Mersenne Digest Wednesday, August 22 2001 Volume 01 : Number 877 ---------------------------------------------------------------------- Date: Fri, 17 Aug 2001 06:34:23 -0000 From: bjb@bbhvig.uklinux.net Subject: Re: Mersenne: 10M digit prime testing question.... On 16 Aug 2001, at 23:58, Jeffrey Euclide wrote: > Just curious of something here, since I've upgraded to a faster > machine, 1.333GHz(1.47 oc'd), 512M DDR on a win2k platform if it would > be prudent to take on factoring 10M digit primes, thanks ahead.... Prudent? In what respect? You're much more likely to find a prime by sticking to "ordinary" LL tests - but obviously much less likely to win $100K. Even with that power, it's still going to take ~3 months per exponent to test 10M digit numbers. So it may be more "fun" to stick to "ordinary" LL tests. Personally I'd run double-checks for a few weeks & check that the results are working into the lucas_v database. Overclocking Athlons is not easy, they overheat very readily when the FPU is driven hard (which Prime95 does). The fact that the system boots clean at 1.47 GHz (presumably 11 x 133 MHz), or even passes Memt25 at that speed, does not mean that Prime95 will be reliable with that much overclocking. You can improve the reliability of Prime95 by running with roundoff checking enabled (Advanced menu) but this will cost about 15% in run speed (for Prime95 only). If you see no errors at all over a year or so, you're probably OK. However bear in mind that hardware glitches (often due to overclocking or overheating) can still creep through even with roundoff checking enabled. If you find that Prime95 _is_ reliable at that speed, please tell me which CPU cooler unit you're using, and whether you can live with the fan noise - I've found that most of the better CPU coolers are rather loud :( 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, 18 Aug 2001 06:18:42 -0700 From: Gerry Snyder Subject: Mersenne: How fast is ECM? I have decided (with prompting from someone with much more experience) to switch one of the CPU's in my dual processor system away from LL testing, because of the slowdown (~30%) when both are running LL. While finding factors the "normal" way is fun, factoring difficult small Mersenne numbers (with ECM) appeals to me even more. I have not been able to find any info on how long it takes, though. I've searched the last 2 1/2 years of the email list, have looked through the Primenet web site, and have even read the Prime95 v 21 help file. I did not find a thing on timing. I'll be glad to RTFM if you can point me to it, and would welcome any timing data from personal experience (with an indication on how things scale with exponent and B1, if possible). If I don't get any answers, I'll find out in a month when the program gets to the ECM entry in the to-do list (currently 100 curves for M751 with B1=44,000,000), but I have no idea whether that will take hours or months on a 1 GHz P3. Since mprime now uses intermediate files for ECM IIRC, I don't need to worry about crashes or other system shutdowns, and my goal is to find chunks of ECM work that take a couple weeks to a month. GIMPS is good! Thank you all, Gerry - -- mailto:gerrysnyder@mediaone.net my work: helping generate data for: http://galileo.jpl.nasa.gov/ _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sat, 18 Aug 2001 10:58:08 -0400 From: George Woltman Subject: Re: Mersenne: How fast is ECM? Hi, At 06:18 AM 8/18/2001 -0700, Gerry Snyder wrote: >If I don't get any answers, I'll find out in a month when the program >gets to the ECM entry in the to-do list (currently 100 curves for M751 >with B1=44,000,000), but I have no idea whether that will take hours or >months on a 1 GHz P3. If I recall correctly, you can roughly expect one hour per curve. Also as a very rough guide if you double the exponent, it will take somewhat more than twice the time. >Since mprime now uses intermediate files for ECM >IIRC, I don't need to worry about crashes or other system shutdowns, Correct. Have fun, George _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sat, 18 Aug 2001 12:09:11 -0400 From: George Woltman Subject: Re: Mersenne: What criteria are the cause for work that gets discredited? Hi, At 04:37 PM 8/16/2001 -0400, CARLETON GARRISON wrote: >I had the recent occassion to see a Double Check LL test - find a >factor. On George's top producer page (the one that gets updated less >than daily or even possibly weekly - how often does it get updated?), Roughly weekly. >it says that finding a later factor or an incorrect LL result will cause >loss of credit. Is this up-to-date stat policy? Yes. The top producers page is basically a tabulation of all the LL tests in my double-checked and single-checked databases. These databases do not track proven bad results and factored exponents. Also, it does not track which program created the LL result. Thus, users that use the newer faster programs are given the same credit as GIMPS veterans that used slower versions. Thus, look at the list as a close approximation to your work contributed relative to other member's work contributed. > Does this apply only to LL stats or is it possible to lose credit for > factoring if later factoring proves more successful? How many such > scenarios are there? That table gives no credit for trial factoring or P-1 factoring. Hope that helps, George _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Wed, 22 Aug 2001 22:06:19 +0100 From: Gareth Randall Subject: Mersenne: Utilities for "headless" computing. Greetings, Intro ===== People running computational projects such as GIMPS normally end up running "headless" computers. These are systems that do not have monitors and keyboards, or perhaps they do but it isn't convenient to use them. Some such headless users are fortunate enough to be able to run their systems continuously, but for those with their systems in bedrooms and the like, it is normally inconvenient to keep turning systems on and off. Content ======= Over time I have written several short utilities to enable headless computers to run with the minimum of effort. These are all for Linux, but could be easily converted to other Unix types. I hereby offer them to all who are interested! Descriptions ============ js-shutdown - ----------- This program shuts down your system when a button is pressed on a joystick. Consequently headless machines can now be powered down (as well as up) without requiring any login, either local or remote. A simple push-button switch can be used instead of a joystick. /etc/init.d/mprime.sh - --------------------- Start and stop mprime automatically. Includes automatic log file redirection. clock-sync - ---------- A simple program to time-step badly drifting clocks on bootup. This script can be run before logging has started. Why use this instead of the more professional NTP daemon? Because otherwise your logs will always begin with the wrong time, and only be corrected later. This program uses expect and rdate. status monitors - --------------- Simple ways of remotely monitoring the progress of calculation programs. These involve simple additions to existing files. Download ======== ftp://lettuce.edsc.ulst.ac.uk/gareth/distrib-utils.tar.gz Thanks to Brian Beesley for generously providing the hosting service. Yours, ======= Gareth Randall ======= _________________________________________________________________________ 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 #877 ******************************