why does BRP4 produces much more errors than S6LV1 ?
30 Apr 2012 20:43:32 UTC
Topic 196304
(moderation:
2,427#286,761 tasks are invalid for the former, whereas only 131#454,949 are invalid for S6LV1 - this correlates to my experiences - I deselected BRP4 from my preferences due to too many errors.
I'm wondering about the root cause for the difference.
why does BRP4 produces much more errors than S6LV1 ?
)
BRP4 runs nicely on my Linux box, even showing graphics.S6LVI has some validate error.
Tullio
The main reason probably is
)
The main reason probably is that GPUs are numerically less reliable than CPUs. BRP4 is currently the only application that offers GPU (CUDA) application versions.
Another thing is that the core comuting routines of the GW Apps (S1LV1, S6Bucket etc.) are hand-tuned assembly code, making these literally identical for almost all platforms (the only exception being Mac OS PowerPC). BRP4 and FGRP1 Apps are compiled individually from C source by different compilers.
BM
BM
RE: The main reason
)
Ah - interesting. For my system then the compiler is the culprit for the errors I've, B7c I do not have a GPU.
Thx
The speed of a GPU is
)
The speed of a GPU is irresistible though..... Love em. My RAC is still climbing, and that's with just one
Hallo Toralf! What type of
)
Hallo Toralf!
What type of errors are you encountered with?
Are they marked as "Completed, marked as invalid" in your Resultslog/Invalid or marked as "Error while computing" in your Resultslog/Error? If they are of the last one, by what error code do they end up? This error code you will find in the appropriate Task ID. And do have all erroneous tasks the same error code?
Kind regards
Martin
Well, I can only see 1 error,
)
Well, I can only see 1 error, all others are too old in the meanwhile - and it for task 284425587 an "Error while computing" with details :
Stderr output
6.12.42
Maximum disk usage exceeded
]]>
I'm wondering b/c there's enough space in the file system.
Hallo Toralf! I had, or have
)
Hallo Toralf!
I had, or have the same kind of error but with LAT tasks and could not clearup this. See here also the answers. So I also disabled FRGP1/LAT, but that doesn´t work or work not safely. So I´ve to abort this tasks by hand. With tasks from BRP4 I´ve other crunching errors, but I´m running Win7Px64SP1.
In case you find the reason for this error, it will be fine, if you report here.
Kind regards and happy crunching
Martin
I run again a BPR4 task (and
)
I run again a BPR4 task (and furthermore switched during that to boinc 7.0.28 and) - the task was finished successfully.
But b/c I'm running Gentoo Linux I cannot say, whether this is purely related to the new boinc version only.
Two days ago I rejoined here
)
Two days ago I rejoined here with my mainsys. So far I have only one invalid result http://einsteinathome.org/workunit/123477212, two cuda pc's produced different results.
But: 12 validated and ~20 are pending, let's see what happens.
So two more failed to
)
So two more failed to validate, both against cuda-pc's.
Normally my mainsys is very reliable, nothing overclocked.
It's intresting, until now all failing wu's came from the newer HD6950, not from the older HD5850. Are there known issues?