Hyperthreading and ristras of hot chillis.
HT has a price production will peak about every 7-10 days with gromacs cores.
Tinker cores will extend the production peak to over two weeks between spikes.
With two exceptions had a wave of gromacs.
Also the proc count is not correct early on when experimenting with the clients Stanford seems to think
I have two more than have actually reported to them. I think it is because of the switch -local since normally
the client reads the machine ID from the registry.
THe switch -forceasm seems to have caused me some real headaches on one PIII dualie not sure why two other identical machines worked fine.
If Stanford sends a wave of tinkers expect to see production drop dramaticly for at least 7-10 days
Two new ones yesterday were tinkers
When setting up clients as a service, or any other way, in win2K or XP you can turn off/disable a number of other services that the typical
desktop may not need or use. Be careful however on a network that employs central/remote administration.
The real question is will doubling up with HT result in an overall production gain. I will need a much longer baseline to answer that question.
Fear not oh mighty Zeus the production cycle should trough for at least several days to come.
The stealth and guile of age frequently defeats youth and enthusiasm.
Except for emergencies I have been on leave for two weeks and had time to tinker.
Tuesday back to the salt mines