Yep, they are indeed BH5s. I don't know why Buffalo changed their part number (prompting newegg to), the modules are the same as the older /WB part number, at least for now. They bear the same sticker, which has the /WB on the end.
I've got tests in every class this week so I haven't had much time to flog them, but they are only a tiny bit better than the Buffalo 3200 CH5 stuff I was running. I can run the GAT Street Racer and F1 settings now, where SR ran slower than turbo on the CH5 and F1 would not allow windows to load. Neither set really likes the "command per clock cycle" option, and with it turned off F1 proved the best GAT setting.
So currently I'm running GAT on F1 and no CPC. The CH5 ran virtually as fast running GAT on Turbo with CPC on, but I don't think it was 100% stable like that. The BH5's have passed a 12 hour prime95 torture test, and while I didn't run the CH5s that long I don't think they would have made it. Turn the CPC off to insure stability with the CH5 and the BH5's spank them pretty well via the increased GAT level.
Sorry I can't tell you how much clock rate they will take, my cpu is topped at at 190fsb. My goal is to wring as much performance as possible from 380MHz, via the tightest possible timings and a big fat load of GAT. As you can see from the PCMark score in the sig, it is a valid approach. What I wouln't give for a c series chip with an 18 multiplier so I could scale the results up and show the validity of the 1:1-moderate clock rate-tight tight tight timings approach. The realized latency (stemming from both the bios timings and the increased efficiency of PAT-like technologies realized in 1:1 mode) makes it more than a match for the latency-challenged async modes.