My P4's don't seem to like the Amber core WU's. (1801, 1804) Using .5-1 GB RAM , so that can't be issue. Just seem to run quite a bit slower (and for way less points ) than a Gromacs WU. Can I -config anything to make it less likely to receive these ?
Nope, the client derives the what WU's to go after based on the architecture of the processsor you're folding with. Really the only way to exert any sort of control over what you get is to run -advmethods, big WU's, or timeless tinkers. Otherwise its entirely at Stanford's discretion as to what you'll get.
I like the Ambers. I get better points on my AMDs and P4s than I do on normal Gromacs but I haven't seen any of them in a while. Then again, I've mostly been getting bigpackets Gromacs and those are the best pointswise for my machines. I do love the 364 pointers though. I get really good PPD from them.
As Stanford prepares more different types of work units to throw into the mix they do it in batches. One way or another they all have to get done so we take what we get.
Even with big units turned on, if the supply is low you could get a tinker or whatever, but it is mostly based on what info your machines sends back to Stanford.
They are all part of the cause, so holdem and foldem its all good in the end