Team drive #3: k=300-400 n=260K-600K

[gd_barnes]

Originally Posted by AMDave

Ah. Thanks.
I'm hoping to find a prime in there somewhere.
7 thousand plus done and still dry.

The average expectancy on a sieve to P=5T for drives 1 and 2 is:

n=300K; 1 in 4000
n=350K; 1 in 4660
n=400K; 1 in 5325

For drive 3 on the sieve to P=6T, they are barely better at 1 in 3970, 4630, and 5290 respectively.

So you've been unlucky so far. My luck is strange. I'm going to stop doing manual reservations if something doesn't change. Bad carma! I'm at 1 in ~9000-10000 on manual reservations but 1 in ~2500-3000 on LLRnet processing. Wierd. But overall, it's barely behind where it should be.

I currently have 7 manual files reserved at n=~370K which are all ~85% done; that's ~160K tests...not a single prime in the group yet. Ugh!


Gary

[mdettweiler]

Originally Posted by gd_barnes

Carlos,

Sorry...I'm going to have to put you off just a little longer. It's borderline on whether we need more k/n pairs in IronBits' server for the rally. Probably not...but I'd rather not take a chance of having to shut down the server to load more pairs during the rally. I'd also like to avoid your testing being too far above where the rally finishes.

Let's give it another 12-16 hours or so to get some response to my question in the rally thread...perhaps by midnight GMT I'll have some files posted here.


Gary

Gary,

For my processing, I don't care whether there is a gap between two ranges that an LLRnet server is running, as long as the higher of the two ranges is inserted into the server after the lower (that's the way it usually is, probably IronBits' port 500 server was an exception). I don't know about Karsten, but I can definitely say that it won't be any harder for me to process the results even with a manual reservation inserted in between the ranges that IronBits' port 5000 server has.

So, long story short, as far as I'm concerned, go ahead, give Carlos the range now. Even if you have to put more in the LLRnet server soon, no problem.

[em99010pepe]

Originally Posted by AMDave

Ah. Thanks.
I'm hoping to find a prime in there somewhere.
7 thousand plus done and still dry.

user=AMDUsers_AMDave
[03/07/08 14:51:40]
607*2^299337-1 is prime!

[Mini-Geek]

Originally Posted by gd_barnes

I currently have 7 manual files reserved at n=~370K which are all ~85% done; that's ~160K tests...not a single prime in the group yet. Ugh!

I think you mean ~16K tests, not ~160K. 160K without a prime would be truly amazing, in a bad way.
Assuming 700 avg k, base 2, avg n 370000, sieved to 5E12 (5T), and 16K tests, that's about a 96.119% chance you'd get at least one prime in all that, with 3.249 primes expected. If it was 160K, those chances would be 99.9999999999992% and 32.489, respectively.

[gd_barnes]

Originally Posted by Mini-Geek

I think you mean ~16K tests, not ~160K. 160K without a prime would be truly amazing, in a bad way.
Assuming 700 avg k, base 2, avg n 370000, sieved to 5E12 (5T), and 16K tests, that's about a 96.119% chance you'd get at least one prime in all that, with 3.249 primes expected. If it was 160K, those chances would be 99.9999999999992% and 32.489, respectively.

LOL. A statistical geek like me. I like it. Yes, of course I meant 16K tests. I'm pretty annoyed with the ranges now. I just checked them again...over 90% complete with no primes...7 ranges total. Argh!

[mdettweiler]

Originally Posted by gd_barnes

LOL. A statistical geek like me. I like it. Yes, of course I meant 16K tests. I'm pretty annoyed with the ranges now. I just checked them again...over 90% complete with no primes...7 ranges total. Argh!

Maybe you should try running a Prime95 stress test on each of your machines to ensure that you're producing good results?

[gd_barnes]

Originally Posted by Anonymous

Maybe you should try running a Prime95 stress test on each of your machines to ensure that you're producing good results?

I'm sure they're perfect. Also, it's 4 different machines running the 7 ranges. None have ever had a false prime, never one of those wierd false '00000...3' residues, no overclocking, relatively cool room all the time, maintenance folks keep the room well dusted, etc. My work laptop with the bad memory chip has done the false prime and wierd residue thing and of course I don't use it for LLRing now.

If I was also hitting well below average primes on LLRnet too, then I'd be wondering but I continue to hit well above average there. Today was an excellent example...hitting one on the next-to-last test on server 300 before changing to server 5000.

Very strange to say the least. I guess it all evens out.


Gary

[mdettweiler]

Originally Posted by gd_barnes

I'm sure they're perfect. Also, it's 4 different machines running the 7 ranges. None have ever had a false prime, never one of those wierd false '00000...3' residues, no overclocking, relatively cool room all the time, maintenance folks keep the room well dusted, etc. My work laptop with the bad memory chip has done the false prime and wierd residue thing and of course I don't use it for LLRing now.

If I was also hitting well below average primes on LLRnet too, then I'd be wondering but I continue to hit well above average there. Today was an excellent example...hitting one on the next-to-last test on server 300 before changing to server 5000.

Very strange to say the least. I guess it all evens out.


Gary

Okay, I see. Yeah, I guess that rules out bad residuals as a problem. As you said, it just must be the way it all evens out then.

[gd_barnes]

Files have been posted for n=400K-420K.

These files will be good size at this n-range. 1 of these n=2K files at n=~400K should take about as long to test as 3 files for the n=200 ranges at n=280K in drive 2.

Carlos, I'm guessing that 5-6 of them will last all cores of your overclocked quad > 1 week.

Karlos and Anon, as big as these are at this n-range, I will drop the file sizes to n=1K pieces at n=450K. 1K was just too small starting out and I didn't want to mess with n=1500 sized files, which would make it more difficult than it already is in scoring.

Anyone can feel free to request an n=1K range of the bigger files now.


Gary

[em99010pepe]

Originally Posted by gd_barnes

Carlos, I'm guessing that 5-6 of them will last all cores of your overclocked quad > 1 week.

Gary, I moved 5 cores to clean 260k-280k in 3-4 days I suppose but then I need to finish my other range from team drive 2. My progress is something like this: 276-280 50% done; 260-276 25% done. After I'll stay here to help and I hope with a new quad-core.

Carlos

[Mini-Geek]

Known primes in unreserved ranges:
395*2^401228-1
307*2^408953-1
373*2^418927-1

Known primes in reserved but uncompleted ranges:
357*2^263518-1
397*2^268601-1
377*2^277810-1
365*2^278054-1
321*2^278837-1
355*2^279801-1
395*2^370458-1
395*2^372674-1
377*2^376274-1
355*2^378477-1
365*2^379426-1
395*2^379448-1
363*2^383088-1
331*2^389727-1
381*2^398511-1

How were these known primes found previously? Fixed k or fixed n searches, that just happen to slightly overlap with NPLB's searches?