28.0M - 28.01M to 65 bits for starters

Boulder · 2004-05-25, 08:46

Hi,

I've started factoring the range to 65 bits. I'll probably go deeper as soon as the round is complete and I'll keep you posted.

Boulder · 2004-07-08, 12:33

Now going at M28005269, only one exponent with a factor found so far

ThomRuley · 2004-07-08, 14:20

One exponent? That's great! One less exponent for LL testing. Your couple hours of factoring just saved the project about a month of LL. Great job.

Boulder · 2004-07-08, 17:15

Well, I would have expected a couple but then again, aren't the factors more rare when you TF the exponents 'deeper' ?

Uncwilly · 2004-07-08, 21:42

Consider that by factoring up to 55bits, many many factors have been removed (maybe 50% of all exponents are eliminated by factoring to 55bit). What you are doing is squeezing a few more out before P-1 will get ~1% of the remaning. All of this proves that it is not prime and saves much time in the hunt.

markr · 2004-07-09, 04:31

All factors are good - they save on LL tests! One factor is well within what's expected: if I'm close in reckoning that you had TF'd 114 exponents from 2^64 to 2^65, you had 30% chance of one factor and 27% chance of two factors.

According to prime95's help file, the chance of finding a factor for an exponent between 2^b & 2^(b+1) is approximately 1/b. The killer is that the time taken doubles. It is very much worth it though - TF will still be cost-effective up to 2^67 for these exponents.

Boulder · 2004-07-09, 05:29

Thanks for the explanations, guys.

I've been thinking about starting TF'ing the remaining exponents to 2^66 as soon as 2^65 is finished but I'm not sure how far I would get before the range will be in PrimeNet. Since I do some real work on the computer (a P4-2.76GHz), it's very hard to estimate the time needed. At the moment the time per exponent is a bit over three hours if I'm not doing anything special. If I shut down all the other processes the time might go down to two and a half hours so it would take about 5-7 hours per exponent when TF'ing to 2^66.

Guess I'll just go with the range as long as I can

EDIT: By the way, is it that TF'ing to 2^64 is not as effective on a SSE2-capable machine as to 2^65?

markr · 2004-07-09, 07:34

Above 2^64, SSE2 gives a big advantage - I don't know by how much.

Below 2^62, Athlons outpace P4s apparently because of some fancy coding that does two-for-one somehow. I don't know any more about why it is (and even this might be wrong!) or where non-SSE2 intel cpus fit in, but I do know that for TF below 2^62 an Athlon XP at 1533MHz is way faster than a P4 at 1800MHz.

In between... not sure.

garo · 2004-07-09, 10:40

These were some tests I ran a while back:

Time taken to complete 14366959 to 0.98%

Code:

To Bit  On PII 450      On P4 2533      Improv  On 1333TB       Improv

59      6               2.5             2.4     1.85            3.25
60      12              5               2.4     3.7             3.25
61      24              10              2.4     7.4             3.25
62      48              20              2.4     14.8            3.25
63      170             37.5            4.5     51              3.33
64      340             75              4.5     102             3.33
65      1540            180             8.5     480             3.21

Boulder · 2004-11-06, 11:18

Hi all,

I had an HD crash two months ago and lost a lot of work. I'm now attempting to factor the range to 2^65 before it hits the network, but can't promise anything. I remember seeing at least one number being factored so there's some motivation at least.

Please notify me if the range goes out, I might not remember to log in often.

Boulder · 2004-11-18, 15:27

OK, I'll release the range as it should soon enter PrimeNet. I emailed the results to George, I nailed that one candidate which I remembered getting factored

2004-05-25, 08:46	#1
Boulder May 2003 3×13 Posts	28.0M - 28.01M to 65 bits for starters Hi, I've started factoring the range to 65 bits. I'll probably go deeper as soon as the round is complete and I'll keep you posted.

2004-07-09, 05:29	#7
Boulder May 2003 100111₂ Posts	Thanks for the explanations, guys. I've been thinking about starting TF'ing the remaining exponents to 2^66 as soon as 2^65 is finished but I'm not sure how far I would get before the range will be in PrimeNet. Since I do some real work on the computer (a P4-2.76GHz), it's very hard to estimate the time needed. At the moment the time per exponent is a bit over three hours if I'm not doing anything special. If I shut down all the other processes the time might go down to two and a half hours so it would take about 5-7 hours per exponent when TF'ing to 2^66. Guess I'll just go with the range as long as I can EDIT: By the way, is it that TF'ing to 2^64 is not as effective on a SSE2-capable machine as to 2^65? Last fiddled with by Boulder on 2004-07-09 at 05:37

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
52.1M - 52.5M to 63 bits	Rde	Lone Mersenne Hunters	9	2008-10-18 00:33
76.5-77M to 63 bits	lycorn	Lone Mersenne Hunters	2	2007-05-25 21:32
64 bits versus 32 bits Windows	S485122	Software	2	2006-10-31 19:14
52.0M - 52.1M to 63 bits	Rde	Lone Mersenne Hunters	4	2006-10-14 18:21
35-35.2 to 62 bits, cont from 61 bits	Khemikal796	Lone Mersenne Hunters	12	2005-12-01 21:35

2004-07-08, 12:33	#2
Boulder May 2003 3×13 Posts	Now going at M28005269, only one exponent with a factor found so far

2004-07-08, 14:20	#3
ThomRuley May 2003 2³·31 Posts	One exponent? That's great! One less exponent for LL testing. Your couple hours of factoring just saved the project about a month of LL. Great job.

2004-07-08, 17:15	#4
Boulder May 2003 3×13 Posts	Well, I would have expected a couple but then again, aren't the factors more rare when you TF the exponents 'deeper' ?

2004-07-08, 21:42	#5
Uncwilly 6809 > 6502 """"""""""""""""""" Aug 2003 101×103 Posts 2⁴·593 Posts	Consider that by factoring up to 55bits, many many factors have been removed (maybe 50% of all exponents are eliminated by factoring to 55bit). What you are doing is squeezing a few more out before P-1 will get ~1% of the remaning. All of this proves that it is not prime and saves much time in the hunt.

2004-07-09, 04:31	#6
markr "Mark" Feb 2003 Sydney 3·191 Posts	All factors are good - they save on LL tests! One factor is well within what's expected: if I'm close in reckoning that you had TF'd 114 exponents from 2^64 to 2^65, you had 30% chance of one factor and 27% chance of two factors. According to prime95's help file, the chance of finding a factor for an exponent between 2^b & 2^(b+1) is approximately 1/b. The killer is that the time taken doubles. It is very much worth it though - TF will still be cost-effective up to 2^67 for these exponents.

2004-07-09, 07:34	#8
markr "Mark" Feb 2003 Sydney 3×191 Posts	Above 2^64, SSE2 gives a big advantage - I don't know by how much. Below 2^62, Athlons outpace P4s apparently because of some fancy coding that does two-for-one somehow. I don't know any more about why it is (and even this might be wrong!) or where non-SSE2 intel cpus fit in, but I do know that for TF below 2^62 an Athlon XP at 1533MHz is way faster than a P4 at 1800MHz. In between... not sure.

2004-11-06, 11:18	#10
Boulder May 2003 3×13 Posts	Hi all, I had an HD crash two months ago and lost a lot of work. I'm now attempting to factor the range to 2^65 before it hits the network, but can't promise anything. I remember seeing at least one number being factored so there's some motivation at least. Please notify me if the range goes out, I might not remember to log in often.

2004-11-18, 15:27	#11
Boulder May 2003 100111₂ Posts	OK, I'll release the range as it should soon enter PrimeNet. I emailed the results to George, I nailed that one candidate which I remembered getting factored