mersenneforum.org  

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

Reply
 
Thread Tools
Old 2012-03-13, 21:23   #34
lycorn
 
lycorn's Avatar
 
Sep 2002
Oeiras, Portugal

2×683 Posts
Default

From 04-02-2008:

51000049,S471375,C65FFF991,Wc1,00000000
51036281,lmurray,largetest,WZ1,00000000
51072661,S471375,C65FFF991,Wc1,00000000
56500057,ThomRuley,Kyle,WZ2,00000000
56500099,ThomRuley,Kyle,WZ2,00000000
56500159,ThomRuley,Kenny,WZ2,00000000
56515961,faniz,gauss,WZ2,00000000
59999999,kollegi,reti,WZ1,00000000
61466891,dyballa,,Wc1,00000000
61666117,dyballa,,Wc1,00000000
61666141,dyballa,,Wc1,00000000
61666691,dyballa,,Wc1,00000000
61666697,dyballa,,Wc1,00000000
61666729,dyballa,,Wc1,00000000
61666981,dyballa,,Wc1,00000000
61668689,dyballa,,Wc1,00000000
61668697,dyballa,,Wc1,00000000
61696627,dyballa,,Wc1,00000000
61999997,dyballa,,Wc1,00000000
67239937,S16447,FAA,Wc1,00000000
69206017,S16447,SB1,Wc5,00000000
70022021,Spherical_Cow,NormNew,WZ1,00000000
77900461,Team_Prime_Rib,DSheets_60,WZ1,0B003300
77900497,Team_Prime_Rib,PM_node_10,WZ2,09000900
77909869,Team_Italia,Paperino,WZ1,0B014103
77909939,hsoftdev17,Geek2,WY1,10001000
79299821,Team_Prime_Rib,DSheets_02,WZ1,00000100
79299959,Net_Force,QA1,WZ1,00000000
86061587,vhf,VHF,Wc2,00000000

Not much difference, except some 61M stuff...

Oooops, not even that. I´d missed ATH´s last post.

Last fiddled with by lycorn on 2012-03-13 at 21:24 Reason: Update
lycorn is offline   Reply With Quote
Old 2012-03-13, 21:38   #35
aketilander
 
aketilander's Avatar
 
"Åke Tilander"
Apr 2011
Sandviken, Sweden

23616 Posts
Smile

Quote:
Originally Posted by ATH View Post
Here are the last lines above 50M from some HRF3.txt:
Thank you ATH and thank you lycorn. The data confirms so far the list, which is very good, and gives us the information that the LL-test for M79299959 was done before 2005 January 10th.
aketilander is offline   Reply With Quote
Old 2012-03-13, 23:41   #36
davieddy
 
davieddy's Avatar
 
"Lucan"
Dec 2006
England

2×3×13×83 Posts
Default

Quote:
Originally Posted by axn View Post
There is an even easier method to get the answer: "No". ECC is a luxury enjoyed by servers (and that too, not all servers). From GIMPS context, "No" is the only answer that makes sense.
I used to work on this stuff. European Community? Hoffman? Huffnung?

Jeez this juice is fine.

Lucan

Last fiddled with by davieddy on 2012-03-13 at 23:48 Reason: Help me out, Paul
davieddy is offline   Reply With Quote
Old 2012-03-14, 00:02   #37
cheesehead
 
cheesehead's Avatar
 
"Richard B. Woods"
Aug 2002
Wisconsin USA

22×3×641 Posts
Default OT

Quote:
Originally Posted by davieddy View Post
I used to work on this stuff. European Community?


- - -

For the bystanders:

The European Economic Community was EEC rather than ECC, of course.

- - -

Quote:
Hoffman? Huffnung?
http://monologues.co.uk/004/Bricklayers_Story.htm

Last fiddled with by cheesehead on 2012-03-14 at 00:11
cheesehead is offline   Reply With Quote
Old 2012-03-14, 02:28   #38
kladner
 
kladner's Avatar
 
"Kieren, ktony"
Jul 2011

3×3,167 Posts
Default

Quote:
Originally Posted by aketilander View Post
Well I am not using ECC for my 100M test because ECC is expensive and the processors which can use ECC are more expensive and slower (and don't use AVX) then the fastest processors on the market AND you will not be sure that you have reached an accurate result.
I went and looked up prices for ECC RAM, and I was surprised at how affordable it actually is. In addition, my Asus board and AMD CPU actually support ECC. If I had been thinking of P95 when I ordered these components, I would have seriously considered ECC.
kladner is offline   Reply With Quote
Old 2012-03-14, 05:43   #39
Batalov
 
Batalov's Avatar
 
"Serge"
Mar 2008
Phi(3,3^1118781+1)/3

3×5×599 Posts
Default

Where were you in 2009? :-)
Batalov is offline   Reply With Quote
Old 2012-03-14, 11:03   #40
aketilander
 
aketilander's Avatar
 
"Åke Tilander"
Apr 2011
Sandviken, Sweden

2×283 Posts
Default

Quote:
Originally Posted by kladner View Post
I went and looked up prices for ECC RAM, and I was surprised at how affordable it actually is. In addition, my Asus board and AMD CPU actually support ECC. If I had been thinking of P95 when I ordered these components, I would have seriously considered ECC.
If you use Intel processors a good idea is to check http://ark.intel.com before you buy ECC memory. They clearly state if ECC is supported or not. Unfortunately the newest and fastest Intel processors don't and the ones supporting AVX don't, I think. Xeon processors usually do. Even though your computer may work with ECC the error correction will not be used if the processor does not support it I think.
aketilander is offline   Reply With Quote
Old 2012-03-14, 11:22   #41
sdbardwick
 
sdbardwick's Avatar
 
Aug 2002
North San Diego County

10100101002 Posts
Default

Xeon E5-2620 supports AVX and ECC.
The E7 series might not support AVX.
IIRC, all Xeons with integrated memory controllers support ECC.

Last fiddled with by sdbardwick on 2012-03-14 at 11:23
sdbardwick is online now   Reply With Quote
Old 2012-03-20, 14:13   #42
Bdot
 
Bdot's Avatar
 
Nov 2010
Germany

3·199 Posts
Default

Quote:
Originally Posted by retina
So then if there is no ECC and the user selects any LL assignment that is expected to take more than one month to complete then very strongly discourage such a test on the basis that it would most probably end up with a useless result. If the user chooses to override the suggestion and runs anyway then so be it and let them deal with whatever happens.
I have submitted over 20 THzDays of LL tests, never any error. Only a small fraction of that was run on a machine with ECC memory. I'm sure many other people have similar error-free runs. Still the discussion here sounds as if that was highly unlikely to happen.

Had I used all my CPU power to finish some M800 exponent (given mprime supported that size), it would have been a similar computing effort. Are you saying the error probability depends on if it is all in one big LL-test or many small ones? Of course I see the difference how much work a single error would render useless. But I'd argue that the average error rate is quite meaningless if you don't know its distribution. (I hope I picked the right term - I'm referring to the case that some machines may tend to produce way more errors than others, and by doing so would also turn in many bad results when running shorter tests.)

Another thought: I see this discussion misses the possibility that memory is working well, but the error is in the CPU. Badly overclocked CPUs, or CPUs running at some limit otherwise may also produce "random" errors, why not bad LL results? Are you willing to restrict longer LL-tests to stock-clocked CPU's?
Bdot is offline   Reply With Quote
Old 2012-03-20, 14:34   #43
kladner
 
kladner's Avatar
 
"Kieren, ktony"
Jul 2011

3·3,167 Posts
Default

Quote:
Originally Posted by Batalov View Post
Where were you in 2009? :-)
Not here, or I might have known better regarding ECC.
kladner is offline   Reply With Quote
Old 2012-03-20, 18:03   #44
retina
Undefined
 
retina's Avatar
 
"The unspeakable one"
Jun 2006
My evil lair

7×757 Posts
Default

Quote:
Originally Posted by Bdot View Post
I have submitted over 20 THzDays of LL tests, never any error. Only a small fraction of that was run on a machine with ECC memory. I'm sure many other people have similar error-free runs. Still the discussion here sounds as if that was highly unlikely to happen.
Yep, So have I, but this says nothing about running a larger exponent with no error. Larger exponents use more memory and hence are more susceptible to problems.
Quote:
Originally Posted by Bdot View Post
Had I used all my CPU power to finish some M800 exponent (given mprime supported that size), it would have been a similar computing effort. Are you saying the error probability depends on if it is all in one big LL-test or many small ones? Of course I see the difference how much work a single error would render useless. But I'd argue that the average error rate is quite meaningless if you don't know its distribution. (I hope I picked the right term - I'm referring to the case that some machines may tend to produce way more errors than others, and by doing so would also turn in many bad results when running shorter tests.)
Running many smaller LLs is less likely to experience an error than running one large LL due to lessor memory usage. If there is an error then for smaller LLs the impact less and the recovery much easier.
Quote:
Originally Posted by Bdot View Post
Another thought: I see this discussion misses the possibility that memory is working well, but the error is in the CPU. Badly overclocked CPUs, or CPUs running at some limit otherwise may also produce "random" errors, why not bad LL results? Are you willing to restrict longer LL-tests to stock-clocked CPU's?
I've never been a fan of overclocking. Penny wise pound foolish. To impose such a restriction would be foolhardy though, since there is no way to reliably detect it and it would also be inappropriate to tell users how to run their systems.

edit: But the most probable cause of error is the memory. Once you solve that then most of the errors would disappear. Worrying about the CPU is a minor problem in comparison.

Last fiddled with by retina on 2012-03-20 at 18:06
retina is online now   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Number of LLR tests completed so far? ellipse Twin Prime Search 26 2019-09-28 17:19
Largest number of LL tests before matching residues achieved? sdbardwick Lounge 1 2015-02-03 15:03
Completed 29M work not showing as completed in GPU72 Chuck GPU to 72 2 2013-02-02 03:25
Largest LL Test Ever Completed jinydu Lounge 40 2010-03-22 20:54
need Pentium 4s for 5th largest prime search (largest proth) wfgarnett3 Lounge 7 2002-11-25 06:34

All times are UTC. The time now is 21:50.

Sun Mar 29 21:50:31 UTC 2020 up 4 days, 19:23, 2 users, load averages: 1.31, 1.33, 1.40

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.