GPU and CPU disagree


log in

Advanced search

Questions and Answers : GPU applications : GPU and CPU disagree

Author Message
Profile David Anderson (not *that* DA)Project donor
Avatar
Send message
Joined: 5 Dec 09
Posts: 108
Credit: 22,950,801
RAC: 5,983
United States
Message 1086841 - Posted: 13 Mar 2011, 21:24:13 UTC

My cpu vs other folks GPUs.

I don't say there is any error here, but what does it mean?
WorkUnit 709859714. A couple GPUs agreed there were 31 pulses
and my CPU says there were 5. The spike count
disagreed too. Maybe an indication of
overheading CPU here?
Or other hardware issue on my computer (5766757) ? Should I be worried?

I've been watching for invalid or error results from my
machines and almost all work tasks are 'validated'. I did have
a brief flurry of Signal 11 (5 tasks?) in February I think.
That machine has 12 real cores (not the 24 Linux reports,
the hyperthreading flag set in the BIOS causes Linux to double
the cpu count. I restrict boinc to the 12 real cores).
Thanks in advance for any advice.
____________

Profile perryjay
Volunteer tester
Avatar
Send message
Joined: 20 Aug 02
Posts: 3377
Credit: 15,488,084
RAC: 11,271
United States
Message 1086852 - Posted: 13 Mar 2011, 22:53:50 UTC - in response to Message 1086841.

You've just experienced the joy of having two GPUs running the old V12 app validate themselves. You were probably right but they matched. One is running a 460 the other is running a 470. The old V12 app throws out nothing but -9s and as, in this case, they can validate each other. Not much you can do about it unless you want to send them both a Private Message asking them to update their optimized apps. I've sent out around a dozen and I've only got back one reply so good luck with that.
____________


PROUD MEMBER OF Team Starfire World BOINC

Profile Lint trapProject donor
Send message
Joined: 30 May 03
Posts: 859
Credit: 26,554,216
RAC: 17,760
United States
Message 1086856 - Posted: 13 Mar 2011, 23:16:54 UTC



Well, at least neither offending pc is owned by "Anonymous". One has 7K+ and the other 10K+ wu's in cache.

I gave up trying to see how many errors/invalid wu's they had. TONS of them!

Martin

Profile Fred J. Verster
Volunteer tester
Send message
Joined: 21 Apr 04
Posts: 3238
Credit: 31,751,856
RAC: 4,205
Netherlands
Message 1093573 - Posted: 5 Apr 2011, 9:47:23 UTC - in response to Message 1086856.

Just plain stupid having an FERMI card, but 'forget' to install
The right V0.37 Installer, instead
the 'older' apps for 200 series of GPU?!. :/

All those false -9 (Result Overflow) or Found Triplets in a Row, have
to be computed again.



____________

Profile Fred J. Verster
Volunteer tester
Send message
Joined: 21 Apr 04
Posts: 3238
Credit: 31,751,856
RAC: 4,205
Netherlands
Message 1093579 - Posted: 5 Apr 2011, 10:36:02 UTC - in response to Message 1093573.
Last modified: 5 Apr 2011, 11:32:28 UTC

Host 5157509, running GTX295 GPUs (too hot?).

Host 5157509, again.

This WU was send to 2 other hosts, who reached a Canonical Result and
ruled out the error on the above mentioned host.

Knowing this, it is possible, a lot more false -9; Result Overflow, could be stored already in DB, but have to be computed, yet again!!
____________

Profile Fred J. Verster
Volunteer tester
Send message
Joined: 21 Apr 04
Posts: 3238
Credit: 31,751,856
RAC: 4,205
Netherlands
Message 1093589 - Posted: 5 Apr 2011, 11:57:57 UTC - in response to Message 1093579.

Could not edit the previous post so I'll continue............

Host 5396281.
http://setiathome.berkeley.edu/show_host_detail.php?hostid=5396281

SETI@Home Informational message -9 result_overflow
NOTE: The number of results detected exceeds the storage space allocated.

Flopcounter: 331167384.000000
Spike count: 30

Host 4294251.
SETI@Home Informational message -9 result_overflow
NOTE: The number of results detected exceeds the storage space allocated.

Flopcounter: 24376980131.228958

Spike count: 14
Pulse count: 16
Triplet count: 0
Gaussian count: 0
called boinc_finish.

Are all -9 , overflow of results being done again? IMHO it has to
be done, bcs errors don't belong in the Science Data Base!


____________

Profile BilBg
Volunteer tester
Avatar
Send message
Joined: 27 May 07
Posts: 2679
Credit: 6,059,590
RAC: 4,072
Bulgaria
Message 1093665 - Posted: 5 Apr 2011, 16:33:18 UTC - in response to Message 1093589.

Are all -9 , overflow of results being done again?


No, if two results validate (even because of wrong app) the WU is considered noisy and is not computed (send) again to another computer.


____________



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

Profile perryjay
Volunteer tester
Avatar
Send message
Joined: 20 Aug 02
Posts: 3377
Credit: 15,488,084
RAC: 11,271
United States
Message 1093737 - Posted: 6 Apr 2011, 2:21:07 UTC - in response to Message 1093665.

Are all -9 , overflow of results being done again?


No, if two results validate (even because of wrong app) the WU is considered noisy and is not computed (send) again to another computer.



What could be very bad is that one of these wrong -9s might be the one that has the signal we are all searching for. :-(

____________


PROUD MEMBER OF Team Starfire World BOINC

The Weasel
Send message
Joined: 6 Jun 99
Posts: 127
Credit: 53,156,207
RAC: 5,370
United States
Message 1093744 - Posted: 6 Apr 2011, 2:35:45 UTC

Very true, I admit that I am not always on top of what is going on in SETI, but I always try and get the most updated version of Boinc/Seti going along with the drivers for my computers. When I actually get a chance to read the forums, these conversations always make me make sure I'm doing just that. I have recently noticed a lot of computers that are trying to verify my tasks with a computer name of anonymous. I usually do not recognise computer names, but that name is starting to stand out. When I check the computer, they always seems to be a new computer that only has 1-7 tasks total in history. Is anyone else experiencing this?
____________

Profile BilBg
Volunteer tester
Avatar
Send message
Joined: 27 May 07
Posts: 2679
Credit: 6,059,590
RAC: 4,072
Bulgaria
Message 1093760 - Posted: 6 Apr 2011, 3:08:34 UTC - in response to Message 1093744.
Last modified: 6 Apr 2011, 3:24:56 UTC


Your Computers are also hidden, so if you are my "WU-partner" I will see in (possibly yours) Computer details "Owner - Anonymous".

I don't know why people hide their Computers - maybe they think we can see the "real" Computer (and files in it) and not just a few details.

Should SETI@home show your computers on its web site? yes

maybe have to be rephrased, e.g.:
Should SETI@home show info about your computers on its web site? yes
Should SETI@home show hardware/software configuration of your computers on its web site? yes
Should SETI@home show info about your computers (type of CPU, GPU, OS, BOINC, etc...) on its web site? yes

http://setiathome.berkeley.edu/prefs.php?subset=project


____________



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

The Weasel
Send message
Joined: 6 Jun 99
Posts: 127
Credit: 53,156,207
RAC: 5,370
United States
Message 1094161 - Posted: 7 Apr 2011, 1:53:04 UTC

After all the years, I never knew that if you hide computers it comes up as that, I will have to adjust my setting. Ty for clearing that up for me. I was starting to wonder how some many units could go to one user, lol.
____________

Profile ignorance is no excuse
Avatar
Send message
Joined: 4 Oct 00
Posts: 9529
Credit: 44,433,274
RAC: 0
Korea, North
Message 1098724 - Posted: 18 Apr 2011, 15:02:13 UTC - in response to Message 1094161.

What people aren't aware of is that the default is Computers hidden. They have to actually unhide them on purpose.
____________
In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

End terrorism by building a school

Questions and Answers : GPU applications : GPU and CPU disagree

Copyright © 2014 University of California