how long will the factor count go?
 

Hey I just started with this program and are now factoring M34430833 to 2^68.

I guess I`m checking if 2^34430833 - 1 is a prime or not. I guess I need to check all primefactors that is less than sqrt of the number but that would be rather huge wouldn`t it (like 2^17215415).

Well how far will it go? Looking at the time left it seem like it will come to 2^97 or something :) Or is this just a "second stage" of the process for this prime? (I guess some1 else been checking the ones up to 2^63)...

Thank you for a reply :smile:

[QUOTE=Xerat]Hey I just started with this program and are now factoring M34430833 to 2^68.

I guess I`m checking if 2^34430833 - 1 is a prime or not. I guess I need to check all primefactors that is less than sqrt of the number but that would be rather huge wouldn`t it (like 2^17215415).

Well how far will it go? Looking at the time left it seem like it will come to 2^97 or something :) Or is this just a "second stage" of the process for this prime? (I guess some1 else been checking the ones up to 2^63)...

Thank you for a reply :smile:[/QUOTE]


Hi,

You are *not* checking to see if it is prime. What you are doing is checking
to see if it has any small prime factors. (which would prove it composite).
If none are found, then a prime test will be done. :bounce:

oki got my answere at [url]http://www.mersenneforum.org/showthread.php?t=666[/url]
I guess :)

p4 1.8 takes about day and a half going to 68

Trial factoring to 2^68 on a Celeron D
 

You are right in the answer you found that the first test phase does trial factoring usually up to 2^68 (but maybe not always it can depend on some things like the size of the exponent, there's a formula/table somewhere). You can get the client to test up to a user specified depth but this tends to confuse the primenet server when it gets the assignment back so is best left alone if you are new.

As you're new, you might find it worthwhile to note your testing started probably around 2^63 because some other person or machine has already checked for any smaller factors before the more lengthy work of bigger numbers was assigned to you.

Trial factoring to 2^68 (only) that is without the subsequent stages or LL testing takes about 46 hours on my Celeron D running at 2.7 GHz. That's slower than Moo's P4 :-(

LL tests take a long time (getting on for 2 months depends on your PC) but I have some machines doing only trial factoring which means my stats get raised quite often, and this maintains my interest between the LL updates.

If you aren't interested in the BIG prize money you needn't do LL testing for 10million digits and above, you could choose to do smaller LL tests instead. The LL test time is roughly proportional to the size of the exponent because it has to go through that many iterations to get the final answer prime or not. As well as this effect, the bigger math takes a little longer per iteration in ranges as numbers are progressively bigger (the way the FFT multiply works). That's why I said ROUGHLY.

If you find a smaller prime among these smaller exponents eg 26 million approx(under approx 34 million which is where the size becomes 10m digits) you might win a small prize when (assuming) we subsequently also find the first 10m+ digit prime (and before anyone else). This is all explained in the Gimps rules off [url]www.mersenne.org[/url]. Such LL tests maybe take 2/3 the time to complete an LL test (depends on the particular exponents, obviously).

From what I remember, once the factoring (early) stages are done, primenet server will record your factoring credit right then before your LL test is finished.
You can therefore at least have some finished work to be proud of on your stats page, although you'll have that long wait before your first complete LL test shows there. If you find that is too long, you could change the time of work you request to smaller LL tests (or double checks which are also likely to be smaller) or factoring only. The client will carry on with the exponent you have queued already, but when it grabs some more work (typically happens about 5 days before finish but configurable) it will obtain and start on the kind you want to be doing.

I would suggest you leave your test running now that its started and stick out the wait, but if you really don't fancy the long waiting, it is possible to stop and return the assignment nicely. To do this you use the advanced menu after putting in password 9876 which enables extra menu items. One of them lets you return the exponent back to the server, but you won't get credit for it.

The only reason you might want to do this is if you initially selected LL testing and with hindsight think factoring only might be a good place to start. Beginning with factoring is suggested in George's documentation anyway, and lets you leapfrog quite a few slow or dormant entries in the factoring producers league table in only a few days/weeks of effort.

In any case, welcome to our collective effort. If you have any other PCs in the house feel free to get them working too.

Are you sure?
 

[QUOTE=moo]p4 1.8 takes about day and a half going to 68[/QUOTE]

Moo, when you say ABOUT do you mean +/- 3 days? LOL

Unless it's a massive overclock beyond 1.8 GHz.

My Celeron is a 2.4 running at 2.7 with dual channel DDR3200 and takes 46 hourish to TF^68 exponents circa 28.5 million.

I have a 3GHz P4 northwood which does work faster than the Celly, but the ratio is not massive, and it's known the P4 isn't very efficient at trial factoring (for a given clock speed compared to say P3).

I got my timings subtracting reporting dates in the prime.log file so they are fairly accurate. If you did this a while ago, then because the exponents available for test these days are bigger, it could be taking longer than you remember unless you benchmark it again from the log for recent tests.

It does depend on what exponent you got to test, but I would think a 1.8 running at 1.8 would take 2 days or more? Just interested. If true you have a mean machine. What do you feed it on?

Moo is a bit off on his time estimate, unless he is only referring to the delta from 2^67 to 2^68. Even then, I suspect it would take longer...

P4's are actually very good at TF above 2^64 when SSE2 kicks in; my P4 1.6GHz is quicker than my AthlonXP 1.6GHz, which would be faster than a 1.6GHz P3.

oki, thx :)

Nah I`ll leave the test running. I got an exam in 1 month and shouldn`t play any games before that.. lol
I tried found the prizelist, but didn`t found it :(.

Cash prizes /rules
 

$$$

The prize rules are explained at [url]www.mersenne.org[/url] click on the "Prizes" menu item at the top left.

From what it says there 3 prizes of $5K have been "won" (no payout until we find the big one though) so there's another $5K waiting for a prime under 10 million digits.

I *think* this is still true - I haven't compared the discoverers lists with the example payout George described (I leave this as an exercise for the reader).

Actually George's rules on payout though generally clear are a little ambiguous on this point. I think it's up to $20K at $5K per prime (in order of discovery ie the first discoveries get prizes) but it's not clear whether subsequent discoveries of bigger ones bump these discoverers out of the prize list (probably not).

Even if there's no money left for finding <10m digit you would still have the fame of discovering it and for a given amount of CPU you are more likely to find a prime by LL testing smaller exponents than the massive ones.

The choice is yours.

[quote]Each new Mersenne prime will receive a [b]maximum[/b] of $5,000[/quote]

3 primes < 10M: 5k each
4 primes < 10M: 5k each
5 primes < 10M: 4k each
...

So yes, the "will receive a maximum" can be interpreted at "they definitely get it" - but I think that's not what's meant.

thx :)

Another queastion on the same subject. Will the LL test stop before 100% if it`s not a prime?