Roundoff Error Message

2004-01-11, 17:09

Hi, my computer has showed a Round Off error while it works on LL. Could be it a problem for the test? What can be wrong on my computer?

Hola, mientras realizaba el Test de Lucas he observado un error del tipo Round Off. Esto puede ser un inconveniente para la fiabilidad del test? Puede ser producto de algun error de hardware o software en mi ordenador?

dsouza123 · 2004-01-11, 17:44

Here is a link to a thread that is about Round off error.

http://mersenneforum.org/showthread.php?threadid=1490

Hope this helps.

2004-01-11, 18:40

OK, thanks.

The chance for bad result is about 50%... Bad news for me :(
My computer was working on 15.020.000 iteration (44,60%)....

Prime95 · 2004-01-12, 04:26

Some round off errors are OK. If the results.txt later said something like "Result is reproducible - ignore previous error" then all is OK.

Wolf · 2004-01-12, 19:07

Quote:

Some round off errors are OK. If the results.txt later said something like "Result is reproducible - ignore previous error" then all is OK.

I got a couple of round off error recently on one of my exponents that later was verified OK.

This is what you should be looking for in your results file

Iteration: X / Y, ERROR: ROUND OFF (0.40625) > 0.40
Continuing from last save file.
Disregard last error. Result is reproducible and thus not a hardware problem.
For added safety, redoing iteration using a slower, more reliable method.
Continuing from last save file

where X is the iteration the error occured and Y is the exponent you are testing.

2004-01-12, 22:33

This is the results.txt message:

[Sun Jan 11 17:15:46 2004]
Iteration: 15028608/33671789, ERROR: ROUND OFF (0.5) > 0.40
Continuing from last save file.

Is my previous work right? How can I know it?

dsouza123 · 2004-01-13, 01:20

If it reverts back to a checkpoint ( savefile ) which holds the state of the Prime95 work before the error happened it should be OK.
It will redo the work done since the savefile was written to disk.

Prime95 · 2004-01-13, 19:28

Quote:

Originally posted by Teseo77Madrid
Iteration: 15028608/33671789, ERROR: ROUND OFF (0.5) > 0.40

Is my previous work right? How can I know it?

The key here is that you did not get a "Disregard last error" message. Your hardware is suspect.

The reported error did not cause your calculation to go wrong. The danger is that you may have had a hardware error that prime95 did not detect. From experience we find that machines that report one hardware error during a test have a 50% chance of producing a correct result.

2004-01-13, 22:55

Will be my computer chance 100% for any other exponent if I complete this exponent and LL result is correct in double check?

2004-02-07, 20:36

Here's the error i got:

FATAL ERROR: Rounding was 0.4997747507, expected less than 0.4
Hardware failure detected, consult stress.txt file.

its a new system and fresh install of win2k.

Any idea what this means? Would appreciate any help :)

flava · 2004-02-07, 22:51

It could be a too aggressive overclock, or too aggressive memory timings, or a faulty memory, or a bad cooler. If your system turns at stock speed and the temperatures are reasonable, try running memtest86 to see if there is a problem with your memory.

2004-01-11, 17:09	#1
Teseo77Madrid 17345₈ Posts	Roundoff Error Message Hi, my computer has showed a Round Off error while it works on LL. Could be it a problem for the test? What can be wrong on my computer? Hola, mientras realizaba el Test de Lucas he observado un error del tipo Round Off. Esto puede ser un inconveniente para la fiabilidad del test? Puede ser producto de algun error de hardware o software en mi ordenador?

2004-01-11, 18:40	#3
Teseo77Madrid 31·211 Posts	Thanks OK, thanks. The chance for bad result is about 50%... Bad news for me :( My computer was working on 15.020.000 iteration (44,60%)....

2004-02-07, 20:36	#10
Unregistered 1255₈ Posts	FATAL ERROR: Rounding was 0.4997747507, expected less than 0.4 Here's the error i got: FATAL ERROR: Rounding was 0.4997747507, expected less than 0.4 Hardware failure detected, consult stress.txt file. its a new system and fresh install of win2k. Any idea what this means? Would appreciate any help :) Last fiddled with by ewmayer on 2005-12-30 at 22:16

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Roundoff error	bcp19	Software	4	2013-02-14 21:23
error message help?	AurKayne	Hardware	3	2005-08-29 09:13
Error message...	Xyzzy	GMP-ECM	2	2005-03-04 20:17
Roundoff Error Penalty	nevarcds	Software	5	2004-08-28 14:29
Error message	McBryce	NFSNET Discussion	2	2003-07-07 11:35

2004-01-11, 17:44	#2
dsouza123 Sep 2002 2·331 Posts	Here is a link to a thread that is about Round off error. http://mersenneforum.org/showthread.php?threadid=1490 Hope this helps.

2004-01-12, 04:26	#4
Prime95 P90 years forever! Aug 2002 Yeehaw, FL 2×3×5²×7² Posts	Some round off errors are OK. If the results.txt later said something like "Result is reproducible - ignore previous error" then all is OK.

2004-01-12, 22:33	#6
Teseo77Madrid 2×23×151 Posts	This is the results.txt message: [Sun Jan 11 17:15:46 2004] Iteration: 15028608/33671789, ERROR: ROUND OFF (0.5) > 0.40 Continuing from last save file. Is my previous work right? How can I know it?

2004-01-13, 01:20	#7
dsouza123 Sep 2002 2×331 Posts	If it reverts back to a checkpoint ( savefile ) which holds the state of the Prime95 work before the error happened it should be OK. It will redo the work done since the savefile was written to disk.

2004-01-13, 22:55	#9
Teseo77Madrid 5²×101 Posts	Will be my computer chance 100% for any other exponent if I complete this exponent and LL result is correct in double check?

2004-02-07, 22:51	#11
flava Feb 2003 2×59 Posts	It could be a too aggressive overclock, or too aggressive memory timings, or a faulty memory, or a bad cooler. If your system turns at stock speed and the temperatures are reasonable, try running memtest86 to see if there is a problem with your memory.