[QUOTE=Dubslow;269084]My next question then became what are the methods that George used to determine the bit depths? He hasn't been here since I asked that, but if anyone else knows, please post.[/QUOTE]

There's not too much math involved into determining the bit depths at present, if any. Going to the tabulated depth, the TF wavefront got too far ahead of the LL wavefront, and George decided to TF one bit deeper for exponents between 53M and 70M, or some such. That process has repeated at least once thus far and I expect it to be repeated time and again as long as enough people do TF.

George's original motivation was noting the TF effort for a given, fixed bit level is proportional to 1/log N. He then solved a minimisation problem: Given, say, 10,000 prime mersenne exponents, how do I eleminate them with the least effort, ON AVERAGE.

We all know that the answer is that it happens at the amount of effort on each type of work where the answer doesn't matter as to which type of work it did. On some single CPU, probably his prototypical 1GHz-Day per day CPU, he found that would happen at the bit depths he tabulated.

Along came GPUs and mfaktc, which TFs 10-100 times faster than CPUs, and the GPUs have all been crowding into the easy, fast work of TF (as opposed to CUDALucas), and you have the present situation. I expect this process to stop after 3-6 more bit levels, particularly as CUDALucas matures....and the TF work becomes equally hard as it was before, since each bit level of TF doubles the amount of compute work involved.

At the moment, I have 1 medium-power GT440 card doing TF, (the other is doing Operation Billion Digits exclusively ATM), and it is producing about as many factors as half a dozen machines of various speeds doing P-1. The haul on the 666 exponents ckdo handed me looks to be 7 or 10 factors in the last two weeks; each of these factors found represents an LL-D test that won't need to be done. I'll be done with that the day after tomorrow.

[QUOTE=TheJudger;268760]Hi!

David: what do you think is a good turnover time for TF assignments?
No matter of the runtime of current assignments I try to receive work for one week at once.

Oliver[/QUOTE]

Not wanting to be argumentative (as usual), but why are there
still 5000 53M and 12000 54M exponents so constipated?

Have GPUs been hijacked by the [url=http://www.youtube.com/watch?v=xoHECVnQC7A]Lone Rangers[/url], or have they gone
on strike?

[QUOTE=davieddy;269342]Not wanting to be argumentative (as usual), but why are there
still 5000 53M and 12000 54M exponents so constipated?[/QUOTE]

Hey, in the last week, between 50M and 60M, 3,475 exponents have been taken from below 71 "bits" to 71 "bits" or above, with 100 exponents eliminated (read: factor found).

How much more do you want?

And why are you so apparently obsessed with this?

As Dr. Silverman himself acknowledged to me here on another thread, [B][I][U]none[/U][/I][/B] of the actual [B][I][U]computational work[/U][/I][/B] done here at GIMPS means anything to mathematics.

Silverman Schmilverman
 

[QUOTE=chalsall;269344]Hey, in the last week, between 50M and 60M, 3,475 exponents have been taken from below 71 "bits" to 71 "bits" or above, with 100 exponents eliminated (read: factor found).

How much more do you want?

And why are you so apparently obsessed with this?

As Dr. Silverman himself acknowledged to me here on another thread, [B][I][U]none[/U][/I][/B] of the actual [B][I][U]computational work[/U][/I][/B] done here at GIMPS means anything to mathematics.[/QUOTE]

Is that a gun in your pocket, or are you just pleased to see me?

Mae West

[QUOTE=davieddy;269348]Is that a gun in your pocket, or are you just pleased to see me?[/QUOTE]

It's a gun. (Joe Kurtz, as written by Dan Simmons in Hard Freeze.) :smile:

[QUOTE=Uncwilly;268147]Try James's calculator: [url]http://mersenne-aries.sili.net/factorbits.php[/url]
I don't know if it has been updated recently.[/QUOTE]I do try to keep it updated whenever I hear about changes (if something changes and my data seems wrong, [i]please[/i] tell me about it!) It is currently based on the numbers that [url=http://www.mersenneforum.org/showpost.php?p=257528]George published on 03-Apr-2011[/url], and that are part of v26.6 (see commonc.h)

I have just made a cosmetic change to my factorbits page: it now shows the lookup table of TF bits vs exponent.