just an informal question - Page 2

CRGreathouse · 2017-07-26, 19:31

Quote:

Originally Posted by GP2

All Mersenne prime exponents p are themselves prime, and therefore either p ≡ 1 (mod 4) or p ≡ 3 (mod 4), except for the trivial case of p = 2. According to the Lenstra–Pomerance–Wagstaff conjecture, the two cases (1 and 3) do not produce Mersenne primes with exactly equal probability.

Right, but that effect goes away with sieving, doesn't it?

chalsall · 2017-07-26, 19:42

Quote:

Originally Posted by CRGreathouse

Right, but that effect goes away with sieving, doesn't it?

Quickly.

LaurV · 2017-07-27, 10:37

Quote:

Originally Posted by masser

Surely the gods would not be so cruel...

BTW what is the ETA for regularly testing exponents near 300M?

you mean 332M, or 100M digits. That is from 50 days per exponent with top computers or top GPUs. A bit faster with top Xeon servers. An i7-6950x may do the work in about 60 days too, with all 10 cores crunching (however this is not the best output, you may get two exponents done in about 100 or 110 days, depends on your system). A classic Titan needs 60 days, and you can short it to ~55 with good cooling (liquid). Newer cards on which I didn't lay my hands yet, can do it a bit faster. If the rumored i9 series (especially i9-7980XE) comes with what is rumored (18 cores 36 threads, 8 memory channels, turbo mode to 4 Gig or so), this will already short the time to half. Which is exactly the "30 days per exponent for top range home computers" we were facing when GIMPS kicked the 10M digits EFF prize. But the number of exponents candidates at that time were 10 times lower....

masser · 2017-07-27, 23:25

Thanks, LaurV. That's not as bad as I thought.

I was more curious about when the "mainline" first-time LL testing "wave" will reach the 300M or the 332M exponent range. 15 years?

fivemack · 2017-07-28, 07:38

Quote:

Originally Posted by masser

Thanks, LaurV. That's not as bad as I thought.

I was more curious about when the "mainline" first-time LL testing "wave" will reach the 300M or the 332M exponent range. 15 years?

Longer, I expect; computers have stopped getting vastly faster, we'll have a bit of a boost if Intel put AVX512 into their normal-end cores but I'm not confident that's something they'll do (the units are physically very big, and use serious amounts of power when running 24/7).

It was fifteen years from 'first 10M tested' to 'last 10M double-checked', but those fifteen years were the last fifteen years of full-bore Moore's Law.

I have just turned forty, and do not expect to see 'last 100M double-checked' before I retire.

henryzz · 2017-07-28, 10:16

Quote:

Originally Posted by fivemack

Longer, I expect; computers have stopped getting vastly faster, we'll have a bit of a boost if Intel put AVX512 into their normal-end cores but I'm not confident that's something they'll do (the units are physically very big, and use serious amounts of power when running 24/7).

It was fifteen years from 'first 10M tested' to 'last 10M double-checked', but those fifteen years were the last fifteen years of full-bore Moore's Law.

I have just turned forty, and do not expect to see 'last 100M double-checked' before I retire.

I suspect that there is potential for cpus to speedup significantly when there is a transition to non-silicon.

davar55 · 2017-08-08, 15:28

Where (in what thread) were guesses for the location of M50 being collected?

thyw · 2017-08-08, 19:38

Quote:

Originally Posted by davar55

Where (in what thread) were guesses for the location of M50 being collected?

Predict M50
http://www.mersenneforum.org/showthread.php?t=20839

science_man_88 · 2017-08-08, 20:18

Quote:

Originally Posted by thyw

Predict M50
http://www.mersenneforum.org/showthread.php?t=20839

I think the funny part is davar55 last posted in that thread February last year.

davar55 · 2017-08-09, 13:11

Maybe my imperfect memory is funny. I remembered the thread,
but not its name or location. And I don't have the access to or
facility with the forum that many here seem to have.

science_man_88 · 2017-08-09, 13:18

Quote:

Originally Posted by davar55

Maybe my imperfect memory is funny. I remembered the thread,
but not its name or location. And I don't have the access to or
facility with the forum that many here seem to have.

Google search ? the forum search ?

2017-07-27, 23:25	#15
masser Jul 2003 wear a mask 2²·419 Posts	Thanks, LaurV. That's not as bad as I thought. I was more curious about when the "mainline" first-time LL testing "wave" will reach the 300M or the 332M exponent range. 15 years?

2017-08-08, 15:28	#18
davar55 May 2004 New York City 5·7·11² Posts	Where (in what thread) were guesses for the location of M50 being collected?

2017-08-09, 13:11	#21
davar55 May 2004 New York City 1000010001011₂ Posts	Maybe my imperfect memory is funny. I remembered the thread, but not its name or location. And I don't have the access to or facility with the forum that many here seem to have.