SETI Issues

loudbob

loudbob

loudbob

Associate
Joined
18 Oct 2004
Posts
1,651
Location
Cumbria
This one goes out to:

TheTross, Grimgoth & The Dogfather.

Well guys, after numerous attempts to get a decent ppd on SETI & Win7 I have finally given up.

I just couldn't get everything working happily together no matter what I tried.

Issues with NVIDIA drivers, Lunatics Optimized App's, CUDA 2.3 dll's (everytime I tried these I kept getting computational errors on every CUDA work unit) and/or dummy vga's.

My ppd was just dire (to say the least).

So, I bit the bullit and re-installed WinXP (dual boot config).

Everything seems to be running ok, I just have to let things settle and get a decent amount of work units to see if it is really stable and to see if my ppd improve.

Wish me luck.
 
hmm, I have to admit, that my error rate is quite high everyday- about 20 or so WUs on my windows 7 rig, all GPU workunits- thankfully each error is after about 1.5 sec of computing time, so....
methinks its a optimised app getting the wrong kind of WU.
but I could be so wrong.

edit: oops i lied, about 4 secs of gpu computing time av on a failed wu.
This is my most common error code:
<core_client_version>6.10.18</core_client_version>
<![CDATA[
<message>
- exit code -6 (0xfffffffa)
</message>
<stderr_txt>

VLAR WU (AR: 0.009600 )detected... autokill initialised
SETI@home error -6 Bad workunit header

File: ..\worker.cpp
Line: 144


</stderr_txt>
]]>
 
Last edited:
Grimgoth, that's not an error - the 'autokill initialised' bit just means that the optimised GPU app you installed is automatically aborting VLAR WUs as soon as they start running on your GPU. (VLAR WUs take an absolute age on GPUs for some reason, although they run perfectly fine on CPUs.) The way around that is to install ReSchedule - it's somewhere on the Lunatics site - and use it to reassign any GPU VLAR WUs to the CPU.

As for W7... I've never used it, so cannot help with that. Running under trusty ol' XP here :D
 
ah thanks TT, do you keep reschedule running all the time? I am guessing so- any particular settings I should be looking out for?
 
we had some discussion on rebranding tool here...
http://forums.overclockers.co.uk/showthread.php?p=15612564#post15612564

I don't run it all the time as the client sometimes fails to log back in to the service. So I only use it when they start sending out lots of VLARs.
Although if they send out too many then my crappy cpu can't keep up with them either so something has to give - currently have about 3 weeks worth of them !
 
I run reschedule every time I download GPU WUs - typically once every day or two. It's not efficient, pretty or even practical, but it works well, though quite what I'll do when I'm away from my PC for any more than a couple of days is anyone's guess...
 
yeah, but the boinc client afterwards has problems reconnecting- so I end up restarting boinc- the way TT does it I believe- wonders why I didn't install it as a service in the first place or something.
Also, I keep rescheduler minimised, rather than getting task manager to run it every so often. adds to the clutter i guess.
Another thing I found: Win 7 is worse than vista- the c:\ProgramData\Boinc folder is so protected, the rescheduler can't list it in the "browse" directory, and I have to type it in manually, even after unhiding the folder and removing the hide system files option. sigh.
 
Hmm, the number of error wu's has dropped to just two today (I guess the rescheduler is working :) )
the two are from the same pute- with the same error:

Work Unit Info:
...............
WU true angle range is : 1.940774
After app init: total GPU memory 939196416 free GPU memory 797998080
Unhandled Exception Detected...
- Unhandled Exception Record -
Reason: Out Of Memory (C++ Exception) (0xe06d7363) at address 0x7C812AFB
Engaging BOINC Windows Runtime Debugger...

erm is that a memory fail in my GTX260? both times at same address. any way of testing memory on a graphics card?
 
I get the 'out of memory' error sometimes too. I just put it down to noisy WUs, as it happens for both GPU and CPU. If it's only happened to a handful of WUs then I'd say there's nothing to worry about :)
 
I was getting that on every wu that one of my GPU's was trying to crunch.

Then I noticed it had dropped it's shader clock to 2D.

After a quick reboot everything was fine.

Anyways, I have since dropped my shader clocks from 1548 to 1476 and everything seems stable again.

I now have most of my HD encoding out of the way too, so I should be able to let my rig crunch away for a bit and perhaps get a decent ppd.

Current pending 42k and slowly rising.

:):):)
 
Since SETI is now not communicating, and one of my rigs is out of gpu work- how do you use rescheduler to rebrand cpu--> gpu?
can't seem to find any settings, and the slider of how much work % to cpu/gpu isn't moving one bit.
 
Under the Settings tab, uncheck 'Only VLar+VHar to CPU'. Go back to the Main tab and then set the slider to whatever percentage you want, then Run.

Bear in mind that if there are no non-VLAR WUs available, the rescheduler will do nothing as it won't put VLAR WUs on the GPU. This is exactly my predicament: 281 CPU WUs, 281 of which are VLAR :(
 
Connection is pretty sporadic at best at the moment. I've noticed that a few WUs are being uploaded but it's very slow going, with most transfers failing.
 
They broke down after the A/C failed a couple of days ago, and are still recovering. One machine here has run out, the other two will within a day.


M
 
You must log in or register to reply here.
Back
Top Bottom