mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Hardware

Reply
 
Thread Tools
Old 2006-04-27, 04:39   #1
Unregistered
 

25×3×5 Posts
Default Error always at exact same spot...

Hi, I have a new AMD system and I'm running Prime95 (just downloaded) to test the processor. I have run Mem86 already and the memory is fine.
However, the torture test always stops exactely at the same point with the same error (Test 1 , 4000 Lucas-Lehmer iterations of m19922945 using 1024k fft length. Fatal error: Rounding was 0.4927285672, expected less than 0.4))

And the stress.txt suggest to post here, in case there is a software bug.

This is my system:
AMD x2 4600+, 2gb OCX dual channel ram, nVidia nForce4 mobo

BTW - according to nTune the processor has a temp of 54-56 degress while running Prime95

Thanks for any help!

Cheers
  Reply With Quote
Old 2006-04-27, 12:14   #2
Cruelty
 
Cruelty's Avatar
 
May 2005

110010110002 Posts
Default

If you are running only one instance of Prime95 and you get temperature of ~56C then your CPU cooler does not perform well.

BTW: Which stress test are you using?
Cruelty is offline   Reply With Quote
Old 2006-04-27, 13:57   #3
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

2×53×71 Posts
Default

You are wise to post here about the possibility of a software bug. It did happen once. However, in this case the torture test has completed test 1 successfully on thousands of CPUs - AMDs included. I suspect the error message is not identical each time, especially the "0.4927285672" in "Fatal error: Rounding was 0.4927285672, ..."

The most likely culprit is memory. Try running the small FFTs in place torture test. If that works then memory is even more likely the culprit.
Prime95 is online now   Reply With Quote
Old 2006-04-28, 03:52   #4
Unregistered
 

29×191 Posts
Default

Thanks for the answers! I have been trying the in-place ffts and they run fine... So I guess it is the memory.

Interesting thing is the Memtest86 gave them a clean bill of health.

Question: I bought the OCZ Platinum 1gb Dual Channel kit. Is there a special sequence to plug in the mems, iaw - does it matter which one goes into the first slot and which one goes into the second?

Also - I changed the BIOS setting for the fans to go to max-speed if the CPU goes above 50 degrees, so now I have more fan noise but the CPU stays below 50. Strange that the Auto-setting would not ramp up the speed...

Thanks so much, I'd love to get my baby to run and use it ;-)

Last fiddled with by Unregistered on 2006-04-28 at 04:01
  Reply With Quote
Old 2006-04-28, 13:16   #5
Cruelty
 
Cruelty's Avatar
 
May 2005

162410 Posts
Default

Quote:
Originally Posted by Unregistered
Question: I bought the OCZ Platinum 1gb Dual Channel kit. Is there a special sequence to plug in the mems, iaw - does it matter which one goes into the first slot and which one goes into the second?
Yes it does matter - consult motherboard manual for right combination of memory slots in dual-channel operation (I assume both modules are identical).
Cruelty is offline   Reply With Quote
Old 2006-04-28, 14:10   #6
Unregistered
 

2·2,281 Posts
Default

okay, will do, thanks!

I ran memtest86+ tonight for 7 hours without any error.

So with the Prime95 running fine inside the FTT and the memtest giving a very clean bill of health to my memory I wonder if the commmunication between the two can be screwed up? How can I check the bus etc to make sure they run the correct way? Any hints for this?

Thanks so much guys for your help!

Cheers

P
  Reply With Quote
Old 2006-04-28, 18:26   #7
Cruelty
 
Cruelty's Avatar
 
May 2005

23·7·29 Posts
Default

Memtest is not so "proficient" at finding memory problems as Prime95 is... in case of Athlon64 this is a little bit more complicated as the memory controller is integrated in the CPU itself. So in your case it still may be a CPU-related problem.

I suggest running "blend" test to verify system stability with A64. For dual-core configurations you would have to launch two instances of prime95 and after first choosing "blend" mode, choose "custom" and set the available memory to half of the suggested amount (you have to do it for both instances separately). If system survives 24h of such testing then you have a prime-stable PC

What you can try to do now is to increase a Vcore by a "notch", increase Vram a bit, or relax memory timings a little bit. You may also fiddle with LDT and chipset voltage, but that shouldn't matter much if you are not overclocking (I assume you don't as you haven't mentioned it).
Cruelty is offline   Reply With Quote
Old 2006-04-29, 03:10   #8
Unregistered
 

22·2,017 Posts
Default

Cruelty,
you bely your bely your name! Thanks so much, will try right away, as the thing still crashes on the "belnd test" with the error I mentioned in my first post. (and, yes, you are right. it just seemed to be the same number always it stops with. The 0.49... is always a little different!)

will get back to you with results and hopefully some success asap. And - no, I am not overclocking.

Cheers
  Reply With Quote
Old 2006-04-30, 22:42   #9
Unregistered
 

188610 Posts
Default

I too didn't understand your choice of id (until now).
  Reply With Quote
Old 2006-05-01, 01:07   #10
moo
 
moo's Avatar
 
Jul 2004
Nowhere

809 Posts
Default

may i susgest you give us some ambiant tempatures
i would like you to do 3 different probes for me
first i would like you to close all running programs and open ntune let it sit for 10 minutes - half an hour, watch some tv or somthing then write down the temp

next i would like you to run one instance of prime95 let it run for a few minutes of you can that wont crash have ntune open, watch some more tv and write down temp

last i would like you to make a copy of your prime95 folder call it prime952 in the same folder you insatlled the origional prime95
open up notepad
type
path to prime95/prime95.exe -A0 -t
path to prime95/prime95.exe -A1 -t
chose save as
then change save as type to all files
and save it as test.bat on your desktop
run that and have ntune open you should see prime95 2 times in task bar
let it run well you watch another tv show and tell me temp if its still running

report those 3 temps here also get the temps of the chipset if you can.
moo is offline   Reply With Quote
Old 2006-05-06, 13:42   #11
FlyingScotsman
 

23·7·11·13 Posts
Default Same Problem

Yo dude I was reading through this post and found I have the same problem
Fatal Error Rounding was 0.49999 , expected less than 0.4
all the test some up with same.
I have a FX-57 running at 3.0 Ghz on an Nf4 mobo.
2gb OCZ platinum 2-3-2-2 1t
Water cooled so no probs with temps
Idle 20 deg C and load (Cpu burn 4 hours) 44 deg C
I thought my system was stable as had no problems with games or movie encoding etc.
Have never had any BSOD's or reboots
why wont this run?
  Reply With Quote
Reply



Similar Threads
Thread Thread Starter Forum Replies Last Post
EC2 Spot Instance stop & resume Mark Rose Cloud Computing 23 2018-01-30 22:32
Unfavorable price trends for AWS EC2 spot instances? GP2 Cloud Computing 14 2018-01-11 02:01
Spot the Difference a1call Miscellaneous Math 34 2017-12-11 00:17
functions which don't have an exact Taylor Series. wildrabbitt Math 6 2017-07-21 13:03
New exact theorem Master Alex Miscellaneous Math 38 2007-03-05 18:30

All times are UTC. The time now is 23:14.


Fri Jul 16 23:14:33 UTC 2021 up 49 days, 21:01, 1 user, load averages: 1.16, 1.35, 1.64

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.