Gromacs
Check out the threads going about Gromacs guys. I'm starting to test some of my rigs out using the -advmethods flag, which forces the use of a server that only gives out Gromacs WU's. I've started a thread to compare Points/Week of the gromacs WU's with the most efficient current Tinker WU's. The following CPU's are likely to benefit using Gromacs although you may have to turn your overclock down just a tiny bit:
Athlon XP
Duron 1 GHz & above (morgan core)
Pentium 4
Pentium 3
Celeron 566 mhz & above (coppermine core)
Note, that this is just comparing P903 and P180/182. There are 166 villins, P183's, and who knows what other little Tinker WU's floating around right now, and they are NOT as efficient as P180/182's. So by forcing Gromacs, you can avoid these less efficient proteins on some of your rigs. Please do some of your own testing if you want. I found a P3 933 got about 17% more points in a week folding P903's as opposed to P180/182's. Gnerma determined that his P4 got an even bigger boost because of the RDram memory bandwidth, about 63% in weekly points!
Ask me anything, and I'll help you set it up. Using core78 really is faster on most CPU's with SSE, definitely on Intels. It's incredibly easy to switch over to core78 and it'll boost your production by hundreds or even thousands of points over the course of a year, per CPU!!!
There also seems to be exceptions with the Linux client. overdoze tells me that Tbirds ARE faster with gromacs using the Linux client than they are with the Tinker core, even though they don't have SSE instructions.
Edit: Even more to know...
Actually, if you're 100% sure the rigs are stable, you may also want to use the flag - forceasm. So it would be "path to the console" -advmethods -forceasm
with a space before each hyphen... The console will disable SSE/3dNow! optimizations following an abnormal shutdown. Since these are what makes the Gromacs core fast, you'll want to have the console force them to come on any time its restarted in case they have been disabled. Now its important to understand that Stanford set the console up to disable these optimizations if it thought there was a problem with stability, since Gromacs is new and still being tested. However, it also turns them off following an improper shutdown, power failure, etc. Once the WU finishes it turns them back on on its own. If you're absolutely sure its stable, use this extra flag to avoid the problem.
I just noticed this on the P3 933 I tested today. I task schedule the client so my friend does't have to be bothered by it (out of sight, out of mind). So I shut it down by killing it in the task manager in order to play a game earlier (didn't want to risk corrupting the Gromacs WU while playing UT2k3). This caused SSE to be disabled when I turned it on again later. The frame time went from 18 minutes/frame with SSE to 40 minutes/frame without SSE!!
So now I added the -forceasm flag to my scheduled task. I know this rig is stable because its not overclocked and rarely ever has an error. It was just the way I'd shutdown. So Gromacs is still touchy and bears watching if you're overclocked fairly heavily and don't use the flag for the sake of Stanford's data's integrity. But you may want the extra flag for insurance against improper shutdowns if you know its stable.