M77909869

guido72 · 2003-10-02, 22:43

Here we are! After 150 days of work, M77909869 has been tested! Thanks to George's Brain power and some improvements in micro-chips Architecture, it has been possible to get in few days what would has needed more than my 30 Years (OK, 31...) with A P90 CPU!
Unfortunately we'll have to run a primality test once again, due to errors occurred in the first run:

Code:

[Wed Apr 16 15:47:48 2003]
UID: Team_Italia/Paperino, M77909869 no factor to 2^72, WZ1: EBC80679
[Wed Apr 16 16:48:23 2003]
Self-test 4096K passed!
[Mon Jun 02 23:07:34 2003]
Iteration: 20308065/77909869, 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.
[Sun Jun 22 14:33:04 2003]
Iteration: 29884957/77909869, ERROR: ROUND OFF (0.40625) > 0.40
Continuing from last save file.
[Sun Jun 22 14:42:36 2003]
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.
[Fri Jul 04 03:18:51 2003]
Iteration: 35569322/77909869, ERROR: ROUND OFF (0.40625) > 0.40
Continuing from last save file.
[Fri Jul 04 03:27:18 2003]
..........
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.
[Wed Jul 09 18:48:35 2003]
Iteration: 38356928/77909869, ERROR: ROUND OFF (0.40625) > 0.40
Continuing from last save file.
[Wed Jul 09 18:54:53 2003]
Disregard last error.  Result is reproducible and thus not a hardware problem.
For added safety, redoing iteration using a slower, more reliable method.
Iteration: 46643868/77909869, ERROR: ROUND OFF (0.40625) > 0.40
[Sun Jul 27 06:44:57 2003]
Disregard last error.  Result is reproducible and thus not a hardware problem.
For added safety, redoing iteration using a slower, more reliable method.
[Tue Aug 19 07:54:47 2003]
Iteration: 57025428/77909869, ERROR: ROUND OFF (0.4375) > 0.40
Continuing from last save file.
[Tue Aug 19 08:01:21 2003]
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.
[Wed Aug 20 11:56:28 2003]
Iteration: 57565601/77909869, ERROR: ROUND OFF (0.40625) > 0.40
Continuing from last save file.
[Wed Aug 20 12:06:20 2003]
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.
[Thu Sep 11 06:43:03 2003]
Iteration: 67535315/77909869, 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.
[Thu Oct 02 21:49:44 2003]
UID: Team_Italia/Paperino, M77909869 is not prime. Res64: 29D550438D436C1E. WZ1: A49C6216,73335676,0B014103

Since this PC proved to be quite reliable, as you may see,

Code:

UID: Team_Italia/Paperino, M8427977 is not prime. Res64: 44CFC8A97A27BAB7. WY1: 7074F093,2364717,00000000
[Fri Dec 27 19:08:13 2002]
UID: Team_Italia/Paperino, M8434003 is not prime. Res64: 23C0828000D45BD2. WY1: DB8CB0B9,7979682,00000000
[Sun Dec 29 19:31:54 2002]
UID: Team_Italia/Paperino, M8434141 is not prime. Res64: 263E57AF7762D1FA. WY1: 62234B96,1323342,00000000
[Tue Dec 31 19:24:47 2002]
UID: Team_Italia/Paperino, M8434189 is not prime. Res64: 9AB2F0501AEB3EEC. WY1: 4A381D61,3136868,00000000
[Thu Jan 02 18:59:03 2003]
UID: Team_Italia/Paperino, M8434199 is not prime. Res64: 78CC30C686F532DB. WY1: 002924BF,6410697,00000000
[Sat Jan 04 18:52:02 2003]
UID: Team_Italia/Paperino, M8434213 is not prime. Res64: C99FE95D3F43240F. WY1: 09109C5F,5472322,00000000
[Mon Jan 06 18:43:26 2003]
UID: Team_Italia/Paperino, M8434259 is not prime. Res64: 3574165AC524A537. WY1: FB8BE249,7418995,00000000
[Wed Jan 08 22:34:11 2003]
UID: Team_Italia/Paperino, M8434403 is not prime. Res64: 13D7A14113C1E148. WY1: 260867B0,1110356,00000100
[Fri Jan 10 21:14:46 2003]
UID: Team_Italia/Paperino, M8434501 is not prime. Res64: 14E5567C8DF15BA5. WY1: A1422D93,5080961,00000000
[Sun Jan 12 19:56:00 2003]
UID: Team_Italia/Paperino, M8434513 is not prime. Res64: FDFE375BBCB4B4E3. WY1: B8A6AF4E,2477769,00000000
[Tue Jan 14 16:19:28 2003]
UID: Team_Italia/Paperino, M8062247 is not prime. Res64: D95D75D21C68AF7D. WY1: 1DCAE6AF,8024445,00000000
[Thu Jan 16 12:52:14 2003]
UID: Team_Italia/Paperino, M8078209 is not prime. Res64: 1FDB1A7A53D29972. WY1: 44531C0A,1221962,00000000
[Sat Jan 18 09:31:16 2003]
UID: Team_Italia/Paperino, M8082079 is not prime. Res64: 796C60E999A8C6A6. WY1: 46930908,4666470,00000000
[Mon Jan 20 06:34:45 2003]
UID: Team_Italia/Paperino, M8090629 is not prime. Res64: 94CAF44886191755. WY1: 414024D8,4473966,00000000
[Wed Jan 22 03:04:02 2003]
UID: Team_Italia/Paperino, M8099873 is not prime. Res64: C4E032E43CD2E8D5. WY1: A36F8A14,1833544,00000000
[Thu Jan 23 23:55:12 2003]
UID: Team_Italia/Paperino, M8105221 is not prime. Res64: 94C9D143CE913DCE. WY1: EDBF801B,3085803,00000000
[Sat Jan 25 20:47:57 2003]
UID: Team_Italia/Paperino, M8118829 is not prime. Res64: FA5B9547F2FD69D5. WY1: 15350E52,5756024,00000000
[Mon Jan 27 17:53:35 2003]
UID: Team_Italia/Paperino, M8144893 is not prime. Res64: 126BBD6E2B9185CE. WY1: 6B53DD5B,1923235,00000000
[Wed Jan 29 16:15:05 2003]
UID: Team_Italia/Paperino, M8401147 is not prime. Res64: 188414A09F9FA29B. WY1: C2AE61E3,7528780,00000000
[Fri Jan 31 16:36:51 2003]
UID: Team_Italia/Paperino, M8401219 is not prime. Res64: C0C835D55BB9BC09. WY1: 66372C53,8250313,00000000
[Sun Feb 02 12:48:23 2003]
UID: Team_Italia/Paperino, M8424289 is not prime. Res64: 327BD4125F8000C3. WY1: 20B4B775,3151847,00000000
[Tue Feb 04 09:05:23 2003]
UID: Team_Italia/Paperino, M8432747 is not prime. Res64: B0C1FDB736F5D1FA. WY1: 134CA2B2,4646041,00000000
[Thu Feb 06 05:14:10 2003]
UID: Team_Italia/Paperino, M8433497 is not prime. Res64: B00D6A4D381E710D. WY1: 12FB50DD,4812975,00000000
[Sat Feb 08 01:17:26 2003]
UID: Team_Italia/Paperino, M8433559 is not prime. Res64: EF0B02EE88091B6D. WY1: 8C657916,5981675,00000000
[Sun Feb 09 21:22:01 2003]
UID: Team_Italia/Paperino, M8435159 is not prime. Res64: E29DDCA22B729F6E. WY1: 09C396A3,3740646,00000000
[Tue Feb 11 17:34:45 2003]
UID: Team_Italia/Paperino, M8435291 is not prime. Res64: B6D972C0B4F3A30C. WY1: 636A8702,2852755,00000000
[Thu Feb 13 13:34:08 2003]
UID: Team_Italia/Paperino, M8435293 is not prime. Res64: 8BF0E9482AE58F26. WY1: BF821682,4035701,00000000
[Sat Feb 15 09:33:32 2003]
UID: Team_Italia/Paperino, M8435303 is not prime. Res64: B5D48FE6361323E7. WY1: E4EF3DB2,6124538,00000000
[Mon Feb 17 20:34:03 2003]
UID: Team_Italia/Paperino, M8436079 is not prime. Res64: BAEE218AA791C5FF. WY1: 6C6E230D,5512191,00000000
[Wed Feb 19 16:59:05 2003]
UID: Team_Italia/Paperino, M8438303 is not prime. Res64: FFEBE74591F4CE9E. WY1: B2148B04,3011723,00000000
[Fri Feb 21 15:40:58 2003]
UID: Team_Italia/Paperino, M8439241 is not prime. Res64: 3BB706F50EBF40F2. WY1: A3B602EF,1499754,00000000
[Sun Feb 23 11:55:57 2003]
UID: Team_Italia/Paperino, M8439791 is not prime. Res64: 40E454339B3682F6. WY1: 31BE10EF,1992380,00000000
[Tue Feb 25 08:02:40 2003]
UID: Team_Italia/Paperino, M8446883 is not prime. Res64: AE33595C33348729. WY1: D3B9B1B5,5902916,00000000
[Thu Feb 27 09:01:25 2003]
UID: Team_Italia/Paperino, M8446951 is not prime. Res64: 44CBA95A2558DEC5. WY1: 5980714D,6008106,00000000
[Sat Mar 01 05:06:27 2003]
UID: Team_Italia/Paperino, M8446961 is not prime. Res64: B9646500CD353718. WY1: B5AE6426,2130982,00000000
[Mon Mar 03 01:18:38 2003]
UID: Team_Italia/Paperino, M8448959 is not prime. Res64: 975C3ECE3E3E9399. WY1: 94A7354E,8164499,00000000
[Mon Mar 10 09:44:24 2003]
UID: Team_Italia/Paperino, M16201489 is not prime. Res64: 2BFFEA31E908A580. WZ1: 37CE8EA6,15932099,00000000
[Mon Mar 17 22:00:46 2003]
UID: Team_Italia/Paperino, M16366837 is not prime. Res64: 2BA8B38946D1B60E. WZ1: A9E61FBE,8164522,00000000
[Sat Mar 22 19:40:14 2003]
UID: Team_Italia/Paperino, M13117837 is not prime. Res64: A458F9725B98563D. WZ1: 852A7E4C,4780783,00000000
[Fri Mar 28 06:38:28 2003]
UID: Team_Italia/Paperino, M14549489 is not prime. Res64: 2A9F640896C12D45. WZ1: 4D2A1018,11509843,00000000
[Fri Apr 04 23:57:52 2003]
UID: Team_Italia/Paperino, M16609123 is not prime. Res64: BFC20CB4A0986166. WZ1: F3DEA079,14965688,00000000

I was wondering if there may be some relation between the size of FFT being used and the error rates, or is better to think about "Summer-Errors" (Last Summer has been really hot, here in Italy...)

GP2 · 2003-10-02, 23:40

Quote:

Originally posted by guido72
Since this PC proved to be quite reliable, as you may see,

[moderators of this forum: perhaps this thread could be moved to the Data forum?]

Guido72,

Please don't take this the wrong way, but that particular machine is not reliable.

I'm sorry to say, this machine was a very unfortunate choice to run such a huge exponent.

Out of the exponents you posted above, 12 are confirmed-bad results, 23 are confirmed-good results, 4 need a double-check, and 1 needs a triple-check.

Overall this machine has 30 confirmed-bad results and 61 confirmed-good results, 35 need a double-check, and 2 need a triple-check.

Of the 30 confirmed-bad results, all but two have zero errorcode. Zero errorcode is a very imperfect predictor of reliability!

Unreliable Team_Italia machines include: dudu, Cygnus_X1, Paperino, Pluto, Laura, pippo. I would suggest they be moved to double-checking only.

Here are the bad exponents from the list you posted above:

8062247,Team_Italia,Paperino,WY1,D95D75D21C68AF7D,8024445,00000000
8082079,Team_Italia,Paperino,WY1,796C60E999A8C6A6,4666470,00000000
8118829,Team_Italia,Paperino,WY1,FA5B9547F2FD69D5,5756024,00000000
8433497,Team_Italia,Paperino,WY1,B00D6A4D381E710D,4812975,00000000
8433559,Team_Italia,Paperino,WY1,EF0B02EE88091B6D,5981675,00000000
8434141,Team_Italia,Paperino,WY1,263E57AF7762D1FA,1323342,00000000
8434189,Team_Italia,Paperino,WY1,9AB2F0501AEB3EEC,3136868,00000000
8434403,Team_Italia,Paperino,WY1,13D7A14113C1E148,1110356,00000100
8434501,Team_Italia,Paperino,WY1,14E5567C8DF15BA5,5080961,00000000
8435291,Team_Italia,Paperino,WY1,B6D972C0B4F3A30C,2852755,00000000
8446961,Team_Italia,Paperino,WY1,B9646500CD353718,2130982,00000000
8448959,Team_Italia,Paperino,WY1,975C3ECE3E3E9399,8164499,00000000

Edit: changed 37 to 35: first time I counted them by hand, second time by script

guido72 · 2003-10-03, 00:48

Quote:

Originally posted by GP2
Guido72,
Please don't take this the wrong way, but that particular machine is not reliable.

I better say thank you for these infos! Where I may find more of them about our PCs?

Quote:

Of the 30 confirmed-bad results, all but two have zero errorcode. Zero errorcode is a very imperfect predictor of reliability!

So how to know immediately if something gone wrong? I was looking just to errorcode to set a "Primestable"" Overclock...

Quote:

Unreliable Team_Italia machines include: dudu, Cygnus_X1, Paperino, Pluto, Laura, pippo. I would suggest they be moved to double-checking only.

I'll do it (and I'll reduce the OCs). But a bad test is always a bad test... So this moving may look usless, or not? May be a relation between exp. size and errors in testing?
What about "Pisolo", "Cucciolo" and "Pepe"?

GP2 · 2003-10-03, 02:16

Quote:

Originally posted by guido72
I better say thank you for these infos! Where I may find more of them about our PCs?

Actually, I just recently started going through the data files to try to identify machines with high error rates. Perhaps in the future I will try to contact teams to let them know about this.

Quote:

So how to know immediately if something gone wrong? I was looking just to errorcode to set a "Primestable"" Overclock...

If the errorcode is nonzero, it's a good indicator of a problem.
But if the errorcode is zero, it's not a good indicator of no problem.

Quote:

What about "Pisolo", "Cucciolo" and "Pepe"?

Pisolo: 0 verified-bad, 0 verified-good, 5 unverified needing a double-check, 7 unverified needing a triple-check.
No verified statistics, but probably bad. The number of results needing a triple-check is a warning flag: it's much more likely that your machine produced the 7 errors rather than 7 independent errors by 7 different people.

Cucciolo: 0 verified-bad, 3 verified-good, 1 unverified needing a double-check, 13 unverified needing a triple-check, 1 unverified needing a quadruple-check.
I'm guessing probably bad.

Pepe: 0 verified-bad, 118 verified-good, 25 unverified needing a double-check, 1 unverified needing a triple-check.
Undoubtedly good.

Edit: changed 26 to 25: first time I counted them by hand, second time by script

ET_ · 2003-10-03, 13:10

Quote:

Unreliable Team_Italia machines include: dudu, Cygnus_X1, Paperino, Pluto, Laura, pippo. I would suggest they be moved to double-checking only.

Cygnus_X1 was an overckocked Celeron.
I tried to make a more conservative overclock and it seems that actually it works fine. Nonetheless I moved it to double-checking.

Thank you GP2

Luigi

eepiccolo · 2003-10-03, 13:19

A cool addition to the future improved Primenet would be to list the reliability of your computers in the individual account report.

GP2 · 2003-10-03, 15:45

Quote:

Originally posted by eepiccolo
A cool addition to the future improved Primenet would be to list the reliability of your computers in the individual account report.

Well, one problem is that we don't have adequate stats on many machines. An error-prone machine that has been doing only first-time LL checks that have all returned zero error code probably won't be detected until the leading edge of double-checking catches up to its lowest tested exponent.

At some point I might ask for volunteers to take on a sample exponents that don't have much chance of being a hidden prime but that will help improve statistics and knowledge of which machines are error-prone.

Actually I'm thinking about proactively contacting some of the teams about their error-prone machines. At least the high-profile ones with public contact info. And at the same time remind them that version 23 is much faster for P4s.

JC · 2003-10-05, 07:40

I'm using this method to be sure my overclocked machines are rock-solid reliable :

1) Run a torture test a few hours
2) Assign only double-checks for a duration of 1 to 2 months
3) Compare results of these double-checks to George's list of residues (updated once a week)
4) If proven reliable, switch to LL

Step 2) is in fact equivalent to a torture test but has two advantages :
- productive !
- very long , 1 to 2 months instead of a few hours.

Do not be too agressive with overclocking, just one single bad result will totally destroy the benefit of overclockling.

GP2 · 2016-08-17, 16:44

Quote:

Originally Posted by guido72

Here we are! After 150 days of work, M77909869 has been tested! Thanks to George's Brain power and some improvements in micro-chips Architecture, it has been possible to get in few days what would has needed more than my 30 Years (OK, 31...) with A P90 CPU!

Almost 13 years later, a double-check was performed and there is a mismatch: M77909869. The double-check took six days.

This was the final unverified result for the machine "Paperino", which is nowadays credited to Luigi Morelli rather than "Team_Italia".

The machine "Paperino" ended up with 51 bad results (assuming the mismatch for 77909869 is eventually verified bad), 89 good results (where 6 of the verified good results were later factored), and 2 unknown results which were factored before being double-checked: M14945083, which had no error code, and M15199087, which had an error code of 01000600.

GP2 · 2016-08-17, 17:17

Quote:

Originally Posted by guido72

What about "Pisolo", "Cucciolo" and "Pepe"?

Quote:

Originally Posted by GP2

Pisolo: 0 verified-bad, 0 verified-good, 5 unverified needing a double-check, 7 unverified needing a triple-check.
No verified statistics, but probably bad. The number of results needing a triple-check is a warning flag: it's much more likely that your machine produced the 7 errors rather than 7 independent errors by 7 different people.

Cucciolo: 0 verified-bad, 3 verified-good, 1 unverified needing a double-check, 13 unverified needing a triple-check, 1 unverified needing a quadruple-check.
I'm guessing probably bad.

Pepe: 0 verified-bad, 118 verified-good, 25 unverified needing a double-check, 1 unverified needing a triple-check.
Undoubtedly good.

Today, all of the above machines are credited to Luigi Morelli rather than "Team_Italia", and all their results have been verified.

The machine "Pisolo" ended up with 12 bad results and 16 good results.

The machine "Cucciolo" was a surprise. It ended up with 2 bad and 32 good results. Actually it shouldn't have been a surprise, since it had mostly been doing double checks of first-time LL results that had a nonzero error code, and that should have been noted at the time. The only bad results turned out to be M14285977 and M14303603 (the latter needed a quadruple check).

The machine "Pepe" ended up with 3 bad results and 171 good results (of which 11 good results have now been factored).

kladner · 2016-08-18, 00:11

Quote:

Almost 13 years later, a double-check was performed and there is a mismatch: M77909869. The double-check took six days.

.

Quote:

..almost 13 years...

I am still awed by the longevity of this project, and that of some of the earliest participants.
Cheers! @ GP2, Luigi, and all of you pioneers!

2003-10-03, 13:19	#6
eepiccolo Dec 2002 Frederick County, MD 2·5·37 Posts	A cool addition to the future improved Primenet would be to list the reliability of your computers in the individual account report.

2003-10-05, 07:40	#8
JC Aug 2002 Belgium 5² Posts	I'm using this method to be sure my overclocked machines are rock-solid reliable : 1) Run a torture test a few hours 2) Assign only double-checks for a duration of 1 to 2 months 3) Compare results of these double-checks to George's list of residues (updated once a week) 4) If proven reliable, switch to LL Step 2) is in fact equivalent to a torture test but has two advantages : - productive ! - very long , 1 to 2 months instead of a few hours. Do not be too agressive with overclocking, just one single bad result will totally destroy the benefit of overclockling.