fond of a factor? Turn yourself in to become inane

[lycorn]

Another one down:

UID: lycorn/asteroid, M27943 has a factor: 9595745432621621468883227471666940222487 (ECM curve 228, B1=3000000, B2=300000000),

132.818 bits

k = 3 × 7 × 2029 × 20873 × 575231 × 47418037 × 7077886967119

Not too rough, not too smooth. Just fine...

[aurashift]

I keep seeing this thread on my way into LMH>100M

http://www.mersenne.ca/exponent/78282691

127 bits, do i win a prize?

[firejuggler]

thats a really nice factor

[PageFault]

Nice find!

Are you going to do your others to B1=44000000? I did a test on my block, M25xxx, and got 27 curves per day. Quite an effort, to finish these 10 or so through to 17100 curves, several years or so ... I'll begin mass deployment in a few days time, is it better to do chunks of each on all cores or the full run on a single core? The latter will take over a year ...

I hope they come up with better hardware real soon, and that I can be positioned to make advantage ... long stretch needless to say, but I am determined. Am I crazy enough to finish this, and embark on the next B1 ???

BTW Stage 2 only takes 750 MB.

Originally Posted by lycorn

Another one down:

UID: lycorn/asteroid, M27943 has a factor: 9595745432621621468883227471666940222487 (ECM curve 228, B1=3000000, B2=300000000),

132.818 bits

k = 3 × 7 × 2029 × 20873 × 575231 × 47418037 × 7077886967119

Not too rough, not too smooth. Just fine...

[alpertron]

Today my computer found a prime factor of the same size (133 bits):

[Thu Jun 04 02:54:08 2015]
P-1 found a factor in stage #2, B1=1000000, B2=30000000, E=12.
M955363 has a factor: 7315800908340060929629104157850801671121 (P-1, B1=1000000, B2=30000000, E=12)

k = 2^3 × 5 × 173 × 3733 × 4111 × 100361 × 102563 × 298631 × 11729009

[lycorn]

Originally Posted by PageFault

Nice find!

Are you going to do your others to B1=44000000? I did a test on my block, M25xxx, and got 27 curves per day. Quite an effort, to finish these 10 or so through to 17100 curves, several years or so ... I'll begin mass deployment in a few days time, is it better to do chunks of each on all cores or the full run on a single core? The latter will take over a year ...

I hope they come up with better hardware real soon, and that I can be positioned to make advantage ... long stretch needless to say, but I am determined. Am I crazy enough to finish this, and embark on the next B1 ???

BTW Stage 2 only takes 750 MB.

I will do several tests with B1=11e6 for the next couple of months (exponents ranging from 12000 to 15000).
I will also pursue the quest for a factor of M1277, using Prime95 for Stage 1 and GMP-ECM for Stage 2. If you have a fair amount of memory available, you could also give it a go.

As for how to split the tests among several cores, there´s not a lot to be gained in doing it one way or the other. If I were you, I would probably do a mix: e.g. test 2 exponents on 4 cores, assigning 2 chunks of the same test to each core. That´s what I´m currently doing...

[firejuggler]

a nice one down pm1 find
M78235639 has a factor 8803993906198449472411631 (82.6 bit, ) found in stage 1
k=5 × 43 × 131 × 3643 × 5827 × 94109

[Madpoo]

Originally Posted by lycorn

I will also pursue the quest for a factor of M1277, using Prime95 for Stage 1 and GMP-ECM for Stage 2. If you have a fair amount of memory available, you could also give it a go.

I'm about halfway through doing P-1 on M1277 at:
B1=5,000,000,000,000 B2=10,000,000,000,000

Methinks it'll finish in a couple more weeks or something. I was bored and thought it'd be fun to allocate a few spare cores to P-1 work on smaller exponents, even though I know ECM would probably be more fruitful, or so I've heard.

I'm also doing these to the same bounds:
M1619
M1753
M2137
M2267
M2273

Each running on one core of a 6-core Xeon (while the other chip in this dual-CPU system is doing LL tests).

I started the M1277 run a while back so it's further along. The other ones will be a month or so before finishing up. Those last two are only at 14% right now.

Speaking of P-1, is there any consensus out there about the fastest program for P-1 work? Would gmp-ecm be something to look at for P-1 as well as ECM, or maybe something else (Windows based)?

[lorgix]

Originally Posted by Madpoo

I'm about halfway through doing P-1 on M1277 at:
B1=5,000,000,000,000 B2=10,000,000,000,000

Methinks it'll finish in a couple more weeks or something. I was bored and thought it'd be fun to allocate a few spare cores to P-1 work on smaller exponents, even though I know ECM would probably be more fruitful, or so I've heard.

I'm also doing these to the same bounds:
M1619
M1753
M2137
M2267
M2273

Each running on one core of a 6-core Xeon (while the other chip in this dual-CPU system is doing LL tests).

I started the M1277 run a while back so it's further along. The other ones will be a month or so before finishing up. Those last two are only at 14% right now.

Speaking of P-1, is there any consensus out there about the fastest program for P-1 work? Would gmp-ecm be something to look at for P-1 as well as ECM, or maybe something else (Windows based)?

If you are doing stage1 all the way to 5e12 it's highly recommended that you take stage2 further than 1e13. Maybe 6e13~1e14 or so.

Yes, ECM is more efficient.

For large enough numbers Prime95 is faster in stage1. GMP-ECM is faster in stage2, depending on bounds and input size.
So in short; it depends. But GMP-ECM can be way faster in stage2.

[LaurV]

For how much ECM was done on these thingies, there is a very slim chance that P-1 finds a factor.

[lycorn]

Yeah, Prime95+GMP-ECM would really be the best bet for the lowest exponents in Primenet ranges, specially for machines with large amounts of memory.