Timing question

[svyerkgeniiy]

I have searched quite a few places and have not been able to find this answer anywhere.

I understand that, for DDR and DDR2 (and presumably for any DDRx in the future), there are two different clock rates: the memory clock rate, and the effective memory rate. For DDR, these might be 200 MHz and 400 MHz, while for DDR2 these might be 200 MHz and 800 MHz.

What confuses me is why latency timings on DDR2 memory are higher than for DDR. I would think that 4-4-4-12 represents memory clocks, meaning that the latency equates to 24 x 5ns or 120ns, but why should DDR2 latency be higher? Maybe it represents effective memory ticks, 24 x 1.25ns or 30ns, but this seems low.

So:
1) which is it, memory clocks or effective memory rate ticks?
2) if it's still memory clocks, why is it higher for the same memory clock rate?

--dv

 

[the_guru]

I don't know if I got your questions right but I'll give it a try...

1. Latency timings are delays in clock ticks.

2. It's higher on different memory sticks since they are more or less capable of transfering data quickly. The circuts on DDR2 memory have higher latecy since the chips can't respond that quickly so in order to not get corrupted data, the memory controller has to wait longer between actions.

 

[svyerkgeniiy]

1. I understand it's clock ticks, but I was wondering if it was the "memory bus side" ticks (e.g., 166 MHz or 200 MHz), or if it was the "CPU side" ticks (e.g., 666 MHz or 800 MHz).

2. If it's "memory bus side" ticks, then why would a "more advanced" technology require a higher latency? Maybe it's an architectural change that causes this. Presumably this extra latency has less impact as memory frequency goes beyond 200 MHz, but DDR2 has been around for a loooong time to not have gone significantly beyond that. What's the maximum now, 266 MHz? Even that has higher latencies of 5-5-5-15 from what I can tell, which is about 30 x 3.75ns = 112.5ns, far longer than the fastest DDR400 latencies of 55ns.

--dv