Mersenne Digest Tuesday, February 13 2001 Volume 01 : Number 816 ---------------------------------------------------------------------- Date: Sun, 11 Feb 2001 15:02:30 -0600 From: "Richard B. Woods" <[EMAIL PROTECTED]> Subject: Mersenne: FFT sizes in benchmark table headings, please? FFT sizes corresponding to exponent ranges seem not to be listed in the benchmark tables at http://www.mersenne.org/bench.htm and http://www.mersenne.org/overclk.htm. They _are_ listed in the rightmost column of the Search Status table at http://www.mersenne.org/status.htm. Could someone add FFT sizes to the benchmark table headings, please? _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sun, 11 Feb 2001 16:30:59 EST From: [EMAIL PROTECTED] Subject: Mersenne: Re: P4 Optimization George Woltman wrote: >George is working on it, but is a long way from completion. Progress is >slow, primarily due to my own laziness. My estimate for a 512K FFT is 0.4 >seconds on a 1.4GHz P4. You can compare that to other machines at >http://www.mersenne.org/bench.htm. Jeff Woods wrote: >Can we assume you meant 0.04? > >According to that site, A P-III 1.0 Ghz is already clipping such a number >at 0.145, and a P4 is already listed there doing such in the current >version at 0.126.... See, there's afactor of 10 speedup already! :) Regarding the time estimate of 0.04 seconds per iteration at 512K FFT length, under optimal load conditions I've gotten around 0.12 sec/iter at this length with Mlucas running on a 500MHz Alpha 21264 with 4MB L2 cache. This is for compiled high-level source code, so scaling the closck speed up to 1.4GHz, subtracting a few tens of % for the smaller caches the L4 will come with initially, and adding them back for the boost hand-tuned assembly (at least in George's able hands) tends to yield, 0.04 seconds sounds about right for what is achievable. Still, even knowing that the current prototype code is likely quite far from optimal for the P4, a factor of ~3 speedup (0.126 --> 0.04 seconds) will be quite a challenge to realize. One of the drawbacks of doing it by hand in assembler...too bad high-quality HLL compilers (i.e. ones capable of giving 80-90% of the performance of laboriously coded and hand-tuned ASM, for complex, data-nonlocal algorithms requiring lots of data prefetch) appear to be nigh-impossible to write for CISCs like the x86 family. I don't want to start a RISC-versus-CISC flame war here, but the fact is, no high-level FFT-based large-integer-multiply code has gotten within a factor of 2 of the performance of Prime95 on the Pentium. - -Ernst _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sun, 11 Feb 2001 22:29:05 +0100 From: Henk Stokhorst <[EMAIL PROTECTED]> Subject: Mersenne: 39.000.000 range L.S., If anyone is working within the range 39.000.000 - 40.000.000, please contact me to avoid double work, I plan to delve for factors over there. YotN, Henk Stokhorst _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sun, 11 Feb 2001 18:12:12 -0500 From: Nathan Russell <[EMAIL PROTECTED]> Subject: Mersenne: P4 speed and implications thereof Let's assume that the P4 is, as George estimates, capable of doing the 512K FFT at 0.04 iterations per second. In this case, with some back-of-the-envelope calculations, I estimate that a thousand P4s (which might well be on PrimeNet in a year, or a little more - the P4 is approaching the status the Tbird was at a year ago, when I bought my current P3-600) could have done the entire 512K runlength in roughly 138 days, or under 5 months, even without the rest of the machines now contributing. Additionally, the P4 is only going to get faster in the next year or two. This frankly makes me wonder how much longer there will be a place in GIMPS for slower machines. I'm not saying that's a bad thing - after all, 486s and original pentiums were the workforce when GIMPS began, and I wouldn't feel comfortable with the amount of opportunity for error involved in using one for GIMPS now. Any comments from the list? Nathan _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sun, 11 Feb 2001 18:48:21 -0500 From: Jud McCranie <[EMAIL PROTECTED]> Subject: Re: Mersenne: P4 speed and implications thereof At 06:12 PM 2/11/2001 -0500, Nathan Russell wrote: >Additionally, the P4 is only going to get faster in the next year or >two. > >This frankly makes me wonder how much longer there will be a place in >GIMPS for slower machines. I wonder how many people are going to have P4s. Presently, the Athlon is faster for most things, and cheaper. I don't know if the P4 will pull ahead of AMD chips for most things, so will people buy them? +-----------------------------------------------------------+ | Jud McCranie | | | | Think recursively( Think recursively( Think recursively)) | +-----------------------------------------------------------+ _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sun, 11 Feb 2001 19:07:46 -0500 From: Nathan Russell <[EMAIL PROTECTED]> Subject: Re: Mersenne: P4 speed and implications thereof Jud, you wrote: >I wonder how many people are going to have P4s. Presently, the Athlon is >faster for most things, and cheaper. I don't know if the P4 will pull >ahead of AMD chips for most things, so will people buy them? My understanding is that they are designed to perform well for graphical tasks; in my experience, people will buy even very expensive computers if it improves the performance they see when doing graphical tasks; gaming in particular comes to mind. Nathan _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sun, 11 Feb 2001 18:17:50 -0600 From: "Jeramy Ross" <[EMAIL PROTECTED]> Subject: Re: Mersenne: P4 speed and implications thereof Nathan Russell wrote: *snip* > This frankly makes me wonder how much longer there will be a place in > GIMPS for slower machines. I'm not saying that's a bad thing - after > all, 486s and original pentiums were the workforce when GIMPS began, > and I wouldn't feel comfortable with the amount of opportunity for > error involved in using one for GIMPS now. This tends to be a thorn in some people's sides when I have tried to get them to run GIMPS on their machines. It would seem that some view the GIMPS project as an 'elite' minded group that requires the use of the more powerful PC's out there due to it's resource demanding nature. They don't want to dedicate several months to a project and as the size of the exponents grow their usefulness shrinks and eventually they will be deemed no longer useful for the project. This view has only been expressed by a few people that I have talked to, but it maybe an opinion that will grow as the month's pass on. One can see that the usefullness of the slower machines is limited as we march on through Primeland (If for no other reason other than it will take so long to complete a task with the slower machines). This undoubtedly will be a natural evolution of the project, but may leave a bitter taste in the mouths of those who are on slower machines. Who knows? I certainly don't know the impact of this on the project....good or bad.. Jeramy _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sun, 11 Feb 2001 16:21:39 -0800 From: "John R Pierce" <[EMAIL PROTECTED]> Subject: Re: Mersenne: P4 speed and implications thereof > I wonder how many people are going to have P4s. Presently, the Athlon is > faster for most things, and cheaper. I don't know if the P4 will pull > ahead of AMD chips for most things, so will people buy them? the P4 is likely gonna ramp up to 2GHz, 3Ghz and beyond faster and farther than AMD can ramp up the Tbird. - -jrp _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sun, 11 Feb 2001 19:49:52 -0500 From: George Woltman <[EMAIL PROTECTED]> Subject: Re: Mersenne: Re: P4 Optimization Hi, At 04:30 PM 2/11/2001 -0500, [EMAIL PROTECTED] wrote: >One of the drawbacks of doing it >by hand in assembler...too bad high-quality HLL compilers (i.e. ones >capable of giving 80-90% of the performance of laboriously coded and >hand-tuned ASM, for complex, data-nonlocal algorithms requiring lots >of data prefetch) appear to be nigh-impossible to write for CISCs like >the x86 family. One of the biggest challenges is discovering the innermost workings of the P4. The P4 documentation does not detail every penalty one can run into. I create lots of test cases and time them, analyzing the timing difference between two nearly identical code fragments can often lead to insights into the P4 architecture. Today's mystery: I've found a case where adding a NOP instruction speeds up the code by 9%. Not just a small loop, the 9% speedup affects the entire 2nd pass of the FFT! As of now I have no theories or explanations. Time to write some more code fragments to figure it out..... Regards, George _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sun, 11 Feb 2001 19:49:52 -0500 From: George Woltman <[EMAIL PROTECTED]> Subject: Re: Mersenne: Re: P4 Optimization Hi, At 04:30 PM 2/11/2001 -0500, [EMAIL PROTECTED] wrote: >One of the drawbacks of doing it >by hand in assembler...too bad high-quality HLL compilers (i.e. ones >capable of giving 80-90% of the performance of laboriously coded and >hand-tuned ASM, for complex, data-nonlocal algorithms requiring lots >of data prefetch) appear to be nigh-impossible to write for CISCs like >the x86 family. One of the biggest challenges is discovering the innermost workings of the P4. The P4 documentation does not detail every penalty one can run into. I create lots of test cases and time them, analyzing the timing difference between two nearly identical code fragments can often lead to insights into the P4 architecture. Today's mystery: I've found a case where adding a NOP instruction speeds up the code by 9%. Not just a small loop, the 9% speedup affects the entire 2nd pass of the FFT! As of now I have no theories or explanations. Time to write some more code fragments to figure it out..... Regards, George _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Mon, 12 Feb 2001 10:40:23 -0500 From: "Joshua Zelinsky" <[EMAIL PROTECTED]> Subject: Mersenne: Prime 95 disk writes. A very minor suggestion: Why not have the next version of Prime95 have an option for disk writes to occur based on the number of iterations, rather than the time passed? If someone's playing a game or running other heavy CPU using programs for some time, then they may end up backing up very few iterations with the time-based backup. On the other hand, if someone is typing, then they could lose very many iterations with the time based backup. Neither of these problems would come with the disk writes being based on the number of iterations. Regards, Joshua Zelinsky _________________________________________________________________ Get your FREE download of MSN Explorer at http://explorer.msn.com _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Mon, 12 Feb 2001 14:09:46 -0600 From: "Griffith, Shaun" <[EMAIL PROTECTED]> Subject: Mersenne: RE: Mersenne Digest V1 #815 On Thu, 08 Feb 2001 23:09:51 -0500, George wrote: > George is working on it, but is a long way from completion. Progress is > slow, primarily due to my own laziness. George might be overclocked -- he is referring to himself in the 3rd person (indirect register mode?). - -Shaun _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sun, 11 Feb 2001 17:58:40 -0800 From: "John R Pierce" <[EMAIL PROTECTED]> Subject: Re: Mersenne: P4 speed and implications thereof > 486 ?? 25->100MHz, factor of 4 > Pentium 60 MHz -> 200? MHz factor of 3.3 P55c (mmx) hit 233MHz, I believe, for almost 4X > other pentiums ?? P6 architecture (which includes PPro, P2 and P3), 166 - 1GHz+ (factor of 6+) _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Mon, 12 Feb 2001 13:29:01 -0500 From: "Brian Last-Name" <[EMAIL PROTECTED]> Subject: Re: Mersenne: P4 speed and implications thereof >This frankly makes me wonder how much longer there will be a place in >GIMPS for slower machines. I'm not saying that's a bad thing - after >all, 486s and original pentiums were the workforce when GIMPS began, >and I wouldn't feel comfortable with the amount of opportunity for >error involved in using one for GIMPS now. I have been working on this project for about five years now. My first two years with a 486DX266, and now with a Cyrix 266. My computer works fine for now, and I do not need any more power (and I lack 2k-3k to invest in video games) I have worked on trial factoring, and have received little credit on the LL testing page. I hate to think that my faithfulness and determination on this project will be overshadowed by passer-by's with hotrodded gaming machines. Isn't this project about the power of numbers rather than the strength of one? Lets face it, RC5 type projects are just guess-my-number with a big number, and SETI is just looking for the nonexistant. I always saw this project as a promotion of better programming and a way to raise awareness of fun in mathamatics. I just hope it does not become a ruler for peoples' computing <ahem> masculinity while bumping the people who put it where it is. Brian [EMAIL PROTECTED] _________________________________________________________________ Get your FREE download of MSN Explorer at http://explorer.msn.com _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sun, 11 Feb 2001 17:40:05 -0800 From: "Stephan T. Lavavej" <[EMAIL PROTECTED]> Subject: Mersenne: The Matrix With respect to the screensaver idea, The difficult thing about visualizing the Lucas-Lehmer compuation is that it's so... abstract. There isn't any immediately obvious way to see its progress other than what's currently done: a readout of the iteration number and clock cycle. However, another idea did spring into my mind. Everyone here is familiar with The Matrix, and the screens of green scrolling numbers and letters displayed at various points in the movie. I suggest that the screensaver might display the intermediate residues in hexadecimal in such a fashion, dropping the residues hexit-by-hexit on the screen such that one residue's worth of digits is displayed every iteration. This way, it's just not a green blurry readout of a list of residues, but a staggered waterfall of hexits. I think it would look nifty. Faster computers would have a faster flowing waterfall of hexits, and it would also depend on the size of the number being tested. It's simple, and it immediately shows the "fastness" of the computation and its intermediate results in a way that humans could see. People would probably want to be able to configure the colors of the hexits displayed, I'm sure. - -*---*------- Stephan T. Lavavej Author of "Mersenne Primes: Development through History, Ongoing Work, and a New Conjecture" at http://www.its.caltech.edu/~stl/paper.html _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Mon, 12 Feb 2001 08:59:10 +0100 From: "Jean Flinois" <[EMAIL PROTECTED]> Subject: Re: Mersenne: P4 speed and implications thereof >De : "Jud McCranie" <[EMAIL PROTECTED]> >� : "Nathan Russell" <[EMAIL PROTECTED]> > | Think recursively( Think recursively( Think recursively)) | > ...(stripped)... No offence meant, but this hurts my feeling of recursivity... maybe something like : Think_recursively(){if (!enough_thinking_done()) Think_recursively();} ? Jean Flinois <[EMAIL PROTECTED]> V-Technologies, Savenni�res t�l +33 (0)2 4172 1077 _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Mon, 12 Feb 2001 21:12:01 -0000 From: "Brian J. Beesley" <[EMAIL PROTECTED]> Subject: Re: Mersenne: P4 speed and implications thereof On 11 Feb 2001, at 18:17, Jeramy Ross wrote: > Nathan Russell wrote: > *snip* > > > This frankly makes me wonder how much longer there will be a place in > > GIMPS for slower machines. I'm not saying that's a bad thing - after > > all, 486s and original pentiums were the workforce when GIMPS began, > > and I wouldn't feel comfortable with the amount of opportunity for > > error involved in using one for GIMPS now. > > This tends to be a thorn in some people's sides when I have tried to get > them to run GIMPS on their machines. It would seem that some view the > GIMPS project as an 'elite' minded group that requires the use of the more > powerful PC's out there due to it's resource demanding nature. They don't > want to dedicate several months to a project and as the size of the > exponents > grow their usefulness shrinks and eventually they will be deemed no longer > useful for the project. So what's new? That bleeding-edge P100 you bought five years ago is a long way short of state-of-the-art now! I would have thought that this is _positive_. Those people who are lucky enough to have today's SoA systems, but do not need 100% of the CPU power on a 24x7 basis, often feel priveleged to be able to participate in a project which really can use the power - even if they've no real interest in the underlying project per se. Sure, that SoA system will look less impressive in a year or two's time - but, then, it can move on to double-checking. Even if they give up after a while, when the system becomes miserably underpowered by the standards of that time, that's still useful input. > This view has only been expressed by a few people that I have talked to, > but it maybe an opinion that will grow as the month's pass on. One can see > that the usefullness of the slower machines is limited as we march on > through > Primeland (If for no other reason other than it will take so long to > complete a > task with the slower machines). > This undoubtedly will be a natural evolution of the project, but may > leave a > bitter taste in the mouths of those who are on slower machines. Who knows? > I certainly don't know the impact of this on the project....good or bad.. Well - perhaps we need something to attract new users - a new search area which would provide short runs with rapid feedback, or some "useful" work which slower systems might complete before they make their journey to the landfill site. With respect to the message I forwarded earlier, perhaps hunting for primes of the form (2^p+1)/3 might be interesting, and with at least a conjectural link to Mersenne numbers. Regards Brian Beesley _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sun, 11 Feb 2001 20:08:06 -0500 From: Jud McCranie <[EMAIL PROTECTED]> Subject: Re: Mersenne: P4 speed and implications thereof At 07:07 PM 2/11/2001 -0500, Nathan Russell wrote: > My understanding is that they are designed to perform well for >graphical tasks; in my experience, people will buy even very expensive >computers if it improves the performance they see when doing graphical >tasks; gaming in particular comes to mind. From what I hear, the P4 is better than the Athlon at streaming data, but worse at everything else. Streaming data is good when you process a lot of data the same way, such as in graphics. But the P4 has a deep pipeline that will be dumped if there is a branch. If I were buying now, I'd get an Athlon. That might change in a few months. +-----------------------------------------------------------+ | Jud McCranie | | | | Think recursively( Think recursively( Think recursively)) | +-----------------------------------------------------------+ _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Mon, 12 Feb 2001 15:44:22 -0000 From: "Brian J. Beesley" <[EMAIL PROTECTED]> Subject: Mersenne: (Fwd) An efficient probable prime test for numbers of the form Anyone who isn't on the NMBRTHRY list will have missed the following interesting & relevant message: - ------- Forwarded message follows ------- Hi all, We have published a new article : "An efficient probable prime test for numbers of the form (2^p+1)/3" Abstract : The developpement of a new probabilistic test for numbers of the form (2^p+1)/3, which have many common properties with Mersenne numbers. This test gave us probable prime numbers with an exponent p<400000 and confirms the "new Mersenne conjecture" with new exponents. You can download it from : - - http://ourworld.compuserve.com/homepages/hlifchitz/ - - http://ourworld.compuserve.com/homepages/hlifchitz/Renaud.html Henri & Renaud LIFCHITZ Paris FRANCE - ------- End of forwarded message ------- _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sun, 11 Feb 2001 20:15:03 -0500 From: Jud McCranie <[EMAIL PROTECTED]> Subject: Re: Mersenne: P4 speed and implications thereof At 04:21 PM 2/11/2001 -0800, John R Pierce wrote: > the P4 is likely gonna ramp up to 2GHz, 3Ghz and beyond faster and farther >than AMD can ramp up the Tbird. Yes, most Intel chips max out at about 2.5 times their initial speed, and they expect the P4 clock speed to go up by at least a factor of 10. Examples (I think these numbers are about right): 8088 4.77 MHz -> 12 or 12 MHz, factor of about 2.3 80286 6 MHz -> 20 MHz, factor of 2.5 80386 16 MHz -> 40 MHz, factor of 2.5 486 ?? Pentium 60 MHz -> 200? MHz factor of 3.3 other pentiums ?? +-----------------------------------------------------------+ | Jud McCranie | | | | Think recursively( Think recursively( Think recursively)) | +-----------------------------------------------------------+ _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Mon, 12 Feb 2001 15:10:48 -0500 From: Nathan Russell <[EMAIL PROTECTED]> Subject: Re: Mersenne: P4 speed and implications thereof On Mon, 12 Feb 2001 13:29:01 -0500, you wrote: >>This frankly makes me wonder how much longer there will be a place in >>GIMPS for slower machines. I'm not saying that's a bad thing - after >>all, 486s and original pentiums were the workforce when GIMPS began, >>and I wouldn't feel comfortable with the amount of opportunity for >>error involved in using one for GIMPS now. > >I have been working on this project for about five years now. My first two >years with a 486DX266, and now with a Cyrix 266. My computer works fine >for now, and I do not need any more power (and I lack 2k-3k to invest in >video games) I have worked on trial factoring, and have received little >credit on the LL testing page. I hate to think that my faithfulness and >determination on this project will be overshadowed by passer-by's with >hotrodded gaming machines. Isn't this project about the power of numbers >rather than the strength of one? Lets face it, RC5 type projects are just >guess-my-number with a big number, and SETI is just looking for the >nonexistant. I always saw this project as a promotion of better >programming and a way to raise awareness of fun in mathamatics. I just >hope it does not become a ruler for peoples' computing <ahem> masculinity >while bumping the people who put it where it is. > >Brian >[EMAIL PROTECTED] I don't see it that way at all. My point in posting the message was to consider how long /any/ older machines - including my own Pentium 3 - - will be able to really contribute. I'm sorry if I upset you - I just thought I'd mention the implications for /all/ of us who don't have the latest hardware. As an aside, your Cyrex 266 is about 3 times as fast as the "original pentiums" I refered to as being obsolete, so I think there's a definate distinction - and hopefully will be for years yet to come. Even a 486 could still be used for trial-factoring, or for ECM, which would take such a machine roughly fifteen minutes in some cases. As regards RC5 and SETI, I believe I started a discussion on that issue when I first arrived on the list (late January 2000, IIRC). You might find that discussion interesting - especially the manner in which some of the senior GIMPS members responded to me. Nathan _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Sun, 11 Feb 2001 21:15:00 -0500 From: Jud McCranie <[EMAIL PROTECTED]> Subject: Re: Mersenne: P4 speed and implications thereof At 05:58 PM 2/11/2001 -0800, John R Pierce wrote: > > Pentium 60 MHz -> 200? MHz factor of 3.3 > P55c (mmx) hit 233MHz, I believe, for almost 4X Was P55c for notebooks? the first desktop Pentiums were 60 MHz. >P6 architecture (which includes PPro, P2 and P3), > 166 - 1GHz+ (factor of 6+) But individually they didn't scale that much. +-----------------------------------------------------------+ | Jud McCranie | | | | Think recursively( Think recursively( Think recursively)) | +-----------------------------------------------------------+ _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Mon, 12 Feb 2001 20:31:51 -0500 From: Jeff Woods <[EMAIL PROTECTED]> Subject: Re: Mersenne: P4 speed and implications thereof At 08:15 PM 2/11/01 -0500, you wrote: Examples (I think these numbers are about right): >8088 4.77 MHz -> 12 or 12 MHz, factor of about 2.3 >80286 6 MHz -> 20 MHz, factor of 2.5 >80386 16 MHz -> 40 MHz, factor of 2.5 >486 The 486 came in at, I think, 33 Mhz, and only went to the DX2, the 66 Mhz model. >Pentium 60 MHz -> 200? MHz factor of 3.3 The original Pentium went to, I think, 266 Mhz. >other pentiums ?? The P-II started at 233 Mhz, and went up to 466 (not counting the Celeron offshoots). _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Mon, 12 Feb 2001 21:08:26 -0500 From: Jud McCranie <[EMAIL PROTECTED]> Subject: Re: Mersenne: P4 speed and implications thereof At 08:31 PM 2/12/2001 -0500, Jeff Woods wrote: >The 486 came in at, I think, 33 Mhz, and only went to the DX2, the 66 Mhz >model. I believe it did come out at 33, and the DX4 went to 100 MHz. The DX2 was 2x the original 33, the DX4 was 3X. +--------------------------------------------------------+ | Jud McCranie | | | | 137*2^261147+1 is prime! (78,616 digits, 5/2/00) | +--------------------------------------------------------+ _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Mon, 12 Feb 2001 18:15:38 -0800 From: "John R Pierce" <[EMAIL PROTECTED]> Subject: Re: Mersenne: P4 speed and implications thereof > > > Pentium 60 MHz -> 200? MHz factor of 3.3 > > P55c (mmx) hit 233MHz, I believe, for almost 4X > > Was P55c for notebooks? the first desktop Pentiums were 60 MHz. P5 was original the 60-66Mhz Pentium w/ a 1x bus multiplier, and 5V I/O(I think?) P54 was the Pentium 75-200 MHz, using a 50 or 60 or 66MHz bus with various core multipliers. P55C was the Pentium with MMX. oh, hey, Mobile Pentium w/ MMX (P55 based) hit 300MHz, so we got a factor of 5X (60-300) all of these had the exact same integer and FPU core, hence they were all P5. also, for completeness... P6 was the Pentium Pro. BTW, just checked, there was a 150MHz P6. P64 was the Pentium-II P68(?) was the Pentium-III Katmai. I'm not sure if there was any such project designation for the current P-III "Coppermine". these chips (and various mobile spinnoffs) all used the exact same core for integer, level 1 cache, execution unit, and FPU. The P2/3 added MMX support. - -jrp _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Mon, 12 Feb 2001 18:17:08 -0800 From: "John R Pierce" <[EMAIL PROTECTED]> Subject: Re: Mersenne: P4 speed and implications thereof > The original Pentium went to, I think, 266 Mhz. .... comprehensive list at http://www.intel.com/pressroom/kits/processors/quickreffam.htm _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Mon, 12 Feb 2001 23:08:06 -0500 From: vincent mooney <[EMAIL PROTECTED]> Subject: Re: Mersenne: The Matrix Clever and I like it. I will use it when it is delivered. At 05:40 PM 2/11/01 -0800, Stephan T. Lavavej wrote: >With respect to the screensaver idea, >The difficult thing about visualizing the Lucas-Lehmer compuation is that >it's so... abstract. There isn't any immediately obvious way to see its >progress other than what's currently done: a readout of the iteration number >and clock cycle. However, another idea did spring into my mind. Everyone >here is familiar with The Matrix, and the screens of green scrolling numbers >and letters displayed at various points in the movie. I suggest that the >screensaver might display the intermediate residues in hexadecimal in such a >fashion, dropping the residues hexit-by-hexit on the screen such that one >residue's worth of digits is displayed every iteration. This way, it's just >not a green blurry readout of a list of residues, but a staggered waterfall >of hexits. I think it would look nifty. Faster computers would have a >faster flowing waterfall of hexits, and it would also depend on the size of >the number being tested. It's simple, and it immediately shows the >"fastness" of the computation and its intermediate results in a way that >humans could see. People would probably want to be able to configure the >colors of the hexits displayed, I'm sure. > >-*---*------- >Stephan T. Lavavej > _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Mon, 12 Feb 2001 20:54:55 -0800 From: "John R Pierce" <[EMAIL PROTECTED]> Subject: Re: Mersenne: Re: P4 Optimization > Today's mystery: I've found a case where adding a NOP instruction > speeds up the code by 9%. Not just a small loop, the 9% speedup > affects the entire 2nd pass of the FFT! As of now I have no theories > or explanations. Time to write some more code fragments to figure > it out..... I wonder if thats a cache line alignment thing with the code? _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Mon, 12 Feb 2001 18:40:34 +0100 From: Guillermo Ballester Valor <[EMAIL PROTECTED]> Subject: Re: Mersenne: Re: P4 Optimization [EMAIL PROTECTED] wrote: > One of the drawbacks of doing it > by hand in assembler...too bad high-quality HLL compilers (i.e. ones > capable of giving 80-90% of the performance of laboriously coded and > hand-tuned ASM, for complex, data-nonlocal algorithms requiring lots > of data prefetch) appear to be nigh-impossible to write for CISCs like > the x86 family. I don't want to start a RISC-versus-CISC flame war here, > but the fact is, no high-level FFT-based large-integer-multiply code has > gotten within a factor of 2 of the performance of Prime95 on the Pentium. Glucas, a C-coded program to L-L test reaches about 55% performance with respect Prime95 when is compiled with GNU/gcc compiler on Pentium and no assembler macros used. After including about three hundred lines of assembler lines, the performance rise to about 65%. Regards. Guillermo. _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Tue, 13 Feb 2001 00:54:20 -0600 From: Ken Kriesel <[EMAIL PROTECTED]> Subject: Mersenne: Processor short family histories The 8088 debuted at 5 (& 8) Mhz; IBM derated it a bit for the PC because 4.77Mhz*3 = 14.318 = 4 * 3.57Mhz (TV color burst frequency). An IBM (pre-XT) motherboard could be pushed to about 7.5 Mhz by splitting the clock signal paths. FPU was separate. Ten Mhz chips were offered. This chip had competition from the 8086, 80186, and NEC V20. Intel offered the 286 with 6, 8, 10, and 12.5 Mhz on one data sheet. AMD got to 16 on this one, but an early data sheet lists 4, 6, and 8 (and says reprinted by permission of Intel). FPU was separate. I don't recall a 286-20. The original IBM AT was designed with CPU clocked at 6Mhz, FPU at 4Mhz. The 386 debuted at 12.5 and 16 Mhz. These had no cache on-chip. On-motherboard cache was common for 25Mhz and above. Intel's offerings topped out at 33Mhz, AMD offered 40 eventually. The 386 core was also throttled down to 16-bit bus width in the 386SX, which had speeds as low as 10Mhz and up to 20 or higher. FPU was separate. Cyrix offered a faster FPU. TI eventually offered a clock-doubled near 486 with cache in a 386 pinout. The 486 was the first to offer an on-chip FPU, came out at 20 and 25 Mhz and went to 100 Mhz (core=3x memory bus) in the Intel line, 133 (4x) elsewhere (AMD?). The 486 socket's performance could be stretched a little further by using a Pentium Overdrive chip from Intel; at 83 Mhz (2.5x) giving 1.7x the performance of a 486-66 in real world finite element analysis (ANSYS). The 486SX was a no-fpu 486, with full memory bus width. The Pentium I's in various subflavors went from 60 & 66 at announcement, to 200 (nonMMX) and 233 in the MMX type. I think the Pentium Pro's went 166, 180, and 200. Pentium II's included the 266 Mhz speed. Ken _________________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers ------------------------------ Date: Tue, 13 Feb 2001 01:32:46 -0800 From: Paul Leyland <[EMAIL PROTECTED]> Subject: RE: Mersenne: P4 speed and implications thereof > Well - perhaps we need something to attract new users - a new search > area which would provide short runs with rapid feedback, or some > "useful" work which slower systems might complete before they make > their journey to the landfill site. With respect to the message I > forwarded earlier, perhaps hunting for primes of the form (2^p+1)/3 > might be interesting, and with at least a conjectural link to > Mersenne numbers. The integer factoring people can always use more cycles, and even antique machines make valuable contributions in a reasonable time. For instance, I have small number of DECstations and a Sun SS2 on my home network all running the ECMNET client. These boxes are perhaps as powerful as a 486DX2-66. Certainly the P90 ECMNET server at home is significantly more powerful than any of them. Until quite recently I was still using a 386-25 for factoring by MPQS. Please contact me, or take a look at http://research.microsoft.com/users/pleyland/factorization/main.htm and the ECMNET site at http://www.loria.fr/~zimmerma/records/ecmnet.html if you'd like more info. Paul _________________________________________________________________________ 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 #816 ******************************
