M40 Again!!!!

[TauCeti]

Originally posted by Dresdenboy
That topic already appeared in the news (in German) of the popular german computer magazine "c't":
http://www.heise.de/newsticker/data/as-18.11.03-000/

(It's the same mag that tested Prime95 on Opteron in april)

Also in http://www.theinquirer.net/?article=12732

[gbvalor]

Well,

At 10:02 EST today I began a verification of new Mersenne Prime. The negative part is I'm not the owner of the system and it can be killed or slowdown in any moment .

Now it runs at 0.047 sec/iter

Regards.

Guillermo

[jeff8765]

George, have you been able to get the save file of the number?

[smh]

Originally posted by jeff8765
George, have you been able to get the save file of the number?

George wrote that the finder of the prime mailed the save file. He himself ran the last 2000 iterations and the program reported the number prime.

[smh]

Now it runs at 0.047 sec/iter

(20e6 x 0.047)/86400 = 10,88 days so with a bit of luck we know before the end of the month.

[GP2]

Originally posted by wpolly
Could anyone tell me when did the Primenet server start to assign 18M/19M/20M/21M/22M+ exponents?

The leading edge of first-time checking reached:

18M : Jan 30 2003
19M: Apr 17 2003
20M: Jul 10 2003
21M: Oct 30 2003
22M: not yet

[pakaran]

Originally posted by GP2
The leading edge of first-time checking reached:

18M : Jan 30 2003
19M: Apr 17 2003
20M: Jul 10 2003
21M: Oct 30 2003
22M: not yet

But keep in mind that if the winning system was a P2 P3 or original Athlon, it could have been working since August. My P3-600 took most of a month to do the 10Mbit first tiime LLs that were being given out at the time.

Alternatively, the user may have been keeping a large queue (if this was one of several machines he was feeding by hand behind a firewall, or a friend's machine that he couldn't count on regularly dialing in - I've known tpr members in both situations).

OR the machine may have had a a minor problem like an unplugged router or a lose NIC that took the user a long time to notice, and in the meantime it was sitting on this prime.

Bottom line: speculation based on "where the first time tests were this month" only should be used for ballpark estimates.

[pakaran]

George, can you tell us whether this number has more or less than 20 mibibits?

[ixfd64]

*claws wall*

Must...wait...to...early...December....

[pakaran]

Originally posted by smh
George wrote that the finder of the prime mailed the save file. He himself ran the last 2000 iterations and the program reported the number prime.

As I've posted before, that's more or less a guarantee against cracking. For any value of iteration n, there may be several, one, or no values at iteration n-1 that will give that value. There may be more than one path from a value 2000 iterations back, but finding even one that works should be outright impossible even if an iteration of the LL test were easy to reverse (which it isn't).

Brian J Beesley posted to that effect on the mersenne mailing list when the last M40 candidate was found, when the backchecking was being proposed.

The problem is that the compiler thought NaN was equal to 0, among other things!

[GP2]

I think I know what M40 is.

Here's my "encrypted" prediction:

40618373

This is NOT the exponent. It's not even prime.

After the number is announced, I'll explain the meaning of "40618373". But only if I'm right