View Single Post
Old 2009-05-01, 22:50   #18
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

1017510 Posts
Default

David,

Later tonight, I'll take the factors from the sieve drive for P=20T-25T and apply them to that file. I'll then send you a file for n=600K-650K to load into port 8000. We're blowing through this drive at break-neck speed.

PCZ and Vaughan,

Could I talk you guys into moving most of your port 8000 machines over to Ironbit's ports 4000 (5th drive) and 5000 (6th drive) within the next 1-2 weeks? Within that time, we really need anywhere from about 5-10 quads running those instead of port 8000. It'd be good if about an equal # were on each port. I have 6 quads running port 5000 on the 6th drive but it is well behind the 5th and 7th drives. Ian has what appears to be about 4-5 quads running the 7th drive. With it ahead of the other 2 drives, we don't need any more on port GB4000 at this time. Only ports IB4000 and 5000 are lacking.

Several reasons:

1. Those drives will test a little faster at the same size for some n-ranges. You'll find slightly more primes per CPU hour at the same size.

2. There are slightly less already known primes on them.

3. The project intent is to keep the lower k's at deeper searched depths than higher k's for the above 2 reasons. I'd like to avoid a situation where k=1400-2000 is at n>650K and k=600-800 is at n<650K. Also, Benson's k=1005-1400 will not be at n=600K by the time we hit n=600K for k=1400-2000.

I don't want to inconvience the non-regular Free-DC searchers so I'm choosing to inconvience the regulars instead. :-)

Initially, the tests will take a little longer than port 8000 because the drives are at higher n-ranges but they won't be for much longer with so many cores on 8000. When all are testing the same size, the lower k-ranges should test faster for some n-ranges.

The next couple of rallies will be run on a combination of ports 4000/5000/8000 assuming that they are at close to the same search depth.

What an excellent situation to have to deal with! :-)


Gary

Last fiddled with by gd_barnes on 2009-05-01 at 22:52
gd_barnes is online now   Reply With Quote