PRPnet Servers for CRUS

[gd_barnes]

This server has been running for years and never had a problem.

[henryzz]

Originally Posted by gd_barnes

This server has been running for years and never had a problem.

Client issue

[gd_barnes]

Sieving for riesel/sierp even/odd k/n is complete to P=1100T. I have loaded n=5M-7M into port 1400.

[mdettweiler]

Originally Posted by gd_barnes

Sieving for riesel/sierp even/odd k/n is complete to P=1100T. I have loaded n=5M-7M into port 1400.

Have you (or anyone else) run the numbers on how many primes are expected in this new range? What about n=3.1M-5M (the approx. gap since the last prime), or n=3.1M-7M?

Maybe my intuition on this is wrong, but it seems like we've hit an especially long "dry spell" on these bases. We had two primes very close together around n=2.2M, then another around n=3.1M, but none since all the way up to (nearly) 5M.

Since the "dry spell" largely represents a range which has been predominantly crunched by me, it's almost enough to make me suspicious there's something funny going on with my machines...but I feel like this reasoning can't be right, since I have no reason to suspect the machines. They're non-overclocked, off-the-shelf Dell desktops; between that and them being modern Intels, which are designed to automatically throttle very well, it seems unlikely that they would be unstable. But unfortunately, they do not lm-sensors installed (they're Linux boxes) and I don't have root so I can't confirm this with hard numbers.

(After we hit n=5M I plan to run them on NPLB port 2000 for a while to just to prove to myself that they can indeed find primes...for my own peace of mind. If the machines are good, the primes should come fast and thick over there...)

[gd_barnes]

There has not been a lot of interest in the CRUS PRPnet servers so what I would like to do is stop port 1300 when base R23 hits n=1.95M and use it for something else. I would ask BOINC or an individual to finish n=1.95M-2M.

We would still keep port 1400.

Thoughts anyone?

[rebirther]

Originally Posted by gd_barnes

There has not been a lot of interest in the CRUS PRPnet servers so what I would like to do is stop port 1300 when base R23 hits n=1.95M and use it for something else. I would ask BOINC or an individual to finish n=1.95M-2M.

We would still keep port 1400.

Thoughts anyone?

Yes, I can take over the rest for BOINC.

[MyDogBuster]

Gary, you can have 1300 now. That last test as abandoned. (machine acting up)

1947784 would be the first test

[gd_barnes]

Port 1300 has been brought down. It looks like Max did n=1947784 after you abandoned it. So the first test will be n=1947856.

Reb, attached is the remaining file for R23 n=~1.95M-2M. I am assuming that BOINC is reserving it now.

[mdettweiler]

Go figure, my machine actually did that last test on 1300 by accident. It's been running on 1400 and grabbed the 1300 test on fallback.

Late to the party but yeah, sounds like a fine plan to me. I didn't have a burning desire to crunch R23 myself, I just like to see the bases <=32 with few-ish k's remaining worked on. Glad to hear BOINC will finish off the range since I haven't had time lately to manage the machines I was running it with.

I don't suppose BOINC would be interested in spending a little time on some other few-k bases <=32? Browsing the pages, the ones available are:

R22 - 1 k at 2M
R27 - 1 k at 3M
R28 - 4 k's at 1M
R30 - 9 k's at 500K, sieve file for 500K-1M
R31 - 2 k's at 1M

S6 - 11 k's at 2M, sieve file for 2M-4M
S9 - 1 k at 2M
S17 - 1 k at 2M, sieve file for 2M-5M
S22 - 1 k at 2M, sieve file for 2M-5M
S26 - 2 k's at 1M
S27 - 1 k at 2M
S28 - 2 k's at 1M
S30 - 2 k's at 500K, sieve file for 500K-1M

S30 looks like an especially attractive target - only 2 k's but still at 500K, and with a sieve file. BOINC could probably knock that out really fast.

A lot of these are low-weight, so the odds of knocking any one of them out aren't as high as they might seem...but if we got them all to (say) 5M, we'd probably get one or two conjectures proven, which would be pretty cool considering these bases have been targets since the beginning of the project.

[rebirther]

Originally Posted by mdettweiler

S30 looks like an especially attractive target - only 2 k's but still at 500K, and with a sieve file. BOINC could probably knock that out really fast.

A lot of these are low-weight, so the odds of knocking any one of them out aren't as high as they might seem...but if we got them all to (say) 5M, we'd probably get one or two conjectures proven, which would be pretty cool considering these bases have been targets since the beginning of the project.

Put it on the recommended list, I will grab them as soon as possible, should take some time.

[gd_barnes]

Originally Posted by rebirther

Put it on the recommended list, I will grab them as soon as possible, should take some time.

I'd rather see many of our 1k bases that are only searched to n=200K pushed higher for much easier proofs so I'm not going to be recommending them. But people are free to do as they wish.

On these sieve files for bases <= 32 searched to a high depth we should check their sieve depth before running them. Under sieving means a lot of wasted CPU cycles.