Interesting, because on my Celly II at 897 (Win98), the 464c is in the range 6% slower than 463. Whereas I was originally getting over 6.2 Mnodes/sec using the OGR long benchmark with a couple of programs running in the background, it dropped about 0.4 Mnodes/sec in 464c. I have not yet tried it on Pentium IIs.
By the way, if you're going to test it, back up your other data first because it seems to start the stub over again. I had just put everything in a backup directory and when I replaced with the old client and old files again, it started the stub again where I left off.