PDA

View Full Version : 32m prime considered 24/7 stable?


Skratch
07-12-2008, 07:02 PM
Just thought I might ask some of you guys since most of us keep tweaking for every little last bit of speed,do you think 32m is stable enough for most ever day tasks?

Its not prime stable but the computer never freezes or bsod,running nice at 475x8 for 3800 on 1.432 v core ram at 570 55515

road-runner
07-12-2008, 07:12 PM
Everyone has a different opinion of stable, myself if I can not pass 1 hour of OCCT I dont use it...

Skratch
07-12-2008, 07:20 PM
Everyone has a different opinion of stable, myself if I can not pass 1 hour of OCCT I dont use it...

ya tru.Does occt strees fsb and memory tho.From all my years of overclocking its bad memory that causes crashes and not really cpu streesing.Like I can be prime stable for an hour but cant complete a super pi 16 mil

jcool
07-12-2008, 07:51 PM
32m prime considered 24/7 stable?

Hell no ;)

Depends on your computer usage though. I run all my rigs 24/7 and never idle (crunching FTW).
Current uptimes are 1 week 6 days for my mainrig and around 1 month for the dual quad.

Obviously you don't care about 24/7 stability if you boot up your pc in the morning, do some light work and start a game for half an hour, then shut it down a few hours later ;)

Telo
07-12-2008, 08:19 PM
Just thought I might ask some of you guys since most of us keep tweaking for every little last bit of speed,do you think 32m is stable enough for most ever day tasks?

Its not prime stable but the computer never freezes or bsod,running nice at 475x8 for 3800 on 1.432 v core ram at 570 55515

I have had many overclocks that go 32M stable 5 times on a row, but when I run Prime95 or OCCT they would crash. So the short answer is NO. You should use 32M test as a first step, then move to harder tests, like Orthos/Prime95 mixed mode; then you do small FFTs on either of the two mentioned before. A rule of thumb is to not consider your OC stable unless it can grind on Prime95/Orthos/OCCT for at least 12hrs [meaning you stopped it after 12hrs, not that an error stopped it]; however, 24hrs is ideal.

Peter949
07-12-2008, 08:42 PM
I don't consider anything less then 8hours of prime blend or small ffts cpu a 24/7 cpu stable. Guess I'm so tired of people doing a few minutes of prime to a hour and think it's stable. But, thats just my opinion.

Leeghoofd
07-13-2008, 12:47 AM
My stability testing :

One hour of Small FFT's, 2 hours of OCCT mixed, Superpi32mb for each core ( so 4 for a quad) and then 3dbenches ( 2005 and 2006 for a few loops ) then my rig if not being gamed upon goes into folding mode... some prefer Linpack but I think it's too harsh... Also I don't like prime for long sessions as when a core crashes it will continue on the other cores making it pretty useless... and a complete waste of time and electricity to me...

dinos22
07-13-2008, 01:04 AM
superpi should always be a part of testing of a new rig as part of full testing

and only time is a full proof 24/7 test
no benchmark has absolutely 100% results and if it did you'd probably have to stress the CPU for so long that it would probably get enough stress for a few years just to find something it doesn't like

LagunaX
07-13-2008, 01:13 AM
Sounds like for what you do 1-2 hours of orthos blend should suffice. That should cover the timespan for stress caused by video conversions/heavy gaming/etc.

kiwi
07-13-2008, 01:14 AM
1 hour occt ram + prime (25.6) blend 4 hours is usually enough. You can add HCI memtest ~30min

Mykou
07-13-2008, 03:07 AM
I consider my rig stable if 4/5 hours prime 25.6 /orthos small fft ( for my 45 dual cores and when ram not o/c ) , 3dmark 06 and run some heavy games .

graysky
07-13-2008, 03:36 AM
Depends on what you do with the machine... for video encoding, I don't use it unless it's >18 h large FFT's stable (p95v25.x) and at least 100 itterations of linpack64 (http://www.xtremesystems.org/forums/showthread.php?t=175729). Below are my runme_xeon64.bat, and linunput_xeon64 if you wanna try it

batch file:@echo off
set KMP_AFFINITY=compact

echo "This is a SAMPLE run script. Change it to reflect the correct number"
echo "of CPUs/threads, problem input files, etc.."

date /t
time /t
linpack_xeon64.exe lininput_xeon64 > win_xeon64.txt
date /t >> win_xeon64.txt
time /t >> win_xeon64.txt
echo "Done: "
date /t
time /t

lininput_xeon64:Sample Intel(R) LINPACK data file (lininput_xeon64)
Intel(R) LINPACK data
1 # number of tests
20000 # problem sizes
20000 # leading dimensions
1000 # times to run a test
4 # alignment values (in KBytes)

Linpack64 is one of the most aggressive CPU stability tests you can run.