2m is a big chunk to be different on 32M, but could even be from cooling solutions or how much stress is on the NB I guess. It's tough to say without seeing a side-by-side comparison. I know there's one at XS somewhere, but I keep missing it
I think a lot of the issues between chipsets and SPi's/latencies is solely based on how the manufacturer tweaks just like you said. The DQ6 I have is definitely WAY messed up, for mem in particular, and apparently it's not the only 965 board that has similar issues with timings and ratios. Mine booted with timings that aren't even in the SPD programming AFAIK and refused to hold manually set timings.
965s may still get a decent bump in performance as BIOS's mature. Many of the Conroe boards out there still have trouble just running them. I've never seen such a streak of BIOS releases from manufacturers...ASUS seems to have a new one every week for the P5W DH. I think the first order of business for manufacturers is to get something that "works" 100% and then hopefully they will concentrate on improving performance, especially for the 965s, since it seems like even the heavyweights aren't immune from buggy boards/BIOSes.
Maybe Abit or MSI are walking quietly and carrying a big stick