Help needed on ports IB4000 & IB5000 vs. IB8000
 

With almost all of the new Free-DCers running port IB8000 on the 10th drive now, I'd like to ask that some of our Free-DC regulars and other folks that are on that port move some or all of those cores to ports IB4000 and 5000 for the 5th and 6th drives in the near future.

Several reasons:

1. The lower k-values will test faster for tests of the same size for some n-ranges. You'll technically get slightly more prime score per CPU hour.

2. There are slightly less already known primes on the 5th and 6th drives. Once again, you'll get slightly more primes/prime score per CPU hour.

3. We're still sieving k=1400-2000 to P=45T. If we wait and test some of those pairs after we complete sieving around May 15th, we'll save some CPU time.

4. We'd like the lower k's (i.e. k=300-1001) to maintain a higher search depth than the higher k's (i.e. k=1400-2000). This keeps it more in line with the project objectives.

Also, as we close in on the completion to n=600K for k=1400-2000, we're passing the search depth of Peter Benson's k=1005-1400 range. It appears that it will still be 1-2 months before he has completed that range to n=600K.

We need about an equal # of cores on ports IB4000 and IB5000. About 6-15 additional quads on each moved over from port IB8000 would help things be more "in balance".

Heck of a job on port IB8000 guys! This is a great problem to have. :smile:

Edit: I'll stick this thread for about a week and then unstick it.


Thanks,
Gary

I have a cache of 50 WU on port 8000. Wasn't there a quick way of flushing them out?

[quote=Flatlander;171931]I have a cache of 50 WU on port 8000. Wasn't there a quick way of flushing them out?[/quote]
batch file:

llrnet -c
[repeat 50 times]

Or program some sort of for loop (or similar loop, I'm not too familiar with programming in batch files) into the batch file that runs llrnet -c 50 times.

Everyone needn't feel like they need to move their machines immediately. Feel free to finish what you have cached and/or move your machines over gradually at your convienience over the next few days.

My intent is that the "trend" be to put a higher percentage of machines on the lower k-values since so much has been focused on the higher k-values for quite a while.

It would be nice to have all k=300-1001 at n=700K by mid-June or so. Ian and I are working the remaining k's that are n<700K for k=300-400.


Gary