> I could be wrong, but I don't think that with that memory you'd want to > clock the system bus past 166 (since it's "DDR" -- "double data rate" > RAM, the actual FSB clock is half of what it says it is)... Actually, this is where most overclockers push on things. They increase the basal RAM clocks, and you can usually tweak it about 1-3% without too much drama. The better your RAM (and to some extent), the more you can tweak. CPUs seem to take it in stride so long as the thermal figures are OK. There is a fun multiplicative effect here too... one additional RAM Mhz translates to 20-30MHz of CPU Mhz. > that after a certain point AMD started locking the multiplier capability > in their athlons since people were overclocking them so much They always locked the multiplier, except in their "highest speed" CPU within a class. For example, the 1.4GHz Thunderbird is completely unlocked. But who cares, because even at specification speeds, it produces enough heat to setoff a localised fusion reaction in your PC's case. AMD had various measures to protect the multipliers, some were trivial to work-around, and others more tricky. With the Pentium 4, Intel pretty much locked that down for good though, but with all of the hyper-fast RAM available (ever wonder what DDR-533 was for?), you can push on the FSB clocks quite easily without too much difficulty. > If you got some faster memory, though, you could probably overclock the > bus. Am I right about this? Correct. I've seen more and more BIOSs which allow you to configure the CPU pre-multiplied-FSB clock. As always, YMMV. I was able to bump my SuperMicro FSB800 P4 servers from 3.2GHz to 3.4GHz, which passed by hardcore 7-day 100% CPU burn-in test (which does md5sums of /dev/random filled memory) without any trouble. Since the box lives in CPU Paradise [a datacentre, with 65F, 55% RH, and lots of forced-air circulation, and not a trace of dust], it's quite happy. =MB= -- A focus on Quality.
