GPU workunit problems?
14 Sep 2013 3:05:14 UTC
Topic 197191
(moderation:
I've taken to aborting Einstein workunits that use the GPU; they don't seem to be reporting right (presumably not running right, either).
Example: PA0023-002D1-131524_0_1_0
It's reporting 01:48:22 elapsed, 02:14:47 remaining, and 1.895% progress. My math skills lead me to suspect progress is much closer to 50% (somewhat short, but not below 2%!)
Wassup?
Thanks in advance.
GPU workunit problems?
)
With your computers hidden it is difficult to see what kind of GPU you use, but if it's an AMD/ATI, then make sure that BOINC is using one less CPU core for CPU work, so that the GPU can use it all. That'll speed up calculations enormously. (On multiprocessors, use at most 100% of the processors -> 99% will always tell BOINC to use one less CPU core).
Else tell details about your hardware.
Also know that BOINC won't know from day 1 what the exact times are that tasks run for, it will have to calculate that and will do so over the run of a tenfold or so of tasks of about the same run time. This does mean you'll have to let tasks run to their end, not abort them.
The Remaining time is Estimated, it says so in the column. It really does mean that, estimated, as no one computer running the work at Einstein is the same as the computer they tested the work on, so they can't possibly give a correct remaining time.
RE: I've taken to aborting
)
Aborting those tasks is not the wisest action to choose. ;-)
If you can't or don't want to follow Ageless's advice, go to your Einstein@home online preferences and disallow the use of your GPU.
Gruß
Gundolf
Computer sind nicht alles im Leben. (Kleiner Scherz)
RE: I've taken to aborting
)
I'm experiencing a similar, but more severe version of what Garry was experiencing. I have a Sony Vaio with a quad-core i7 and an Nvidia GeForce 735M GPU, plus 8GB RAM, so resources shouldn't be an issue (at least, I don't think so).
My E@H WUs are running indefinitely and show 0.000% progress. The current WU requires "0.5 CPUs + 1 intel_gpu GPU", according to the BOINC Manager. This WU was scheduled for a runtime of 03:29:32 hours, but has now run for just over 6 hours with zero progress showing. I've already aborted the first one that had been running for over 10-1/2 hours with zero progress, and now this next one is behaving similarly. When I shut down BOINC and restarted my computer after a Windows update, the WU restarted from scratch and behaved exactly the same way.
I just set the CPU percentage at 99%, as you had suggested to Garry, and I'll give it the originally allocated amount of time as additional runtime, but I'm not expecting different results. Everything else that I'm running with BOINC seems to run and complete normally, including WUs that also use the GPU, such as SETI@home.
The name of the Work Unit currently running is:
p2030.20130202.G203.78+00.82.N.b1s0g0.00000_2207_1
There are still 7 other related WUs waiting to start that will likely behave in a similar manner, unless you can guide me in a different direction.
This behavior seems to have just started very recently, because until now I have over 56,000 points for Einstein@Home.
Your guidance would be greatly appreciated. If you can't see my computer, please let me know how to make it visible to you.
Thanks...
Respectfully,
Dave Haber
A potential clue to the
)
A potential clue to the errant behavior... I've discovered that my Nvidia GeForce 735M GPU update software updates the GPU driver version to R331, but for some reason, my Sony Vaio update software wants to update it back to R326. There seems to be a driver update battle going on behind the scenes, and the GPU appears to be just trying to adapt to the changes. I'm no GPU expert, so I can't speak from knowledge, but these events are just what I can observe as a layman.
As further information, I upgraded my operating system from Windows 8 to Windows 8.1. Windows 8.1 puts the Start button back onto the Desktop (No, it is NOT in the initial version of Windows 8! Microsoft listened to its customers and eventually reintegrated a Start button -- MUCH better...).
Are there any known issues with the Nvidia GeForce 735M GPU and Windows 8.1 on a Sony Vaio quad-core i7?
Meanwhile, I've made sure that my GPU driver version remains at R331, and I've re-enabled E@H to accept more tasks.
I'll post any further observations in this message thread.
You said it was running an
)
You said it was running an intel GPU task. That would be in the iGPU that the i7 has not the Nvidia. There was a bug where intel GPU tasks were trying to run on an AMD GPU but I thought that one had been dealt with.
Maybe you can run GPU-Z while it's running and see which one is actually processing the task. Also the startup messages in the BOINC event log to do with what GPUs it found might help.
BOINC blog
i would like to bump this
)
i would like to bump this forum, i have the exact same setup as Dave and started to get the exact same problem last night, i only have 1,000 EaH credits, and am very new to this project in general (bonic)
I tried up-dateing my HD 4000 driver, but nothing is working. stays at 0.000% while other projects chug away. I set my processor to 75% after reading other possible solutions, and nothing is working for me :(
Dave, did you find a solution? or has anyone else found a solution?
I am running windows 8.1 - and my intel driver is 10.18.10.3345
I am sorry, I also forgot to
)
I am sorry, I also forgot to mention, that the first couple (I saw 3 once) units that it tries give a computation error, then windows 8.1 notifies me that the Driver has unexpectedly crashed and recovered.. and then that is when it hangs on 0.000%
Thank you for your help and advice!
Jospeh
P.S. What does RAC: stand for?
Hi Joseph and welcome to
)
Hi Joseph and welcome to Einstein!
A few days ago I tried updating my Intel graphics driver to version 10.18.10.3345 and I could not get it to work, same problem as you are having, instant error on start of the task with an "access violation". I also have a Intel HD 4000.
My solution to the problem was to completely uninstall the driver and go back to what I had before, version 9.18.10.3257, the previously released version. That got things moving again.
The difference here is that I'm running Win 7 on a desktop while your running Win 8.1 on what seems to be a laptop so I can't guarantee it will work for you.
RAC = Recent average credit, read more here http://boinc.berkeley.edu/wiki/Computation_credit
Thank you much Holmis, and I
)
Thank you much Holmis, and I will check out that page. And yes this is a laptop, Sony Vaio SVF15A1CCXB
I will try stepping back my driver and see if that gets me any results.
Thanks for your input!
Hello, I have also problems
)
Hello, I have also problems with some Einstein WUs.
After the WU start with normal % count up, it hang btw stop counting.
Also the GPU usage goes down, so the calculation will Stop.
That´s bad because the resource will not be give free for the next WUs, so I lost points.
Here a screen shot and a Link to my error results:
http://einsteinathome.org/host/6882331/tasks&offset=0&show_names=1&state=5&appid=0
http://abload.de/img/einstein_errorexj30.jpg
If you need more information, please tell me which.
Best regards
WindHund@P3D