Hello,

I was playing with overclocking one of my Quadra 650/800s, and I noticed
something a little unexpected.

I switched the 33 MHz 68040 (which was running at 40 MHz) for a 40 MHz,
and set that to run at 48 MHz. Afterwards, I got bus errors, so I took out
some of the slower (70 ns) memory.

It ran fine, and I got these speeds from the distributed.net client:
RC5:   36,855 keys/sec
OGR:  161,048 nodes/sec
Not too bad for an '040.

But to finish the computer, I took out the one 16 meg SIMM and put in a
pair of 8 meg SIMMs, knowing that these Quadras take advantage of memory
interleaving, and got these results:
RC5:   40,329
OGR:  171,513

Damn! That's a big difference! Has anyone else noticed big differences
like this with interleaved memory? It is especially surprising because the
distributed.net clients are heavily optimised and should fit almost
entirely into the CPU's cache.

Comments? Thoughts?

Thanks,
John Klos