just finished the biggest (successful) LL test

[joblack]

As far as I know I just finished successfully the biggest LL test:

[Dec 29 18:02] M123456811 is not prime. Res64: A41D7D17044F74xx. We4: xxx
[Dec 29 18:02] Setting affinity to run helper thread 1 on any logical CPU.
....
[Comm thread Dec 29 18:02] pnErrorResult=0
[Comm thread Dec 29 18:02] CPU credit is 605.4956 GHz-days.
[Comm thread Dec 29 18:02] ==END==

Unfortunately it's not a prime number (it even looks very nice). It needed more than a year with two cores running.

Anyway, who got more CPU credits? ;)

[TimSorbet]

Not quite the biggest, these two beat it out:
http://www.mersenne.org/report_expon...p_lo=332197123
http://www.mersenne.org/report_expon...p_lo=150000091
By the way, here's a listing with a lot of information on exponent statuses:
http://www.mersenne.org/primenet/
And here's a list of LL results for exponents over 100M:
http://www.mersenne.org/report_LL/?e...B1=Get+LL+data
Third-biggest, still quite impressive! Unfortunately, with a non-0 error code, this test is more likely (than average) to be bad.
Please hide the last two hexdigits in the Res64 so nobody is able to resubmit that result and get all that credit without any work.
By the way, I should hope you really did TF farther than the 65 bits and no P-1 that PrimeNet knows about before running this monster LL.

The error code indicates 1 ROUNDOFF > 0.4 error occurred. From this thread, it looks like about 40% of numbers with that sort of error code are usually bad. Considering this and the length of time this number took, I'd give it a 50/50 chance for actually being correct.

[moebius]

Unfortunatley it is a Suspect LL.

[joblack]

Originally Posted by Mini-Geek

Third-biggest, still quite impressive! Unfortunately, with a non-0 error code, this test is more likely (than average) to be bad.
Please hide the last two hexdigits in the Res64 so nobody is able to resubmit that result and get all that credit without any work.
By the way, I should hope you really did TF farther than the 65 bits and no P-1 that PrimeNet knows about before running this monster LL.

The error code indicates 1 ROUNDOFF > 0.4 error occurred. From this thread, it looks like about 40% of numbers with that sort of error code are usually bad. Considering this and the length of time this number took, I'd give it a 50/50 chance for actually being correct.

The error occurred before I disabled the error corrections (and I downclocked the cpu) so I'm pretty confident it will be right. We'll see in a few years.

Tell me more about the correct way to proceed TF and P-1 (up to which bits?) before running such a number? I thought Prime95 will do the necessary (and useful) tests for itself? Any receipts?

[davieddy]

I think <2025 is pie in the sky.

David

[joblack]

Originally Posted by Mini-Geek

Not quite the biggest, these two beat it out:
http://www.mersenne.org/report_expon...p_lo=332197123
http://www.mersenne.org/report_expon...p_lo=150000091
By the way, here's a listing with a lot of information on exponent statuses:
http://www.mersenne.org/primenet/
And here's a list of LL results for exponents over 100M:
http://www.mersenne.org/report_LL/?e...B1=Get+LL+data
Third-biggest, still quite impressive!

Damn it, there are always bigger fishes in the ocean. ;-)

[davieddy]

Originally Posted by joblack

Damn it, there are always bigger fishes in the ocean. ;-)

You asked for it:
Too many fish in the sea

David

[cheesehead]

Originally Posted by joblack

Tell me more about the correct way to proceed TF and P-1 (up to which bits?) before running such a number? I thought Prime95 will do the necessary (and useful) tests for itself? Any receipts?

What did your Test= worktodo line look like (without the assignment ID, please)?

Test=aID,exponent,TFdone,P-1done

where TFdone = bit level to which TF had previously been done

and

P-1done = 1 if P-1 has been done, else 0

If your worktodo line was:

Test=xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx,123456811,73,1

for instance,

then prime95 would think that it had already been TFed to 2^73 (which is the default limit for exponents from 115300000 to 147500000), and that standard P-1 had already been done.

It should have looked like:

Test=xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx,123456811,65,0

which indicates that no P-1 has been done, and TF to only 2^65, matching the PrimeNet database entry for that exponent.

- - -

(I'm tracking down what prime95 would do if the last two parameters had been left off.)

[petrw1]

Exponent Status Data  
123456811 No factors below 2^65 
 Suspect LL A41D7D17044F74__ by "joblack" on 2010-12-29 
 Assigned LL testing to "joblack" on 2010-12-07 <====================
 History no factor from 2^62 to 2^63 by "GIMPS Visualization" on 2009-03-06 
 History no factor from 2^63 to 2^64 by "Richard" on 2010-05-30 
 History no factor from 2^64 to 2^65 by "Richard" on 2010-12-06 
 History A41D7D17044F74__ by "joblack" on 2010-12-29

[TimSorbet]

Originally Posted by joblack

Tell me more about the correct way to proceed TF and P-1 (up to which bits?) before running such a number? I thought Prime95 will do the necessary (and useful) tests for itself? Any receipts?

If you look in results.txt, it will list any TF or P-1 that was completed. If you have anything to report and it's not reporting automatically, (which, unless there are complicating factors like the TF and P-1 being run on a different computer, it's clearly not - though you could try triggering a manual communication to see if it does anything different) you could report it at http://www.mersenne.org/manual_result/. If you use Test= and gave it the appropriate (or lower-than-actual) TF and P-1 levels, (which it does any time PrimeNet gives it to you) it will have run both to the optimal level automatically. If you used Test= and gave it inaccurately high TF/P-1 levels/completions, it will skip work based on that. If you used AdvancedTest= (which is the same as clicking Advanced > Test in the GUI), it will skip all TF/P-1. If you know exactly how you gave it to Prime95, you can know exactly what you can expect it to have done. And you can also look at the results.txt and see what it actually did.

Originally Posted by cheesehead

(I'm tracking down what prime95 would do if the last two parameters had been left off.)

Looks like you might have forgot about this or can't do it right now, so I did it: it ("Test=123456811") is the same as if you told it no TF or P-1 had been done ("Test=123456811,0,0"), so it starts TFing at the beginning. This would duplicate a little work, but only about as much as the first bit level you need to test would take, (i.e. from 0 to 2^65 is roughly the same work as from 2^65 to 2^66, and you needed to do 2^65 to 2^73) a minuscule amount for a test this large.

[Batalov]

If I remember correctly, read it sometime ago on this forum -- "if someone decides to finish the TF from 65 to 73 bits and finds a factor - the LL credit will be voided."