k=33661: Prime or bug ? [Archive]

Zuzu

10-19-2007, 07:50 PM

Hello,
The "Max n" of the current test window is stuck at 14277480 for k=33661, compared with 14292-14293K for the other tests. Furthermore the number of pending tests seems to be decreasing for that k, meaning that probably no new tests are assigned.
The last time I've seen that it was for k=27653 when the "max n" was stuck at the value of 9248397, a few hours before a prime was announced.
So, is there a new prime or is it just a bug ?

vjs

10-20-2007, 09:18 AM

Could be either... just have to wait and see.

Joh14vers6

10-21-2007, 12:19 PM

So far there is still no change in the situation.

Zuzu

10-21-2007, 01:00 PM

When looking at www.seventeenorbust.com/secret we have this (excuse me for the bad formatting):

2:43 pm EST, 21 Oct. 2007
Secret Statistics
Server's Assignment Queues 0 minutes old Last 24 Hours
Queue Name # Tests Min n Max n # Done n Increase
dropped-tests 0 n/a n/a 10 +∞
error-fix 0 n/a n/a 0 n/a
first-pass 13618 14277744 14999911 109 +-19519
garbage 0 n/a n/a 0 n/a
global 0 n/a n/a 0 n/a
largest-prime 0 n/a n/a 0 n/a
missing-test 0 n/a n/a 0 n/a
(not queued) n/a n/a n/a 1 n/a
residue-recovery 0 n/a n/a 0 n/a
second-pass 54096 7034832 13467934 10 +-694
Overall 67714 7034832 14999911 130 +-694

So, the "Min n" of firstpass is stuck at 14277744 which is the n value assigned to k=33661 next to 14277480 according to the Sob.dat datasets.
I doubt this is a concidence and my opinion turns to a bug. Still wish I'm wrong...

vjs

10-21-2007, 09:20 PM

I don't think I'm going too far out on a limb here by saying that its more than likely a prime. Just keep on keeping on until the annoucement. It will certainly take a few days to verify then we will know for certain. Don't drop any tests or anything in the time being.

Keroberts1

10-22-2007, 03:09 PM

well the biggest encouragment I can take from this is that no admins have denied a prime yet. So I'm keeping my hopes alive until we hear from them.

engracio

10-22-2007, 08:53 PM

I hope they would implement the recently discussed faster prime verification. AFAIK the last prime took longer than previous primes to verify and submitted as prime.

engracio

10-23-2007, 09:08 PM

Is it soup yet???????:umm:

e:)

vjs

10-24-2007, 08:29 AM

Humm...

Maybe someone should e-mail Louie or Dave just incase??? LOL...

Zuzu

10-24-2007, 09:07 AM

Thank you for your suggestion. As having initiated the post I just now wrote a private e-mail to both, precising the situation. Hope a reply soon...

vjs

10-24-2007, 11:12 AM

Hey!! it's a possibility... but unlikely.

The steps are basically, Louie or dave or someone re-runs the test using the SoB client to verify the prp "possible prime".

Then they send it to someone else who runs a primality proving program.

Along with some communication back and forth this takes sometime. I wouldn't expect to hear anything this week.

hhh

10-24-2007, 11:41 AM

well the biggest encouragment I can take from this is that no admins have denied a prime yet. So I'm keeping my hopes alive until we hear from them.

I don't remember ever to have heard any confirmation or dementi whatsoever from the admins before the final answer. I think it is ok like this, and we can play this game whenever we want.

Please remember, though, that even if the handing out of candidates has been stopped because of a positive result, that doesn't mean that the result is correct.

We'll have to wait, that's for sure.

H.

Zuzu

10-24-2007, 11:54 AM

OK, OK sorry, didn't pay attention to the meter:bonk:
Nevertheless it's a bit of a step back from the last prime which went nearly unnoticed beforehand:thumbs:
So, inshallah...

vjs

10-24-2007, 02:08 PM

Not a step back...

It's just a slow annoucement :-)

They may be taking time to put together a press release or something. Also you never know how long the other steps are going to take... wait for computer time, people on vacation etc.

No matter how long it takes there has not been any lost time projectwise.

Zuzu

10-25-2007, 09:31 AM

OK I understand. It took more than 5 weeks to verify the last prime compared with 1 week for k=27653 at n=9.2M. So this time we'll have to be patient :)
On the other hand a false positive (error in PRP residue computation) would be detected faster, so the more we wait...

vjs

10-25-2007, 11:06 AM

Yup.. That's it exactly.

And the longer we wait the more likely the larger the prime, I have a feeling this is going to be a firstpass prime and therefore the largest SoB has ever found, at approximately n=14M this ranks pretty high on the global scale as well.

This also begs another question... anyone know what I'm thinking about?

About to step on the soap box.

engracio

10-25-2007, 11:56 AM

Ummm, Is it soup yet??????? :umm: I've been waiting for the email/im stating it was one of my boxes who found it.:rotfl: Guess have to wait a bit longer.:D

e:)

Brucifer

10-25-2007, 02:28 PM

Yup.. That's it exactly.
This also begs another question... anyone know what I'm thinking about?

About to step on the soap box.

aww come on.. step up on the box. :thumbs:

vjs

10-25-2007, 07:02 PM

Well currently building the box... more later.

Guilherme

10-25-2007, 07:18 PM

What is the soap box?

vjs

10-25-2007, 09:18 PM

jasong

10-25-2007, 10:05 PM

Double checking... and error rates.

I and others have talked alot about it in the past, currently its not a big issue.
One can revisit previous posts but based on past experience nothing will change.

Although, we did get one prime out of it... so it's an effort that was not wasted in the past.
I think the questioner meant,"What does the term 'soap box' mean?"

It means that someone is about to vigorously try to convince a large number of people to agree with one of their opinions. It comes from early United States history, when people would basically go out to a street corner, stand on whatever is both convenient and causes them to be easily seen, and give a speech.

umccullough

10-27-2007, 04:51 PM

tqft

10-27-2007, 06:42 PM

Scores are nice but not my only reason.

My machine is doing 2nd pass - and not a clunker Pentium D 805 - 2 CPU's at 2.66Ghz, and it gets about 22 out of 24 hours to crunch them at the moment.

Like one or two others I suspect a prime is lurking in 2nd pass. I want to make a t-shirt this prime is mine (score no ego yes).

Also cleaning out 2nd pass - I believe we are catching up to first pass - slowly but steadily, given the relative times to complete tests it is possible, depending on the machine ratio and how well sieved and factored the numbers in between are - will give the project some finality when all 17 are in the bag. Easier to do now when a payoff of a prime will get us closer to our goal.

I want to see the problem looked at hard - when all 17 are done, I want someone to take all the data and ask why is that number the lowest that does not produce a prime ever. And get some answers - it is at least worth doing a Masters thesis on, if not more.

Also I don't know if the scoring will be backdated if/when it is changed.

Sceptic

10-28-2007, 07:28 AM

Since Zuzu's post at 20 Oct, the pending tests for 33661 has dropped from ~480 to 345 (as of writing).
How many days does a n=~14M test take on a PC these days?

Sceptic

engracio

10-28-2007, 12:48 PM

Since Zuzu's post at 20 Oct, the pending tests for 33661 has dropped from ~480 to 345 (as of writing).
How many days does a n=~14M test take on a PC these days?

Sceptic

Depending on the box, I guestimate from 10 days to 2 weeks.:thumbs: Several more days.:umm: All signs point to a prime, I hope. :D

e:)

KriZp

10-28-2007, 02:21 PM

I've been keeping track aswell :) the timestamp is HH.MM.DD.MM.YYYY Time is GMT+1, rest copy/paste from SoB.com/stats

33661 72834 tests 380 tests 11047415 14271696 22.32.26.10.2007
33661 72814 tests 403 tests 11047415 14265936 10.17.25.10.2007
33661 72802 tests 424 tests 11047415 14265936 18.32.24.10.2007
33661 72790 tests 438 tests 11047415 14259912 16.55.23.10.2007
33661 72775 tests 451 tests 11047415 14259912 02.12.23.10.2007

Jwb52z

10-28-2007, 04:46 PM

As I write this right now, the number of tests for this K is 344.

Zuzu

10-29-2007, 09:00 AM

Now it's 334 down from ~510 when I first noticed the anomaly on Oct 18. Having done systematic statistics from Oct 20 (5:19 pm EST) till now the nb of pending tests decreased from 497 to 334 and the sum completed+pending decreased from 73237 to 73202. For the other k's the nb of pending tests varied +/-30 max and the C+P increased by about 60 to 200 depending on k density.
Having browsed through Mersenne CPU benchmarks (www.mersenne.org/bench.htm), a Mersenne test with exponent 14.2 M is expected to be completed in 2.5 to 4 days for the best processors (Core 2 Duo, Pentium 4, Athlon 64) ;thus, if this is relevant for the Proth PRP tests the prime discovered is expected to be firstpass and a false positive should have been ruled out. OTOH AFAIK the complete proof takes some time, so...be patient:)

axn

10-29-2007, 10:14 AM

Having browsed through Mersenne CPU benchmarks (www.mersenne.org/bench.htm), a Mersenne test with exponent 14.2 M is expected to be completed in 2.5 to 4 days for the best processors (Core 2 Duo, Pentium 4, Athlon 64) ;thus, if this is relevant for the Proth PRP tests the prime discovered is expected to be firstpass and a false positive should have been ruled out. OTOH AFAIK the complete proof takes some time, so...be patient:)
LLR would use a FFT size of 1536K which will take roughly twice as long. Proth would take roughly 3x-4x as much as LLR. Thus, if Proth.exe is being used for verification, the ETA is 15-30 days. :eek:

engracio

10-29-2007, 10:39 AM

LLR would use a FFT size of 1536K which will take roughly twice as long. Proth would take roughly 3x-4x as much as LLR. Thus, if Proth.exe is being used for verification, the ETA is 15-30 days. :eek:

I hope they implemented the recently discussed faster prime verification. Otherwise it is something that need to be look at in the future or not. AFAIK the last prime took longer than previous primes to verify and submitted as prime. A prime is a prime is a prime.:)

Zuzu

10-29-2007, 11:05 AM

So, the PRP test duration is expected from 5 to 8 days with fastest processors ; this is consistent with the "Max tested" data: the highest n is 14322298 corresponding to a test assigned just 5 days ago. That begs me 2 questions:
1. Why is there such a difference in FFT size between LLR and Mersenne prime tests ?
2. If a false positive check takes at least 5 days what is the procedure: for a given k, if there is a hint for a prime, is this check (2nd LLR test) done before or after the stop of new tests ? I would guess it is before for reasons of "risk management" but am not sure. If my guess is correct the confirmation would be expected by ~ 10 days.

Jwb52z

10-29-2007, 02:22 PM

It's down to 328 tests now.

Sceptic

10-29-2007, 05:13 PM

So, the PRP test duration is expected from 5 to 8 days with fastest processors ;
[Some other text in between...]
If my guess is correct the confirmation would be expected by ~ 10 days.
But we don't know what hardware they run the PRP test/verification on.
We can only speculate what type of hw they have (unless someone has inside information and share it with us :) ).
My guess is: (They run the verification on a PIII 1GHz, underclocked to 900MHz + an oversized CPUcooler just to make sure the run is stable enough so the result is trustworthy)^2 :D
That's why it takes so long to confirm the prime :harhar:
(...Or they like to read our silly posts :music: )

Sceptic

robin_benson

10-30-2007, 08:47 PM

http://primes.utm.edu/primes/status.php

Zuzu

10-30-2007, 09:21 PM

A secondpass ??? 'Tis the season, see Riesel ! OTOH 15 days to check and still "unverified because in process" in CC list ???:D And only #10, not the so hoped #6:cry:
Anyway if confirmed, very good news for the project and...:whistle:

vjs

10-30-2007, 10:48 PM

where is that soap box...

:weggy:

axn

10-30-2007, 11:17 PM

Dang! That was a twist in the plot! But why is it being verified by prime pages? Aren't all SoB primes supposed to be "Externally verified", meaning Prime Pages just accepts it on the project's word???

Jwb52z

10-30-2007, 11:49 PM

Well axn, maybe I am just completely wrong here, but maybe the prime is of significant size or has a certain quality that they want an additional check for accuracy. How many digits would this prime have in general?

tqft

10-31-2007, 12:00 AM

http://www.free-dc.org/forum/showthread.php?p=115519&posted=1#post115519

Joe O

10-31-2007, 07:57 AM

axn

10-31-2007, 08:43 AM

Well axn, maybe I am just completely wrong here, but maybe the prime is of significant size or has a certain quality that they want an additional check for accuracy.

Actually, the idea is that 1) projects that find /really/ big primes will do their due diligence and 2) prime pages doesn't have the resources to verify primes above a certain size (case in point -- the current find is still being verified at the time of this writing :cry:). In this particular instance, I am pretty sure that the original positive PRP test plus the project's own verification is more than sufficient validation -- any further test is wasted clock cycles.

I see this more as a coordiation issue -- particulary since all other SOB primes have been previously verified as "External". If you see the link here (http://primes.utm.edu/primes/search.php?Discoverer=SB*&Style=HTML), you can see that, except for the smallest four, everything else has verification status as "External". You can find the definition of "External" here (http://primes.utm.edu/primes/help/page.php#status)

PS:- The current verification could take the better part of a week. I hope somebody just kills that run and marks it as "External".