mersenneforum.org  

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

Reply
 
Thread Tools
Old 2020-03-20, 07:18   #1
jas
 
"Simon Josefsson"
Jan 2020
Stockholm

100102 Posts
Default My DC mismatched - how to doublecheck?

Hi!



I just completed a double check that mismatched:


https://www.mersenne.org/report_expo...0866657&full=1


The machine has been quite solid and has ECC memory, but maybe hard drives got corrupted? Or the earlier check was invalid.


How do I re-test this on another machine? I have never requested a manual assignment, is that the way to go?



Perhaps someone else could double check this number?


/Simon
jas is offline   Reply With Quote
Old 2020-03-20, 07:45   #2
sdbardwick
 
sdbardwick's Avatar
 
Aug 2002
North San Diego County

23×29 Posts
Default

Wow, that was one of mine from ages ago!

Hmm, the machine that did the first time test was very reliable and not overclocked. Interesting to see which one is correct.

You can't double check your own work, but somebody will pick that exponent up soon.
sdbardwick is online now   Reply With Quote
Old 2020-03-20, 09:36   #3
S485122
 
S485122's Avatar
 
Sep 2006
Brussels, Belgium

3·5·101 Posts
Default

There is a thread for requesting triple checks : Strategic double and triple checks (PRP's and P-1's too).

It is true that it is difficult to find :

mersenneforum.org > Great Internet Mersenne Prime Search > Data > Marin's Mersenne-aries > Strategic double and triple checks (PRP's and P-1's too)

One level up would be more logical, but its place in the folder hierarchy is historical.

Jacob
S485122 is offline   Reply With Quote
Old 2020-03-20, 11:35   #4
ATH
Einyen
 
ATH's Avatar
 
Dec 2003
Denmark

22×709 Posts
Default

I started a triple check: https://mersenne.org/M50866657
ATH is offline   Reply With Quote
Old 2020-03-20, 13:21   #5
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

23·167 Posts
Default

Quote:
Originally Posted by jas View Post
Hi!
I just completed a double check that mismatched:
It happens. About 2% of LL tests are bad. Do another on a different exponent and see how it goes. Or do a PRP with GEC and see how many errors it detects. Some hardware can be very reliable on ordinary computing and yet not reliable enough for Mersenne primality testing. The well tuned code of a good primality test software puts extraordinary stress on the hardware. So much so that chip manufacturers use them for chip testing.
kriesel is online now   Reply With Quote
Old 2020-03-20, 13:32   #6
jas
 
"Simon Josefsson"
Jan 2020
Stockholm

228 Posts
Default

Quote:
Originally Posted by ATH View Post
I started a triple check: https://mersenne.org/M50866657

Yay, thank you!



The disk is non-RAID ext4 so disk corruption may happen, however the program was never halted during the test of that number so I'm not sure how relevant disk corruption could be.


/Simon
jas is offline   Reply With Quote
Old 2020-03-20, 13:42   #7
jas
 
"Simon Josefsson"
Jan 2020
Stockholm

2·32 Posts
Default

Quote:
Originally Posted by kriesel View Post
It happens. About 2% of LL tests are bad. Do another on a different exponent and see how it goes. Or do a PRP with GEC and see how many errors it detects. Some hardware can be very reliable on ordinary computing and yet not reliable enough for Mersenne primality testing. The well tuned code of a good primality test software puts extraordinary stress on the hardware. So much so that chip manufacturers use them for chip testing.

The machine has completed 12 tests before -- 2 DC, 4 LL, 4 PRP -- since I started it on february 12th. It is a Dell R620 server (not overclocked) with 2xE5-2650v1, hosted at a co-location facility with good cooling. I monitor temperature (CPU, IPMI etc) and it looks stable.


I bought the machine cheap on ebay though :) I never visually inspected the CPUs, but 'cpuid' says 'GenuineIntel' -- is that guarantee enough that it isn't a ES/QS CPU?


Let's see what the triple check says...



/Simon
jas is offline   Reply With Quote
Old 2020-03-20, 14:46   #8
kuratkull
 
kuratkull's Avatar
 
Mar 2007
Estonia

22·3·11 Posts
Default

Hi, just to chip in about CPU issues. The GenuineIntel doesn't matter, all CPU's are genuine. Nobody makes "counterfeit Intel CPUs", well not working ones at least. To know if it's an ES you can look at the markings on the CPU heatspreader. Though I would assume any utility capable of giving out CPU details would work also. I don't know what cat /proc/cpuinfo would say about an ES. But most likely you have a functional CPU, it just might not like the workload:


I have a Ryzen CPU running LLR that crashed with BSOD before a BIOS update. Now the same CPU just overheats the motherboard VRM's so I have to clock down the CPU. And I have another Intel CPU that gave me bad LLR residues - occasionally. I detected the error when it started to tell me that it found 10 primes almost sequentially. Of course doublechecking on another CPU found no primes. I ran LLR for Riesel primes, but the tightly optimized CPU instructions are mostly the same - they are both based on the gwnum library. Both CPUs work completely normally and without crashes otherwise.

Last fiddled with by kuratkull on 2020-03-20 at 14:56
kuratkull is offline   Reply With Quote
Old 2020-03-22, 14:43   #9
LaurV
Romulan Interpreter
 
LaurV's Avatar
 
Jun 2011
Thailand

100001011001002 Posts
Default

Quote:
Originally Posted by kuratkull View Post
Nobody makes "counterfeit Intel CPUs"
You seems to be pretty sure about that
We may have posted here long time ago a story from the time we were working in the far east, and not far away from our factory was an obscure 4-storied building, looking more like a villa, about which we never knew what they do, until once, when their "guilao customers" came, tested the "products", chose the best xx% of them, and scrapped the others, by burning them in a big oven. There, local people have a very good entrepreneurial skills, we have no idea what was burning in the ovens, but the following weeks they sold us 80486 processors for a very cheap price, which we used for long time, or resold back home. On them, it was printed something about genuine intel parts being made in USA....
LaurV is offline   Reply With Quote
Old 2020-03-23, 05:00   #10
ATH
Einyen
 
ATH's Avatar
 
Dec 2003
Denmark

22×709 Posts
Default

Quote:
Originally Posted by jas View Post
Hi!



I just completed a double check that mismatched:


https://www.mersenne.org/report_expo...0866657&full=1
Your result is correct: https://mersenne.org/M50866657
ATH is offline   Reply With Quote
Old 2020-03-23, 05:10   #11
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

3×2,281 Posts
Default

These are the closest results (time-wise) from the bad computer:

Code:
51167521	Amy Pond	3570K-blu	74E6AFBCA445EF__		2013-06-11 20:31
51247087	Amy Pond	3570K-blu	080CD993FECE53__		2013-06-24 06:38
51258113	Amy Pond	3570K-blu	22639D22B33C1E__		2013-06-05 09:18
51415153	Amy Pond	3570K-blu	0D29D283D6D99D__		2013-06-08 15:07
In case any of these are unassigned, someone might want to grab them.
Prime95 is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Doublecheck efforts; S66/S79 to start with gd_barnes Conjectures 'R Us 16 2014-08-07 02:11
Doublecheck always have shifted S0 value? ATH PrimeNet 11 2010-06-03 06:38
All things doublecheck!! masser Sierpinski/Riesel Base 5 44 2006-09-24 17:19
DoubleCheck vs LL assignments Unregistered PrimeNet 9 2006-03-26 05:48
doublecheck - results TheJudger Data 4 2005-04-04 08:54

All times are UTC. The time now is 17:19.

Fri Jun 5 17:19:44 UTC 2020 up 72 days, 14:52, 1 user, load averages: 1.32, 1.52, 1.52

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, 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.