Switching back to tinkers?
shwaip
bluffin' with my muffin Icrontian
Would there be any points benefit for the team if we took the SMx machines off of gromacs, and back to tinkers, now that they're worth "more"?
0
Comments
I already took the -advmethods out of SM21. For the first time since I got her she broke 700pts/week. I left -forceSSE as i stil get about 60% gromacs even without -advmethods.
Even with -advmethods in place, my boxes will be passing Icrontic2 in pointage tomorrow sometime. The interesting thing is the turnin timeframes for Tinkers, as well as pointages. P4s are so bad at Tinkers that the repointaging actually gives a tinker advantage to slower boxes-- but, the P4s are blazing through Double Gromacs. Since there are now three WU grades out there, and points are more balanced, might not really matter over a month.
Folding is evening out points some.
- Current technology fast P4 or Athlon.
- One instance of Folding Running.
- Hyperthreading turned off with the P4.
1) Points generated from an Athlon running Tinkers versus P4 running Gromacs will be roughly equivalent.
2) P4 running a Double Gromacs will best any other machine running any other protein model.
P4 in hyperthreading mode with two instances of Folding, any combination of two work units, will outproduce any other machine running any single work unit. The only exception to this would be the P4/HT running two low-point work units versus another P4/non-HT running a Double Gromacs.
Where the points production becomes exceptionally lucrative is when a hyperthreaded machine runs two instances of folding, one a Double Gromacs and the other a Gromacs. I don't recall ever having been issued two Double Gromacs simultaneously.
Did I confuse the non-Intel folks? It just sounds nuts unless you've actually observed the points generation yourself. (Not an Intel fanboy here. My ratio of AMD to Intel builds is 8:1.)
AMDs - Get as many tinkers as possible.