mersenneforum.org 332.2M - 333.9M (aka 100M digit range)
 Register FAQ Search Today's Posts Mark Forums Read

 2016-01-22, 17:18 #606 Uncwilly 6809 > 6502     """"""""""""""""""" Aug 2003 101×103 Posts 5·7·311 Posts Here is the status report for the range from 332192831 to 332399999 (plus other data out to 332999999): Code: Date of data 2016/01/20 Average bit depth for first 100 expos 79.01 Average bit depth for first 1000 expos 78.48 Average bit depth for first 5000 expos 76.29 Average bit depth for first 10000 expos 75.39 100th active expo (no factor found) 332198443 1000th active expo (no factor found) 332247557 5000th active expo (no factor found) 332448943 10000th active expo (no factor found) 332696087 Unitless total effort number 414,523,392 Number of first 100 expos to 2^79 81 Number of first 1000 expos to 2^78 831 Number of first 5000 expos to 2^77 1917 Number of first 10000 expos to 2^76 3702 Number left in (classic) range 4027 Estimated expos in range to be removed 121 (by taking all expos to 2^79) Estimated expos in range to be removed 263 (by taking all expos to 2^82) Code: Bit # at bit level 74 151 75 983 76 1039 77 693 78 356 79 752 80 35 81 13 82 2 83 1 84 0 85 2 P-1 628 There have been a few good P-1 factors found, mainly by Aurashift. Attached image has arrows indicating the previous (last report) lowest expo for the 74, 75, and 76 bit levels. Attached Thumbnails
 2016-02-15, 21:04 #607 aurashift   Jan 2015 11×23 Posts http://www.mersenne.org/report_expon...2277943&full=1 http://www.mersenne.org/report_expon...2279839&full=1 Got a few more coming in a few weeks.
 2016-02-15, 22:06 #608 Uncwilly 6809 > 6502     """"""""""""""""""" Aug 2003 101×103 Posts 5×7×311 Posts Cool. There was a 170 bit composite factor found via P-1 recently. http://www.mersenne.org/report_expon...2259013&full=1 82 and 88 bits for the factors
 2016-02-24, 19:31 #609 aurashift   Jan 2015 11·23 Posts Thanks to the venerable MadPoo for helping me figure out a numa issue with some new servers by using AffinityScramble2, I've currently got the ability to bring in 32 100M digit exponents every ~50 days. (at least until these servers get used for something.) edit: that is, once they start LL. Last fiddled with by aurashift on 2016-02-24 at 19:33 Reason: I wanted to.
 2016-03-04, 23:36 #610 TObject     Feb 2012 19516 Posts Holly Batman, aurashift! Can I be your friend?
2016-03-05, 20:00   #611
aurashift

Jan 2015

25310 Posts

Quote:
 Originally Posted by TObject Holly Batman, aurashift! Can I be your friend?
sure, but there's a vetting process and registration fee.

 2016-03-22, 21:01 #612 aurashift   Jan 2015 FD16 Posts I have a good feeling about 333,333,223 and 333,333,307. Or I just think they look neat. About 25 days before we start seeing results... lost 6 blades to projects (12 sockets)
 2016-03-23, 00:24 #613 TObject     Feb 2012 34·5 Posts So, if it is one exponent per blade, each blade is 15-20% faster than a Titan [at the 100M digit range LL], plus the ECC memory benefit. If it is two exponents per blade, then it is another 50% faster. What does a system with 32 such blades cost? What does a system with 16 such blades cost? Last fiddled with by TObject on 2016-03-23 at 00:29
2016-03-23, 02:53   #614
aurashift

Jan 2015

FD16 Posts

Quote:
 Originally Posted by TObject So, if it is one exponent per blade, each blade is 15-20% faster than a Titan [at the 100M digit range LL], plus the ECC memory benefit. If it is two exponents per blade, then it is another 50% faster. What does a system with 32 such blades cost? What does a system with 16 such blades cost?
It is a single blade with two sockets, which can handle two exponents. Each socket has 16 physical cores (this model anyway...you can max out at 18 cores per socket right now).

As for cost... let's just say metric fktons. (...really...at 16-32 the blades and the infrastructure...hmm...think cost of a house. I'm privileged to get to use these until they get provisioned.)

I'm glad you commented, I was just wondering what the GPUs could do in the 100M range the other day...

Last fiddled with by aurashift on 2016-03-23 at 02:55

 2016-03-23, 18:36 #615 TObject     Feb 2012 1100101012 Posts Yea, a Titan (the original, non-overclocked) does a 100M digit exponent in about 60 days (maybe a few days less if you are willing to be more adventurous with FFT sizes), with current CudaLucas. But it is really better to run two side-by side and compare residues along the way. Cool info on the blades. I need to win a lottery, pronto; so I can get some. ;) Last fiddled with by TObject on 2016-03-23 at 18:40
 2016-04-21, 00:34 #616 Uncwilly 6809 > 6502     """"""""""""""""""" Aug 2003 101×103 Posts 2A8516 Posts Here is the status report for the range from 332192831 to 332399999 (plus other data out to 332999999): Code: Date of data 2016/04/20 Average bit depth for first 100 expos 79.01 Average bit depth for first 1000 expos 78.52 Average bit depth for first 5000 expos 76.54 Average bit depth for first 10000 expos 75.81 100th active expo (no factor found) 332198443 1000th active expo (no factor found) 332247649 5000th active expo (no factor found) 332450803 10000th active expo (no factor found) 332700301 Unitless total effort number 457,326,592 Number of first 100 expos to 2^79 81 Number of first 1000 expos to 2^78 850 Number of first 5000 expos to 2^77 2086 Number of first 10000 expos to 2^76 3895 Number left in (classic) range 4006 Estimated expos in range to be removed 111 (by taking all expos to 2^79) Estimated expos in range to be removed 252 (by taking all expos to 2^82) Code: Bit # at bit level 74 0 75 931 76 1059 77 742 78 407 79 751 80 95 81 16 82 2 83 1 84 0 85 2 P-1 686 There has been some serious firepower thrown at this range of late. Note the rise in expos taken to 80 bits. We are now over 61% factored out for the 332M range as a whole. The first Attached image has arrows indicating the previous (last report) lowest expo for the 74, 75, and 76 bit levels. Attached Thumbnails

 Similar Threads Thread Thread Starter Forum Replies Last Post Uncwilly GPU to 72 64 2013-03-31 02:45 JuanTutors PrimeNet 8 2012-12-06 13:47 JuanTutors Lounge 6 2012-02-21 07:36 __HRB__ Riesel Prime Search 0 2010-05-22 01:17 Unregistered Information & Answers 10 2010-03-24 20:16

All times are UTC. The time now is 07:10.

Fri Jan 27 07:10:20 UTC 2023 up 162 days, 4:38, 0 users, load averages: 0.94, 0.86, 0.83