Mersenne Digest Sunday, April 9 2000 Volume 01 : Number 716 ---------------------------------------------------------------------- Date: Fri, 7 Apr 2000 19:00:30 -0400 From: "Ethan O'Connor" <[EMAIL PROTECTED]> Subject: RE: Mersenne: Prime 95 Speeds I worked that excel sheet up using the information available at http://www.mersenne.org/bench.htm; the .xls file is available at http://web.mit.edu/zudark/www/MersenneSpeeds.xls I haven't updated it with the newest numbers from mersenne.org, but hopefully it's still useful to you. - -Ethan O'Connor - -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On Behalf Of [EMAIL PROTECTED] Sent: Monday, March 27, 2000 1:20 PM To: [EMAIL PROTECTED] Subject: Mersenne: Prime 95 Speeds Recently - someone posted an excel sheet comparing speeds of processors running Prime95 Is this spreadsheet still available? Lawrence...... - --------------------------- _________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Fri, 7 Apr 2000 18:46:04 -0700 From: "John R Pierce" <[EMAIL PROTECTED]> Subject: Mersenne: mprime on a Dell PowerEdge Ugh. I think I mentioned I had a new PowerEdge 4400 server in my labs with dual 600MHz P3-XEON coppermines running on a ServerWorks chipset that has dual PC133 SDRAM busses? Well, I fired up mprime on it today, and running two concurrent instances, have gotten several ERROR: SUMOUT things, each time both processes checked out within a few minutes of each other. The CPUs are running cold as can be (that thing has massive cooling). I left it running a single instance to see if its a SMP issue... - -jrp _________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Fri, 07 Apr 2000 19:20:40 -0700 From: Stefan Struiker <[EMAIL PROTECTED]> Subject: Re: Mersenne: mprime on a Dell PowerEdge Fascinating (for me). How much memory do you have, and what is each XEON cache size? And (pleading ignorance) what OS are you running? Best Regards, Stefanovic John R Pierce wrote: > Ugh. I think I mentioned I had a new PowerEdge 4400 server in my labs with > dual 600MHz P3-XEON coppermines running on a ServerWorks chipset that has > dual PC133 SDRAM busses? Well, I fired up mprime on it today, and running > two concurrent instances, have gotten several ERROR: SUMOUT things, each > time both processes checked out within a few minutes of each other. The > CPUs are running cold as can be (that thing has massive cooling). I left > it running a single instance to see if its a SMP issue... > > -jrp > > _________________________________________________________________ > Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm > Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers _________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Fri, 07 Apr 2000 22:38:33 EDT From: "Nathan Russell" <[EMAIL PROTECTED]> Subject: Mersenne: ECM time needed? How long will each ECM curve on M727 take? I'd like to run a few in May, when I'm done with my current work, but don't know how many to set up. To put it another way, how many curves will take about a week on a p3-600 running 16/7? I checked the various FAQs, but couldn't find this information anywhere. Thanks, Nathan Russell, unofficial GIMPS pet newbie :-) ______________________________________________________ Get Your Private, Free Email at http://www.hotmail.com _________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Fri, 7 Apr 2000 20:52:37 -0700 From: "John R Pierce" <[EMAIL PROTECTED]> Subject: Re: Mersenne: mprime on a Dell PowerEdge Stefanovic asks... > Fascinating (for me). How much memory do you have, and what is each > XEON cache size? And (pleading ignorance) what OS are you running? 1.2GB (!), 256k cache per ( :( the 0.18u coppermine xeons are only available this way), and RedHat Linux 6.1, special Dell SBE version. George says... > What a shame. It could be a memory problem instead yeah, ugh. Well, the system came with 256MB of PC133 ECC registered SDRAM (2 * 128MB), I've fortified it with an additional 1GB of PC133 ECC registered SDRAM (4 * 256MB). I would *hope* any memory error would trip a ECC error. Tomorrow, I might pull the 1GB out and leave just the original 256MB to see what happens... Lets see whether a single instance runs all night... If it does, then I'm gonna have to suspect a cache coherency problem which is likely architectural, ugh. This machine uses the ServerWorks III LE chipset (formerly Reliance Computer Corporation, they make high end SMP server chipsets for Compaq, Dell, HP, SuperMicro and others). - -jrp _________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Fri, 07 Apr 2000 21:52:10 -0700 From: Stefan Struiker <[EMAIL PROTECTED]> Subject: Re: Mersenne: mprime on a Dell PowerEdge If I had added the 1 Gig, and it wouldn't play I'd pull it. Sounds so gorgeous otherwise, it's just got to fly...arghhh. John R Pierce wrote: > Stefanovic asks... > > Fascinating (for me). How much memory do you have, and what is each > > XEON cache size? And (pleading ignorance) what OS are you running? > > 1.2GB (!), 256k cache per ( :( the 0.18u coppermine xeons are only available > this way), and RedHat Linux 6.1, special Dell SBE version. > > George says... > > > What a shame. It could be a memory problem instead > > yeah, ugh. Well, the system came with 256MB of PC133 ECC registered SDRAM > (2 * 128MB), I've fortified it with an additional 1GB of PC133 ECC > registered SDRAM (4 * 256MB). I would *hope* any memory error would trip a > ECC error. Tomorrow, I might pull the 1GB out and leave just the original > 256MB to see what happens... > > Lets see whether a single instance runs all night... If it does, then I'm > gonna have to suspect a cache coherency problem which is likely > architectural, ugh. This machine uses the ServerWorks III LE chipset > (formerly Reliance Computer Corporation, they make high end SMP server > chipsets for Compaq, Dell, HP, SuperMicro and others). > > -jrp > > _________________________________________________________________ > Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm > Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers _________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Fri, 7 Apr 2000 23:18:41 -0600 From: "Aaron Blosser" <[EMAIL PROTECTED]> Subject: RE: Mersenne: mprime on a Dell PowerEdge >yeah, ugh. Well, the system came with 256MB of PC133 ECC registered SDRAM >(2 * 128MB), I've fortified it with an additional 1GB of PC133 ECC >registered SDRAM (4 * 256MB). I would *hope* any memory error would trip a >ECC error. Tomorrow, I might pull the 1GB out and leave just the original >256MB to see what happens... > >Lets see whether a single instance runs all night... If it does, then I'm >gonna have to suspect a cache coherency problem which is likely >architectural, ugh. This machine uses the ServerWorks III LE chipset >(formerly Reliance Computer Corporation, they make high end SMP server >chipsets for Compaq, Dell, HP, SuperMicro and others). Well, that'd be a shame if it was architectural. I had a chance to test out a Poweredge last year...just a quad PIII Xeon though, before the new chips were out. It performed admirably under NT and running 4 instances of NTPrime. I must say though, the Compaq 4 way "blew" it away (well, by a few percentage points) when running multiple instances on each CPU. I recently got some info on Compaq's new Proliant servers that'll be using the new CPU's... I'm under NDA, so I'm not 100% sure what info is already public and what isn't. :( I'm sure I can at least say that the new machines use the latest Xeon's, run at 133MHz FSB, use Compaq's own architecture, as always. They're avoiding the Intel chipset, and the Compaq rep had some funny stuff to say about Intel's chipsets... They have the latest 64 bit/66 MHz PCI slots, etc. etc. etc. Looking really good. I wouldn't mind getting my hands on one of the 8-way Proliant 8000/8500 boxes. With 4 CPU's and about 1GB or so, they're a steal at about $42,000...probably less through the other vendors. Let's see...what else...they'll support a lot more memory, 8 or 16GB I seem to recall... Of course, you'd need Windows 2000 to support that much. Anyway, expected release dates for those nice little boxes are from this month out to June for all the different models. Basically, they're revamping the entire Proliant (and workstation) lines to include support for 133MHz FSB and all the new chips. And what with the 820 and 840 chipsets being somewhat lousy, it's nice having Compaq's own stuff on there, so you get real SDRAM support and all the other goodies. Aaron _________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sat, 8 Apr 2000 03:17:09 -0700 From: Russel Brooks <[EMAIL PROTECTED]> Subject: Re: Mersenne: mprime on a Dell PowerEdge John R Pierce wrote: > yeah, ugh. Well, the system came with 256MB of PC133 ECC registered SDRAM > (2 * 128MB), I've fortified it with an additional 1GB of PC133 ECC > registered SDRAM (4 * 256MB). I would *hope* any memory error would trip a > ECC error. I thought ECC memory was supposed to Correct any errors and continue, not just detect them. cheers... Russ _________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sat, 8 Apr 2000 13:03:48 +0200 From: "Hoogendoorn, Sander" <[EMAIL PROTECTED]> Subject: RE: Mersenne: Prime 95 Speeds >I worked that excel sheet up using the information available at >http://www.mersenne.org/bench.htm; Goerge, There still isn't a link to this page from youre pages. The current link still points to the old benchmark page _________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sat, 8 Apr 2000 07:00:30 -0600 From: "Aaron Blosser" <[EMAIL PROTECTED]> Subject: RE: Mersenne: mprime on a Dell PowerEdge >John R Pierce wrote: >> yeah, ugh. Well, the system came with 256MB of PC133 ECC registered SDRAM >> (2 * 128MB), I've fortified it with an additional 1GB of PC133 ECC >> registered SDRAM (4 * 256MB). I would *hope* any memory error would trip a >> ECC error. >I thought ECC memory was supposed to Correct any errors and continue, >not just detect them. That's the idea. ECC can correct single bit errors and detect (but not correct) 2 bit errors. Compaq's memory design on their serers has ECC-2 which corrects 2 bit errors and can detect (but not correct) 4 bit errors. I don't know how common ECC-2 is among server designs from other companies... Typically, such things are considered of less importance on a workstation, which is also why you don't find RAID (or even SCSI) on many workstations either... Aaron _________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sat, 8 Apr 2000 16:42:37 -0000 From: "Brian J. Beesley" <[EMAIL PROTECTED]> Subject: RE: Mersenne: mprime on a Dell PowerEdge On 8 Apr 00, at 7:00, Aaron Blosser wrote: > >I thought ECC memory was supposed to Correct any errors and continue, not > >just detect them. > > That's the idea. ECC can correct single bit errors and detect (but not > correct) 2 bit errors. And hopefully log the occurrence of any memory faults i.e. when a bit correction needed to be made. This gives you a good idea about how close to the wind you're sailing. One bit corrected per month is probably acceptable. > > Typically, such things are considered of less importance on a workstation, > which is also why you don't find RAID (or even SCSI) on many workstations > either... ... but price is a bigger issue. Most home users don't drive a system hard enough for the ultimate in speed to be a problem (RAID is most commonly used in conjunction with stripe sets to increase disk I/O bandwidth); as for reliability, home users, and purchasers of desktop systems for business, blame any hangs & crashes on the OS anyway, so why bother building reliable hardware when you can sell crap for $20 less? Regards Brian Beesley _________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sat, 08 Apr 2000 11:10:40 -0700 From: Eric Hahn <[EMAIL PROTECTED]> Subject: Mersenne: M727 has a factor?!?!? Hi!! Tell me I'm wrong... and if not, what happened?? I just made a slight error in adding a P-1 factor assignment to the WORKTODO.INI file for M727 and came up with the following result (on screen): P-1 on P727 with B1=30, B2=1 P727 stage 1 complete. 116 transforms. Time: 0.018 sec. (4659194 clocks) Stage 1 GCD complete. Time: 0.001 sec. (164887 clocks) P727 has a factor: 11633 and in the RESULTS.TXT file: [Sat Apr 08 10:43:37 2000] P-1 found a factor in stage #1, B1=30, B2=1. UID: Net_Force/V20, P727 has a factor: 11633 This meets all the criteria too.... 1) 11633 is PRIME. 2) 2kp+1 = 2*(8)*727+1 = 11633 3) 8n+1 = 8*(1454)+1 = 11633 4) 2^p (mod n) = 2^727 (mod 11633) = 1 Eric P.S. the error in question was: Pminus1=727,1E16,0,0,0 _________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sat, 8 Apr 2000 12:18:18 -0700 From: Will Edgington <[EMAIL PROTECTED]> Subject: Mersenne: M727 has a factor?!?!? P-1 on P727 with B1=30, B2=1 P727 stage 1 complete. 116 transforms. Time: 0.018 sec. (4659194 clocks) Stage 1 GCD complete. Time: 0.001 sec. (164887 clocks) P727 has a factor: 11633 This meets all the criteria too.... 1) 11633 is PRIME. 2) 2kp+1 = 2*(8)*727+1 = 11633 3) 8n+1 = 8*(1454)+1 = 11633 4) 2^p (mod n) = 2^727 (mod 11633) = 1 11633 divides M1454 where 1454 = 2*727, but 11633 does not divide M727. Your #4 calculation has a bug, probably a rounding error; the correct result is 11631. In fact: M( 1454 )C: 11633 M( 1454 )C: 52068472442119144511578580563 M( 1454 )C: 59803996769241650545074361210286131 M( 1454 )D That is, M1454 is considered to be completely factored even though it is a multiple of M727, which is known to be composite but has no known prime factors. There are other cases like this in the data. >From my "reverse method" program, I should now have all factors less than about 1.6 billion for _any_ Mersenne number with an exponent less than 2^30 (just over a billion). Will http://www.garlic.com/~wedgingt/mersenne.html _________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sat, 8 Apr 2000 16:26:02 -0400 (EDT) From: "David A. Miller" <[EMAIL PROTECTED]> Subject: Re: Mersenne: M727 has a factor?!?!? > >Hi!! > > Tell me I'm wrong... and if not, what happened?? >P727 has a factor: 11633 Look carefully: that's P727, which denotes 2^727+1. You must have accidentally specified "factor 2^n+1" when you started the work. This small factor was reported because there is no entry in LOWP.TXT for P727; it was removed after P727 was finished last year--by me, incidentally. Alas, P727's evil twin M727 remains unfactored. David A. Miller [EMAIL PROTECTED] _________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sat, 08 Apr 2000 15:20:11 -0700 From: Eric Hahn <[EMAIL PROTECTED]> Subject: Re: Mersenne: M727 has a factor?!?!? Will Edgington wrote: > > P-1 on P727 with B1=30, B2=1 > P727 stage 1 complete. 116 transforms. Time: 0.018 sec. >(4659194 clocks) > Stage 1 GCD complete. Time: 0.001 sec. (164887 clocks) > P727 has a factor: 11633 > > This meets all the criteria too.... > 1) 11633 is PRIME. > 2) 2kp+1 = 2*(8)*727+1 = 11633 > 3) 8n+1 = 8*(1454)+1 = 11633 > 4) 2^p (mod n) = 2^727 (mod 11633) = 1 > >11633 divides M1454 where 1454 = 2*727, but 11633 does not >divide M727. Your #4 calculation has a bug, probably a >rounding error; the correct result is 11631. Well, I went back and did it by hand!! You're right about #4... BTW, George wrote that what I got was a result of a parsing error on the part of Prime95 (it did 2^727+1, not 2^727-1). R = 1 727 = 1011010111 E=727 D=1 R= 2 A= 4 E=363 D=1 R= 8 A= 16 E=181 D=1 R= 128 A= 256 E= 90 D=0 R= 128 A= 7371 E= 45 D=1 R= 1215 A= 5531 E= 22 D=0 R= 1215 A= 8804 E= 11 D=1 R= 6133 A=11370 E= 5 D=1 R= 4008 A=11004 E= 2 D=0 R= 4008 A= 119 E= 1 D=1 R=11632 A= 2528 E= 0 *11632* Eric _________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sun, 09 Apr 2000 15:00:04 EDT From: "Nathan Russell" <[EMAIL PROTECTED]> Subject: Mersenne: Another ECM question To start with, thanks to all those who answered my first question. What I want to know is as follows: If no factors are found for an exponent after running the needed curves for a 55 digit factor, what is done? Are additional curves run, or is that exponent put aside to await improvements in software? Thanks in advance Nathan ______________________________________________________ Get Your Private, Free Email at http://www.hotmail.com _________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sun, 9 Apr 2000 23:44:19 +0200 (MET DST) From: [EMAIL PROTECTED] Subject: Re: Mersenne: Another ECM question > Nathan Russell asks > To start with, thanks to all those who answered my first question. > > What I want to know is as follows: If no factors are found for an exponent > after running the needed curves for a 55 digit factor, what is done? Are > additional curves run, or is that exponent put aside to await improvements > in software? The exponent would probably be put aside unless another algorithm is known for completing the factorization. Nathan's earlier question was about M727, the first Mp with no known prime factors. I anticipate somebody (perhaps NFSNET) will finish this by ECM or SNFS within two years. Peter Montgomery _________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sun, 09 Apr 2000 20:51:54 -0700 From: Stefan Struiker <[EMAIL PROTECTED]> Subject: Mersenne: Clockers Curiouser and curiouser: Working on nearly the same exponent, 9.7 million, I notice that our PIII 500 takes only 119.5 million clocks per iteration, whereas our 733EB requires 124.8 million to do the same. The 500 has 512K 1/2 speed off-chip cache and runs 128MB of SDRAM, with a 100MHz bus and was cheap. The 733 has 256K full-speed on-chip cache and runs 256MB of RDRAM (RAMBUS), 133MHz bus and was not so cheap. An Athlon of my acquaintance does all of this in 114.6 million clocks, but that's another story, because the Athlon is supposed to decode more instructions per cycle, and generally ace it in the floating-point department.. Can anyone explain the Pentium discrepancy? Best Regards, Stefanovic _________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ End of Mersenne Digest V1 #716 ******************************
