from the line "Computation Error: no candidates found for p=xxxxx." which, given that it's a GPU task, matches the problem in the other thread.

scepter0, your friend's computer on the other hand, was getting a different error, on the tasks you linked to (the link had &offset=5400, so he's obviously errored out a *lot* of tasks): "Error: API mismatch: the NVIDIA kernel module has version 256.53, but this NVIDIA driver component has version 260.19.21." because he didn't reboot after upgrading the NVIDIA driver.

However, His more recent tasks are getting the same error as you: "Computation Error: no candidates found for p=xxxxx"
5) Message boards : Proth Prime Search : Computation Errors - XML Error (Message 28598)
Posted 3617 days ago by aaronhaviland
Hi sceptre0! no wonder I've started getting more credit than you (Been watching the Ubuntu team's rankings the last couple days)

See this thread: http://www.primegrid.com/forum_thread.php?id=2839

Short version, right now, it seems either you'll have to downgrade away from the 260 driver series, or specify a cmdline parameter in your app_info.xml... but you're not using an app_info.xml

Also, FYI,
For the nvidia driver version,

apt-cache policy nvidia-current

should tell you something like:

nvidia-current: Installed: 260.19.06-0ubuntu1 Candidate: 260.19.06-0ubuntu1 Version table: *** 260.19.06-0ubuntu1 0 500 http://ubuntu.media.mit.edu/ubuntu/ maverick/restricted i386 Packages 100 /var/lib/dpkg/status

That's your driver version. Or you could launch nvidia-settings and see that information also.
6) Message boards : Proth Prime Search : gpu task problem (Message 28570)
Posted 3617 days ago by aaronhaviland
http://www.nvnews.net/vbulletin/showthread.php?t=157563 New nvidia BETA linux driver released today. It's lacking an updated changelog, so no way to know what's different/new.

I just tried with this new driver, and continue to have the same problem; just in case anyone else is inclined to test the new version, it won't help this.

They did however fix being able to change the clockspeeds through nvidia-settings, though, so some people using the 260 driver might find it useful.
7) Message boards : Proth Prime Search : gpu task problem (Message 28487)
Posted 3619 days ago by aaronhaviland
As for -m 2 and a GTS 250, though, that's just way smaller than it should have to be.


I've tried all values between 2 and 8. Most bail-out within 30 seconds, some sooner. Should I try higher values?

Also, I've rebuilt the version in your git repo against CUDA 3.2, and get the exact same behaviour. (and roughly the same performance), so CUDA version doesn't seem to matter.
8) Message boards : Proth Prime Search : gpu task problem (Message 28478)
Posted 3619 days ago by aaronhaviland
it looks like adding "-m 8" to the command line may make the results better for Fermi users.


Likewise, it would seem that "-m 2" allows my GTS 250 to run these new WUs without errors.
[Return to PrimeGrid main page]
DNS Powered by DNSEXIT.COM
Copyright © 2005 - 2020 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 0.45, 0.50, 0.67
Generated 25 Oct 2020 | 22:51:58 UTC