Ok perhaps I wasn't to clear.
Let’s take DDR 400MHz as an example. The frequency at which data is sent from the DIMM is 400MHz or 400 million times per second. So the time between each bit of data being sent is 1/400,000,000 seconds or 1/400e06 seconds if you will let me express it in scientific notation. If you work this out you will find that it is equal to 2.5 nano-seconds (ns), where a nano second is 1 billionth of a second.
Now back to the point, latencies for ram are expressed in clock cycles. The clock cycles at half the data rate for DDR ram or 200 million times per second, so one complete clock cycle takes 5ns. If there is a delay of 2 clock cycles for a given operation, this would mean a delay of 10ns.
Right, now we look at DDR2 800MHz, externally the clock is cycling at 400 million times per second (or twice as fast as DDR 400) so a latency of 2 in this case means a delay of 5ns (if we use the external clock speed). If we use the internal speed of the actual chips in the ram this is a clock rate of 200 million times per second and hence a delay of 2 of these clock cycles would give you a delay of 10ns again.
To sum up DDR400 with a latency of 2 will mean a delay of 10ns on any given operation (CAS, tRAS...).
DDR2 800 with a latency of 2 will mean a delay of 5ns (if you use the external clock rate for your calculation) or 10ns (if you use the internal one).
I am asking if a latency of x clock cycles on DDR2 is referring to external or internal clock cycles. If it refers to internal clock cycles then the latencies of DDR 400 and DDR2 800 are directly comparable, i.e. latency of 2 on DDR 400 is equal (in terms of actual time delay) to a latency of 2 on DDR2 800. If however the latencies of DDR2 refer to the external clock cycles then a latency of 2 on DDR 400 would be equal to a latency of 4 on DDR2 800.
I am assuming that you know how DDR2 works here, if you would like an explanation (all be it a rough one, as I am no expert on this) ask and I shall try.
Hope that’s a bit clearer.