PPD difference, cache ??

[Markfw]

I have many boxes. Most of my Q6600's with 2 smp clients are doing ~1600-1700 (each client). I just got an E2200 to use to support a new 8800GT (it will only be doing the GPU client alone once the GPU is running). Anyway, to stability test before the card comes, I now have it down to 2800, and I think its stable or close to it. If performance was linear, it should do ~1300 ppd, but it only foes 750 !!!!! My Q9450 does 1850 ppd. I think the small cache on the E2200 is killing it. I am glad its only serving the gpu client long term, but $80 for reail cpu and motherboard was hard to resist. Add a 8800GT@$115, and for a little over $200, you have another ~5000 ppd ! (we will see if it kills it supporting a GPU)

E2200= 1 meg
Q6600= 4 meg (each client)
E6400= 2 meg Not running smp client to tell ppd anyway, just gpu.

Any other feedback or observations are welcome.

 

[Cutthroat]

Cache makes a difference on some WU's and not on others. I've noticed in particular that p2665 is heavily dependent on cache.

 

[Insidious]

I don't have two PCs that will lend themselves to apples-apples comparison. Too many other things are different.

But I do seem to remember when X2s of differing cache sizes and Opterons were in vogue everyone was comparing them... the general consensus was that cache did make a considerable difference in the ability to fold quickly.

-Sid

 

[Markfw]

I just got the E2200 exchanged for an E7200. Its running at 2933 instead of the E2200 @ 2800, but the ppd went to 1600 ! Over doubled, so cache really does matter for F@H, at least the 2665 as noted.