Page 1 of 1
Posted: Fri Jan 09, 2009 5:42 am
Although I'm 10+ hrs prime stable in windows VUx64sp1 at default bios settings Prime24 Fails immediately "Fatal ERROR Writing to Temp File"
At this point I have tested it on many different bios setting and various clocks with the same result. As far as testing is concerened all other app that I use seem to work fine ie: Memtest. I believe this may be a bug so I thought I would report it. I see the ubcd50b8 is out so I will try it instead and report back.
I do not believe at this point it is a windows or driver problem correct me if I'm wrong. All my hardware and software use up to date drivers patches firmware and bios. Also at this point I have not disabled any CPU Configuration settings from within bios all is at default.
Posted: Fri Jan 09, 2009 6:33 am
The bug is solved in UBCD50b8.
Though, thanks for reporting the bug.
Posted: Fri Jan 09, 2009 11:21 am
No.... sorry ubcd50b8 doesn't work either.
Posted: Fri Jan 09, 2009 1:49 pm
mprime is a silly program. It wants a certain amount of free space for writing its temporary files. I reserved 12MB for it. It only writes files which need less than 100 bytes in size (check the file sizes in /tmp/torture-test/torture.*/results.txt or /tmp/torture-test/torture.*/local.ini*). Each time you run one instance of mprime it wants that amount of free space on the filesystem. It worked fine for me (2 cores).
To solve your problem, run the following at the command line (size=20M
should be more than enough, set to a higher value if it is not the case):
Code: Select all
mount -o remount /tmp -o size=20M
Is it correct that you have a quad code in your PC?
Change the link of Mersenne Prime Test
on the homepage to: http://www.mersenne.org/freesoft/
Currently it points to http://www.mersenne.org/freesoft.htm
(redirects to http://www.mersenne.org/freesoft/
There is a new version of mprime (December 15, 2008): mprime v25.8.
I will incorporate it in the next CPUstress image (and abandon v23 and v24).
How much memory may I use to avoid the "Fatal ERROR Writing to Temp File" message so that it can run on most PCs (minimum RAM in PC: 32MB or 64MB)? Maybe the new version doesn't have this problem (not tested yet).
I quickly scanned the code and the following error messages can't be found in the new version, but are in the v24 version:
"FATAL ERROR: Writing to temp file."
"FATAL ERROR: Reading from temp file."
So maybe the bug is gone.
Posted: Thu Feb 05, 2009 8:14 pm
This program works perfectly fine with no errors in ubcd50b10. If you are still getting errors please repost again.
Posted: Fri Feb 06, 2009 4:14 pm
on a dual core I think mprime works
but on a quad core I just installed /tmp is too small and I have to remount after each boot
that's not very nice
It would be nice to increase tmp in the image
Posted: Sat Feb 07, 2009 8:10 am
It will be fixed in the next release of the CPUstress image.
I think that I will stick to v23 and v24 of mprime.
At first look v25 needs more extra file to run and I think it isn't better for stress testing than v24.
Posted: Sat Feb 07, 2009 9:01 am
I did a few tests and it seems I need at least 17m for my Q6600 quadcore
Posted: Sat Feb 07, 2009 9:29 pm
gloutch wrote:I did a few tests and it seems I need at least 17m for my Q6600 quadcore
Maybe its a good idea, icecube, to add a dualcore/quadcore option that requires more ram to start with...because most people that have multicore's have more than 512mb of ram to begin with. Maybe 30mb to be safe.
Im sure you already thought of this now so...
Posted: Sun Feb 08, 2009 6:21 am
StopSpazzing wrote:Im sure you already thought of this now so...
Yes, I have thought to something like this. But the mprime program shouldn't require so much MB of space in the temp directory, because it writes only files of a few bytes to it (at least when using the torture switch).
Posted: Thu Mar 05, 2009 9:16 am
An updated version of the CPUstress test image is in UBCD50b12.