When will NPLB be the top project...

[Flatlander]

My guess for 4th by score: Mar 3rd 2010.

Yes I did misunderstand your BBQ ref. Gary, but I was only joking anyway.
(btw I hate real BBQs. Somebody fetch the Senna Tablets! )

[Brucifer]

Originally Posted by gd_barnes

You guys both missed my meaning of BBQ here. I wasn't referring to an NPLB rally with Free-DC BBQing everyone. I was referring to a BBQ where NPLB BBQ's RPS for 1st place on # of primes!!

Regardless, the point should be clear. BOINC would only be used for extremely annoying huge tasks that are otherwise relatively thankless and that few people would consider doing. Double-checking might be another example.

Gary

Thank you for clarifying the BBQ thing, I too misunderstood what you had said.

I haven't looked at the drives listing before posting this, but once we finish wading through the port 9000 stuff, if there isn't already a port set up for doing double check work then put one up and I'll start running double check stuff. All that kind of stuff is still part of the project and has to get done too.

[mdettweiler]

Originally Posted by Brucifer

Thank you for clarifying the BBQ thing, I too misunderstood what you had said.

I haven't looked at the drives listing before posting this, but once we finish wading through the port 9000 stuff, if there isn't already a port set up for doing double check work then put one up and I'll start running double check stuff. All that kind of stuff is still part of the project and has to get done too.

Well, currently we have the Doublecheck Drive #1 (k=3-1001, n=100K-260K) which we would like to have completed by the end of the year, but there's a catch: all the machines run on it must be reasonably guaranteed to be stable. This is because we don't have first-pass residuals for any of those numbers, so we have to "blindly" trust the quality of the doublecheck residuals, so to speak.

This is why we don't put that drive on a server--because otherwise there's no way we could control who connects to it, possibly with a machine that hasn't undergone stability testing.

If you're absolutely sure of the stability of your machines, then we can set up a private server for you loaded with doublecheck work. That would be great since the doublecheck drive has been moving rather slowly.

I noticed you sent me a PM about this--I'll respond to that with some more details regarding this.

Max

[henryzz]

could we do a pm-to-get-the-port server for the current doublecheck effort maybe?

[mdettweiler]

Originally Posted by henryzz

could we do a pm-to-get-the-port server for the current doublecheck effort maybe?

Hmm...that's not a bad idea. However, then we have the problem of the port number showing up in the stats...

One option would be to give the doublecheck server a different port number in the stats than the one you connect to. For example, it could be listed in the stats as GB100 (which is a port we would never use anyway since it's <1024 and thus can't be run as a non-root user on Linux), yet in reality the port number would be something rather more obscure.

This wouldn't be hard to do, since all I'd have to do is tell my results copy-off script to put a different port number in the filename (which is where the DB gets that information). This is, of course, assuming that it would be run as a GB server; David, I presume it would be similarly possible to do something like this on one of your servers?

[IronBits]

Anything is possible

[MooooMoo]

Originally Posted by MooooMoo

...Now, the difference between the top two projects are only 32 primes (923 vs. 891).

Wow!

There's a huge battle going on for that #1 spot. Half a month after my original post, the number of primes in the top 5000 list found by RPS and NPLB are 985 and 960. The gap between both projects was under 10 primes a few days ago, but RPS has pulled away a bit recently.

Also, this is the first time that more than half of the top 5000 primes were found by only three projects - RPS (985), NPLB (960), and Primegrid (559).

[gd_barnes]

Have you taken a look at RPS lately? If so, you'll know why. They have 4 relatively new low n-range drives, all created in the last 3 months plus 2 more that they are working on that aren't formally announced...k=3400-4000 and k=6000-7000.

We actually tied them for 15 mins. and then poof, it was gone. lol You can see in the k=1400-2000 n>350K primes thread from a little over a week ago when a prime was submitted that gave us the tie.

They did what they had to do and are now pulling away again. My hat's off to 'em. If we somehow are able to catch them again, it will likely be after PrimeGrid passes us both.

It would be nice if RPS would attempt to fill in the non-top-5000 ranges of the k's that they search. We spend about 15% of our resources double-checking and another 15% doing new non-top-5000 ranges and will not stray from that just to get to the top spot. Bruce, alone, has likely upwards of 25-30 cores or more just searching k=1801-1999 for n=200K-350K right now and has probably completed 70% of the entire range of k=1401-1999 for that same non-top-5000 range. Going after the top spot was just a "side effect" of the efforts that we do and not a specific goal that we had in mind. I think that it is best for the prime searching community to fill in all of the ranges and then do double checks at the low n-ranges; slowly creeping upwards on those as computer speeds/capacity increase.

On another note: I'm still not particularly fond of a thread of this nature. I let it stand this time because I said I'd be OK with it to you if we got within 30 primes of them and even added my own 2 cents and prediction to it. But I don't suggest elaborating on it much in the future. Thanks.


Gary

[Brucifer]

I wouldn't really call this a race to be #1 as we have multiple people here that have not been concerned with the top-5000. We've had several involved in sieving, double-check, and cleaning up low weight holes. If in fact the project was going for #1, then we wouldn't have this other stuff going on. :-) But then that wouldn't be keeping with the project name and goals. :-)

[Flatlander]

Oh dear.

[kar_bon]

is correct again!

RPS = 984
NPLB = 969