mersenneforum.org RPS 11th Drive: Search for more megabit primes
 Register FAQ Search Today's Posts Mark Forums Read

 2013-08-27, 08:43 #144 Lennart     "Lennart" Jun 2007 25×5×7 Posts At the rate at which primes have been added to this list in the past, the 307855 digit prime "2217*2^1022660-1" may only be on the list for about 4 weeks. 2217*2^1022660-1 is Prime Lennart
2013-08-27, 10:40   #145
pinhodecarlos

"Carlos Pinho"
Oct 2011
Milton Keynes, UK

121D16 Posts

Quote:
 Originally Posted by Kosmaj Although, frankly speaking, with my limited resources I'm not that interested in searching for primes that will last only a few weeks on Top-5000.
Me too.

 2013-08-27, 11:45 #146 Lennart     "Lennart" Jun 2007 25·5·7 Posts I think it is best to leave this range or cut it down to 2000
 2013-08-27, 11:49 #147 pinhodecarlos     "Carlos Pinho" Oct 2011 Milton Keynes, UK 4,637 Posts I think this drive should be shut down. Leave the numbers to NPLB. Let's concentrate on k's<300. Carlos
2013-08-27, 12:24   #148
Thomas11

Feb 2003

111011100012 Posts

Quote:
 Originally Posted by Lennart I think it is best to leave this range or cut it down to 2000
A cut down to a smaller number of k might be a good option (instead of abandoning the whole drive). But maybe 2000<k<2100 is still too large. We could also consider to pick just the high weights, or (alternatively) the k's which had been primeless so far.

 2013-08-27, 12:38 #149 pinhodecarlos     "Carlos Pinho" Oct 2011 Milton Keynes, UK 10010000111012 Posts Let's look at our Top 5000 primes. The first prime with k 2xxx appears in position 184, the second in position 2110, and so on, so we are wasting time with this drive.
2013-08-27, 13:07   #150
Thomas11

Feb 2003

77116 Posts

Quote:
 Originally Posted by pinhodecarlos Let's look at our Top 5000 primes. The first prime with k 2xxx appears in position 184, the second in position 2110, and so on, so we are wasting time with this drive.
Carlos, I agree with you that the most economic search would be to concentrate on the smallest k's only (let's say k<300). But shall we therefore drop all other drives?

Of course, we currently have quite a few drives - maybe just too many.
If we are about to abandon any, the 11th drive would be the first to consider.

But as long as there is interest in this drive, we should keep it alive, but without jeopardizing the other drives by drawing too much computing power...
(There might also be individuals who don't care about their Top5000 score and who crunch "their" numbers just for fun.)

Regarding a cut down to a smaller number of k's:
I checked the 7th drive (which covers the range k=2000-3000). The average Nash weight is 2901, with individual weights ranging from 1659 (k=2265) to 3883 (k=2835).
In comparison, the 11th drive contains quite a few k's with even larger weights. Here are the largest ones with weight>3000:
Code:
 k  weight
----------
2289 3107
2061 3124
2271 3180
2079 3221
2085 3345
2127 3437
2259 3437
2253 3507
2205 3532
2103 3935
2043 4049
2211 4509
So, how about picking just those 12 k's, driving them to n=1290k and merging them into the 7th drive?

2013-08-27, 13:27   #151
pinhodecarlos

"Carlos Pinho"
Oct 2011
Milton Keynes, UK

463710 Posts

Quote:
 Originally Posted by Thomas11 Carlos, I agree with you that the most economic search would be to concentrate on the smallest k's only (let's say k<300). But shall we therefore drop all other drives?
Yes, drop all drives above k>300. That's my opinion.

 2013-08-27, 14:39 #152 Beyond   Dec 2002 44410 Posts I was planning on alternating between the 8th and 11th drives to help advance both drives. My thinking is we only needed to get both drives to n=1.20M to stay ahead of PrimeGrid's Proth search which is producing many more primes than their SG search. Proth is the one we need to worry about. The SG search could go on for years and will never increase in digits so we can catch up and pass at some point down the road. The 8th drive once ahead of the proth search can be easily be maintained with just a few machines active. Pulling ahead should only take a couple of weeks at current rate. The 11th drive is a bit more problematic due to the number k's involved. Per Lennart's post with his 100 cores engaged it will be a long slog to reach our goal of overtaking the proth search. I could bring 50+ cores to the effort, more once the summer is over, but when Lennart's and my interest wains we would be stuck losing ground again. So, I would vote to cut the number of k's to a more manageable level, but what that level should be?
2013-08-27, 14:57   #153
Thomas11

Feb 2003

3×5×127 Posts

Quote:
 Originally Posted by Thomas11 So, how about picking just those 12 k's, driving them to n=1290k and merging them into the 7th drive?
From counting the number of candidates in the sieve file it turns out that picking just those 12 high weight k's from the original 126 k's would reduce the work load to about 20% (19.05%).

Picking the range k=2000-2100 would be roughly 33%. Quite comparable...

I would vote for one of those options.

BTW.: I could also bring a few cores (about 20) into this effort.

Last fiddled with by Thomas11 on 2013-08-27 at 15:00

2013-08-27, 23:09   #154
Lennart

"Lennart"
Jun 2007

100011000002 Posts

1021-1025k complete 2 Primes

2171*2^1023034-1 is Prime

Lennart
Attached Files
 Drive_11_1021-1025k.log.zip (282.3 KB, 41 views)

 Similar Threads Thread Thread Starter Forum Replies Last Post Kosmaj Riesel Prime Search 372 2020-05-31 13:38 Kosmaj Riesel Prime Search 455 2020-03-20 07:50 pinhodecarlos Riesel Prime Search 132 2017-05-14 02:39 Kosmaj Riesel Prime Search 20 2011-11-17 04:10 lsoule Riesel Prime Search 140 2009-03-02 15:01

All times are UTC. The time now is 17:39.

Wed Aug 12 17:39:47 UTC 2020 up 26 days, 13:26, 0 users, load averages: 1.93, 2.08, 2.16