[quote=henryzz;146040]now drive 4 is nearly done what is the priority drive 1 or 3?[/quote]

Drive 1. I already moved 4 cores to server C443.

[quote=em99010pepe;146041]Drive 1. I already moved 4 cores to server C443.[/quote]
Yes, I guess you could say Drive 1 is the higher priority--actually, since both Drives 1 and 3 are at approximately the same n-level, they're at equal priority, but since Drive 1 is somewhat larger than Drive 3, it will require more resources to move at the same pace as Drive 3 and thus it is the higher priority.

Early this morning, I finished the last manual ranges from the 4th drive. I just moved 5 quads to port 400 1st drive. The other 4 are now catching up on other misc. reservations. As those slowly finish up, I'll look to move 1 quad to the 3rd drive if it gets behind.

Technically, we need about a 6 to 1 ratio of port 400 (1st drive) vs. 5000 (3rd drive) to keep them moving ahead about equally. That's because the 1st drive has 6 times as many k's.

Karsten and Max, I promise I'll get to the 4th drive results in the next 1-2 weeks. I've already done some preliminary sorting and verification work on them but keeping up verifications and web page updates at CRUS has been a large effort.


Gary

For example why don't we move all cores to drive 3 to finish it in one month or so? Just thinking loud, I would do that.

Carlos

[quote=em99010pepe;146145]For example why don't we move all cores to drive 3 to finish it in one month or so? Just thinking loud, I would do that.

Carlos[/quote]


That would mean the 1st drive would need a very intense effort to finish by Jan. 31st, which would offer little choice to people. Jan. 31st is the date we're shooting for now, although I still think Dec. 31st is possible. But I think we'd need more Free-DC help on several rallies to accomplish that.

But don't let me stop anyone. If you like the slightly lower k's and less tests per n-range in the 3rd drive, go right ahead.

I'm also considering starting the k=1005-2000 effort for n=50K-100K in the next 2-3 weeks. I think Bruce's sieve depth is great enough for such a low n-range. As popular as the Riesel base 3 mini drive is at CRUS, it's obvious that there is some 'pent up' demand for smaller tests that find quite a few primes. It will also be nice to start cleaning up behind Benson's mess that was left for k=1003-1400.


Gary

Delay k=1005-2000 effort for two months. You are contradicting yourself because by starting another effort you will delay drive 1 and 3.
But this figure can change if we manage to bring up a good stats page...

One thing I have to say, I hate doublechecks....

Carlos

[quote=em99010pepe;146160]Delay k=1005-2000 effort for two months. You are contradicting yourself because by starting another effort you will delay drive 1 and 3.
But this figure can change if we manage to bring up a good stats page...

One thing I have to say, I hate doublechecks....

Carlos[/quote]

Yeah, I know. Many people are averse to double-checks. That's why we try to bury them in larger efforts. The idea is that about 3-5% of CPU time is spent on double-checks so that you don't feel like you're getting 'stuck' with them.

Here is what I'm thinking about 1005-2000 at a low n-range: Doing it will likely attract more attention to the project and we'll get some people to test some higher n-ranges after doing some of the lower n-ranges. About double-checking the range; n=50K-100K is largely unsearched. Double-checks would likely amount to < 5%. Benson only did above 300K.

We'll publicly discuss the timing of unveiling k=1005-2000 beofore doing so. Waiting until the 1st and 3rd drives are done would not be a big deal.


Gary

[QUOTE=gd_barnes;146193]Doing it will likely attract more attention to the project and we'll get some people to test some higher n-ranges after doing some of the lower n-ranges. About double-checking the range; n=50K-100K is largely unsearched. Double-checks would likely amount to < 5%. Benson only did above 300K.
[/QUOTE]

Benson tested the range from about n=170k to 476k for 1005<=k<=1399 (see his prover code L80).
the range n=50k-100k was not completly tested, only some k's, so this effort is 90%-95% new and not doublecheking.

542.0-.2 is complete, no primes.
Estimate for 542.2-.4 is 11/3.
Estimate for 542.4-.6 (finishing this range) is 11/12.

reserving n=544K-547K for port 400

Taking 570K-575K for LLRnet G4000. This is so that I can have some k/n pairs to load into a new LLRnet server that Gary and I are hosting on one of his machines as a testbed for some Perl-based results handling scripts I'm developing. I'm purposely taking a range that's far above the current leading edge, so that this range won't be quite as high priority for a while. :smile:

Once I've gotten this server fully up and running, I'll post the server/port info on the forum so other can connect to it. :smile: