Funny (probably fake...) result

[lycorn]

While trying to figure out the reason for the spike in Primenet throughput, I stumbled upon these 2 lines of results:


Smok_bmv Manual testing 72533261 NF 2015-08-25 12:52 0.0 1726784.8 no factor for M72533261 from 2^79 to 2^89

Smok_bmv Manual testing 72533261 F 2015-08-25 12:52 29.3 3283.9227 5723559834261931083792841

While the 2nd line is surely genuine, the first one is complete BS.

[petrw1]

Originally Posted by lycorn

While trying to figure out the reason for the spike in Primenet throughput, I stumbled upon these 2 lines of results:


Smok_bmv Manual testing 72533261 NF 2015-08-25 12:52 0.0 1726784.8 no factor for M72533261 from 2^79 to 2^89

Smok_bmv Manual testing 72533261 F 2015-08-25 12:52 29.3 3283.9227 5723559834261931083792841

While the 2nd line is surely genuine, the first one is complete BS.

The factor is more than 82 bits....

[petrw1]

Originally Posted by lycorn

While trying to figure out the reason for the spike in Primenet throughput, I stumbled upon these 2 lines of results:


Smok_bmv Manual testing 72533261 NF 2015-08-25 12:52 0.0 1726784.8 no factor for M72533261 from 2^79 to 2^89

Smok_bmv Manual testing 72533261 F 2015-08-25 12:52 29.3 3283.9227 5723559834261931083792841

While the 2nd line is surely genuine, the first one is complete BS.

The factor is more than 82 bits....

The fastest GPU with stats here http://www.mersenne.ca/mfaktc.php does 1230G / day.
I am not suggesting he has that card .... only taking the maximum case.
And since as far as I know TF is single threaded it would take that GPU almost 4 years.
Though I suspect it was not even out 4 years ago...

Mind you it was only 1 month earlier that TheJudger got it to 75 bits to allow Smok_bmv to take it from there to 89 bits.

Hmmmmmmmmmmmmmmmmm

Even taking it from 75 - 82 bits to get that factor should take a card like that 37 days.

[lycorn]

And the factor found is 82.xxx bits, which makes a NF from 79-89 a false result. Note both tests were (?) on the same exponent

[chalsall]

Originally Posted by petrw1

Even taking it from 75 - 82 bits to get that factor should take a card like that 37 days.

A possibility is it was actually found by P-1, but submitted as TF.

[lycorn]

That´s possible, giving the smoothness of K:

K = 2^2 × 3 × 5 × 163 × 436399 × 9244351

[petrw1]

Originally Posted by chalsall

A possibility is it was actually found by P-1, but submitted as TF.

Good point. I've had that happen to me.....just nowhere near this magnitude.

[Madpoo]

I filtered some info from the raw logs (user id/computer name), but here's the info:

UID: <user>/<machine>, no factor for M72533261 from 2^79 to 2^89 [mfaktc 0.20 barrett87_mul32_gs]
UID: <user>/<machine>, M72533261 has a factor: 5723559834261931083792841 [TF:82:83:mfaktc 0.20 barrett87_mul32_gs]

The factor was indeed found by TF, according to that. And since TF work by non Prime95 clients doesn't have the checksum and is accepted on the honor system, yeah, that 2^89 entry must be bogus. Not saying he did it on purpose though... he did find a factor above 2^82 which can't be faked, so there was some work that went into this.

My guess is some cutting and pasting happened and that "89" should have been "82" ? Although every other entry was just a 1-bit range... so that's kind of weird too.

That user has taken some work above 2^80 in the past, e.g.:
M999999059

But nothing beyond 2^82. In fact, this appears to be the first time this user went beyond 2^82 at all.

Still, this user has found a lot of factors in the past, so I'd assume it was a mistake and nothing intentional.

[LaurV]

Copy/paste mistakes don't happen like that... My feeling is that the guy decided to give himself some more TF credit (i.e. intentional). He was thinking that once factored, the exponent is cleared anyhow and nobody will bother with it, and no harm done for the project either. He was just a bit greedy to go to 89, just with an 84 or 85 and no one would observe

[axn]

Originally Posted by Madpoo

he did find a factor above 2^82 which can't be faked, so there was some work that went into this.

As indicated above, with high probability, it was a P-1 find.

[Madpoo]

Originally Posted by axn

As indicated above, with high probability, it was a P-1 find.

Except for this:

UID: <user>/<machine>, M72533261 has a factor: 5723559834261931083792841 [TF:82:83:mfaktc 0.20 barrett87_mul32_gs]

Would mfaktc report "TF:82:83" if it was really found by P-1?

Or did you mean he may have faked that line?

It still seems like, based on the user's past history, that it would be out of place to purposely edit that, but yeah, it's hard to see how that could have been munged accidentally with cutting and pasting.

Anyway, summary is, we know it wasn't done from 79-89, and at most that would have been 79-82 since the factor was found in 82-83.

I'm not sure what it would take to adjust everything. Correcting the log entry and the factored-to level is easy enough, but then there are some factoring credits, etc. I'd have to leave that to George, so I'm happy to wait on his input.