A P4 2.8 will make about 750 to maybe as much as 900 points a week if it was running 24/7.
Ok, I did a search and this was the closest information I could find to answer my question. My question today is.. Does the Celeron absolutely blow at floating point math? I had the interesting observation that right now, all 3 of my folding machines have strains of the exact same protien. Yes, I realize that doesn't mean it's still a perfect benchmark, but I've been watching all 3 systems for the past couple days and they seem to get very similar PPD for all gromacs anyway, and right now they all have p563_BBA5_ext. Here's what I have:
Pentium III 1Ghz, 133Mhz memory
Athlon 2400+ running at 995Mhz, memory at 266Mhz due to mobo incompatibility with the mobile chips
Celeron 2.6 o/c'd to 2730Mhz, 420Mhz memory
So basically two either stock, or underclocked, and one 4% overclocked. I already confirmed the Celeron is stable by running mprime (prime95 for linux) for 8 hours in torture test mode. No errors, which is what you'd expect for such a modest o/c. The memory is DDR400 by the way. Anyway, here's my current PPD listing for each server
Now I'm not too surprised that the Athlon has about a 45% performance advantage over my old P3 considering how much more advanced the Barton core is, plus the memory is running twice as fast. I already determined through earlier testing that memory speed seems to play a pretty significant role in speed for both mprime and folding.
Now in theory, I was thinking the Celeron should still have about a 2:1 advantage over the Athlon, but in reality it posts up scores that are only 33% faster. This is sad! Unfortunately, both servers are linux so I don't have any other type of benchmarking software I can load up to compare the two. Still, 56ppd is just shy of 400ppw, which is roughly half what it sounds like a non-Celeron P4 should be capable of at the same clockspeed. So what's the deal? Should I bite the bullet and just get a normal P4? Is the Celeron really that bad at math?