LMH status

[garo]

And for the first time in recent memory, the Recent Results page shows a full hour's worth of results!

[petrw1]

Originally Posted by S485122

Do you mean "All exponents (above 5 M) are either factored to 63 bits or assigned." ? Between 1 M and 5 M there are a lot of (unassigned) exponents that have only been factored to 61 or 62 bits and afew not even tot that level...

Jacob

I actually meant 10M...but 5M works too.

[petrw1]

Originally Posted by garo

And for the first time in recent memory, the Recent Results page shows a full hour's worth of results!

Except when GIMPS Visualization (and a couple other big players) drop a whole bunch of Manual Assignments results in...they we might only see 5 or 6 minutes.

Some day, when the TF lit levels are high enough (68?) even they will have trouble filling the full hour.

[markr]

Everything above 5M is now done to 63 bits.

[petrw1]

I recall that at some point in the last year or so chalsall commented that if we could factor ALL exponents to 64 bits or higher it would clean up his visualization graphs/charts.......or something to that effect.

So just because this kind of stuff is fun for me I did some estimating (thanks to http://mersenne-aries.sili.net/credit.php).

To take all exponents to 64 bits the highest of which is in the 8M range and the lowest is our friend 1061 - - -
would take about 435,000 GhzDays (almost the sum total of the top-5 lifetime TF'ers)
of which almost 300,000 is for those exponents below 1M
of which about 157,000 is for those below 100K
of which about 63,000 is for those below 10K.
and 2776 just to get 1061 to 64 bits.

PS I DO realize all the reasons why doing so would have little overall benefit to the project beyond personal interest and I am NOT seriously considering doing so.

[mdettweiler]

Originally Posted by petrw1

I recall that at some point in the last year or so chalsall commented that if we could factor ALL exponents to 64 bits or higher it would clean up his visualization graphs/charts.......or something to that effect.

So just because this kind of stuff is fun for me I did some estimating (thanks to http://mersenne-aries.sili.net/credit.php).

To take all exponents to 64 bits the highest of which is in the 8M range and the lowest is our friend 1061 - - -
would take about 435,000 GhzDays (almost the sum total of the top-5 lifetime TF'ers)
of which almost 300,000 is for those exponents below 1M
of which about 157,000 is for those below 100K
of which about 63,000 is for those below 10K.
and 2776 just to get 1061 to 64 bits.

PS I DO realize all the reasons why doing so would have little overall benefit to the project beyond personal interest and I am NOT seriously considering doing so.

How many GHz-days can a high-end GPU (say, GTX 460) crank out in a day? In other words, how many GPU-hours do all those figures translate to? I do wonder if some of them would come down into the realm of feasibility that way.

[ET_]

Originally Posted by mdettweiler

How many GHz-days can a high-end GPU (say, GTX 460) crank out in a day? In other words, how many GPU-hours do all those figures translate to? I do wonder if some of them would come down into the realm of feasibility that way.

I'm testing 1000037 from 63 to 64 bits just for fun on my GTX275.

It will take about 45 minutes to crunch 1.963239 GHz-days.

Or 62.823648 GHz-days per day.



Luigi

[chalsall]

Originally Posted by petrw1

I recall that at some point in the last year or so chalsall commented that if we could factor ALL exponents to 64 bits or higher it would clean up his visualization graphs/charts.......or something to that effect.

LOL... That's not exactly what I said. But your analysis is interesting. And it should be clear that anyone trying to do this kind of work should not expect to find any new (or, at least, very, very few new) factors.

What I did say is that having the lower bounds brought up would allow me to raise the lower bound for the graphs. Specifically those exponents <60.

I actually have one of my quad cores working on this (with the permission of George) using a self-compiled executable (the publicly available executable won't accept such low exponents). I send George the results from time to time.

Unfortunately (for GIMPS), I've become interested in distributed photorealistic rendering, so that machine (and several others) are very busy rendering scenes most of the time. The machine still works these low exponents when it's not busy rendering....

[mdettweiler]

Originally Posted by ET_

I'm testing 1000037 from 63 to 64 bits just for fun on my GTX275.

It will take about 45 minutes to crunch 1.963239 GHz-days.

Or 62.823648 GHz-days per day.



Luigi

Okay, so using a figure of 62.823648 Ghz-days/day, that means that it will take:

6924 GTX275-days to complete all exponnets to 64 bits
4775 GTX275-days for those exponents below 1M
2499 GTX275-days for those below 100K
1003 GTX275-days for those below 10K
44 GTX275-days for M1061

Anyone with a GPU up to a 44-day M1061 job?

Oh, and we can get all exponents up to 64 bits in a year with 18 GPU crunchers working full time on it.

[ET_]

Originally Posted by chalsall

Unfortunately (for GIMPS), I've become interested in distributed photorealistic rendering, so that machine (and several others) are very busy rendering scenes most of the time. The machine still works these low exponents when it's not busy rendering....

POV-Ray? I used id a lot in the last 10 years to render scenes and animations...
It was PVM-aware. I guess how could be translated into CUDA...

Luigi

[ET_]

Originally Posted by mdettweiler

Okay, so using a figure of 62.823648 Ghz-days/day, that means that it will take:

6924 GTX275-days to complete all exponnets to 64 bits
4775 GTX275-days for those exponents below 1M
2499 GTX275-days for those below 100K
1003 GTX275-days for those below 10K
44 GTX275-days for M1061

Anyone with a GPU up to a 44-day M1061 job?

Oh, and we can get all exponents up to 64 bits in a year with 18 GPU crunchers working full time on it.

The good news: GTX 460 and higher are nearly 3 times faster than my GTX 275.
The bad news: mfaktc only works with exponents higher than 1,000,000

Luigi