Error during LL test — should I continue?

Chuck · 2012-05-10, 17:52

This is the first time this computer has received an error message during LL testing. Last night at 2 AM:

Code:

 
[May 10 02:27] Iteration: 14990000 / 45167567 [33.18%].  Per iteration time: 0.032 sec.
[May 10 02:29] Iteration: 14995072/45167567, ERROR: ROUND OFF (0.5) > 0.40
[May 10 02:29] Continuing from last save file.
[May 10 02:29] Resuming primality test of M45167567 using FFT length 2400K, Pass1=640, Pass2=3840
[May 10 02:29] Iteration: 14986736 / 45167567 [33.18%].
[May 10 02:29] Possible hardware errors have occurred during the test! 1 ROUNDOFF > 0.4.
[May 10 02:29] Confidence in final result is fair.
[May 10 02:31] Iteration: 14990000 / 45167567 [33.18%].  Per iteration time: 0.031 sec.
[May 10 02:31] Possible hardware errors have occurred during the test! 1 ROUNDOFF > 0.4.
[May 10 02:31] Confidence in final result is fair.
[May 10 02:36] Iteration: 15000000 / 45167567 [33.20%].  Per iteration time: 0.031 sec.
[May 10 02:36] Possible hardware errors have occurred during the test! 1 ROUNDOFF > 0.4.
[May 10 02:36] Confidence in final result is fair.

The warning repeats with each iteration message. Should I kill this, release the exponent and go on to another one?

Chuck · 2012-05-10, 19:21

What I decided to do:

I blew out the CPU radiator (really not much dust), and cut back the overclock a couple of notches since summer is coming and it is heating up here in Florida.

Stopped the worker, restored a backup file from Windows Backup from the day before the problem, and continued.

Chuck

Dubslow · 2012-05-10, 21:33

Should be fine. If you look in undoc.txt, you can control how often/how much the error message prints; you'll also notice that Prime95 also resorted to an older save file, though admittedly not as far back as you did; it just went back to the last 30-min checkpoint file. Between that, your manual restore, and the fact that it was only one error, you're probably fine. (PrimeNet will immediately reassign the exponent because of the non-zero error code, but I would expect that test to match yours.)

Chuck · 2012-05-11, 02:09

Will it cause any grief when the worker checks in with PrimeNet? The % complete is going to be a little smaller than it was at the last check-in.

Dubslow · 2012-05-11, 02:19

Nope. There are some of the exponents way below 45M which experience similar things -- and yet PrimeNet doesn't expire them. Check in the Data forum. (The difference it those have been going on for months like that; you have nothing to worry about.)

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
how to continue test after computer update or Prim95 close?	STCB	Information & Answers	2	2017-12-04 16:53
How to use Test > Continue button	Minter	Information & Answers	13	2009-11-09 19:17
error rates and P-1 test	drakkar67	Prime Sierpinski Project	9	2008-05-26 14:29
Torture Test error	Unregistered	Hardware	27	2005-12-29 15:37
Should we continue to crunch after an error occurs??	outlnder	Hardware	9	2003-02-12 10:13

2012-05-10, 19:21	#2
Chuck May 2011 Orange Park, FL 1110001010₂ Posts	What I decided to do: I blew out the CPU radiator (really not much dust), and cut back the overclock a couple of notches since summer is coming and it is heating up here in Florida. Stopped the worker, restored a backup file from Windows Backup from the day before the problem, and continued. Chuck

2012-05-10, 21:33	#3
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 3·29·83 Posts	Should be fine. If you look in undoc.txt, you can control how often/how much the error message prints; you'll also notice that Prime95 also resorted to an older save file, though admittedly not as far back as you did; it just went back to the last 30-min checkpoint file. Between that, your manual restore, and the fact that it was only one error, you're probably fine. (PrimeNet will immediately reassign the exponent because of the non-zero error code, but I would expect that test to match yours.)

2012-05-11, 02:09	#4
Chuck May 2011 Orange Park, FL 2·3·151 Posts	Will it cause any grief when the worker checks in with PrimeNet? The % complete is going to be a little smaller than it was at the last check-in.

2012-05-11, 02:19	#5
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 7221₁₀ Posts	Nope. There are some of the exponents way below 45M which experience similar things -- and yet PrimeNet doesn't expire them. Check in the Data forum. (The difference it those have been going on for months like that; you have nothing to worry about.)