Bases 6-32 reservations/statuses/primes

[LaurV]

Wow! Very nice hit, congrats!

[KEP]

Originally Posted by LaurV

Wow! Very nice hit, congrats!

Thanks ... yes, it was a nice hit, specially because it was a low weight k, so theoretically the odds of reaching the goal of proving these stubborn conjectures should have gotten easier

[MyDogBuster]

I'm stopping R24 and S24 at 260K due to lack of resources.
I'm losing computers left and right and I'm down to 4 from 14. (Q6600's and the like)
Most of them over 12 years old.

R24 tested 222K-260K
6 primes found 68 remain

S24 tested 222K-260K
2 primes found 71 remain

Results emailed - Bases released

[gd_barnes]

Sounds like me. I'm down to 7 from 12. Most 8-10 years old.

[henryzz]

Originally Posted by gd_barnes

Sounds like me. I'm down to 7 from 12. Most 8-10 years old.

For primality testing 6700k is about 5x the speed of a Q6600. You would save on power quickly by replacing with 2-3 pcs.

[mdettweiler]

Originally Posted by henryzz

For primality testing 6700k is about 5x the speed of a Q6600. You would save on power quickly by replacing with 2-3 pcs.

I can attest to this. I'm mostly crunching with Haswells (4xxx's) these days, hopefully soon to be supplemented by some recent 6xxx/7xxx (Skylake and Kaby Lake) models. Though I'm not paying for power on these, strictly on the basis of computing power I've been downright amazed at what they can do. The big reason for this (as I understand it) is the addition of the FMA3 instructions, which are extremely well-suited to the sort of computations involved in LLR/PRP testing. (For this reason, 4/5/6/7xxx models are pretty close in terms of speed, at least in my observations - the big boost came with the 4xxx series which added FMA3.) Just one of these machines makes everything that came before it look like chump change for LLR.

For comparison - the machines I'm allowed to use for this at work (the bulk of my resources) consist of a roughly 50/50 split between first-gen Core i5/i7's (Nehalem - Gary, these are like your "God Almighty" box) and fourth-gen (Haswell, i.e. 4xxx's). The rough numbers I ran showed that one Haswell is worth about 8 - yes, eight - Nehalems. At this point I barely even bother crunching on the Nehalems because they're not worth the management effort of having to restart the clients when the machines get rebooted or used by someone else (I have a script set up to kill the prpclients whenever someone else logs on).

Incidentally, thanks to these machines, port 1400 is closing in on n=5M in (I estimate) about 8-10 days. How's that sieve coming for the next range? (I was on track to finish this range last year by end of summer, but that plan vanished because I lost access to most of the machines for a while. I've got many of them back now, hopefully long-term.)

[gd_barnes]

Originally Posted by mdettweiler

Incidentally, thanks to these machines, port 1400 is closing in on n=5M in (I estimate) about 8-10 days. How's that sieve coming for the next range? (I was on track to finish this range last year by end of summer, but that plan vanished because I lost access to most of the machines for a while. I've got many of them back now, hopefully long-term.)

Good timing. The sieve to P=1.1P (1100T) is scheduled to finish early on the 19th so I will load up n=5M-7M sometime that day.

[MisterBitcoin]

S15 k-range 50M-100M reached n=25K. There are 480 k´s remain in this range. Results mailed.

The next part (100M-1G) is at 9K.
The S7 range is at 12,6K.

[Cruelty]

S10 tested till n=2M - sieving
R10 tested till n=1.79M

[MisterBitcoin]

Reserving S25 from n=300K up to 350K.

[MisterBitcoin]

Originally Posted by MisterBitcoin

Reserving S25 from n=300K up to 350K.

I´m releasing this reservation now, due to very slow progress, it will need nearly 1 year to reach n=350K. I recommend usage of BOINC on this base.

Results attached.