The "40" I believe refers to the latency of the memory modules. With RDRAM, the faster the bus, the lower the access latencies (which was one of the big complaints people had about RDRAM when it first came out, compared to SDRAM). PC1066 uses 32ns chips (though there appears to be PC800-45 and PC1066-35, but Intel chipsets don't support those).
As was said, there is a noticeable difference in performance by using faster memory, due both to lower latency and the considerable increase in memory bandwidth. (PC800=3.2GBps, PC1066=4.2GBps).