My question is this: With the bus speed now hitting 1600MHz, have we come back to a position where running memory in sync with much better timings is preferable? We can now run the RAM @ 800MHz in sync but looking at the available DDR-3 modules, some of them are twice the bus speed! - unfortunately the timings have gone up accordingly. I did notice on the Skulltrail review that the memory was 1:2 ratio.
I am assuming that raw clock speed of such a greater margin is going to easily outweigh the lower latency figures, but it would be nice to have a test on THG with various CPUs / memory modules compared.
I have been building machines since the K6-2 days and synchronous operation has always been the way to go but lately it seems that Ram speed has far outstripped the bus speed increases that we get.
Maybe when Intel & AMD are both using onboard memory controllers (CSI / QPI & Hypertransport respectively) the situation will be clearer. Certainly I am sure that the new Common System Interface speed from Intel will be far greater than the 1600MHz that we have now which means that sync might be the way to go again. Hopefully the link will work properly - employees have already quit Intel over the dispute over which way to go with the memory controllers.
I am assuming that raw clock speed of such a greater margin is going to easily outweigh the lower latency figures, but it would be nice to have a test on THG with various CPUs / memory modules compared.
I have been building machines since the K6-2 days and synchronous operation has always been the way to go but lately it seems that Ram speed has far outstripped the bus speed increases that we get.
Maybe when Intel & AMD are both using onboard memory controllers (CSI / QPI & Hypertransport respectively) the situation will be clearer. Certainly I am sure that the new Common System Interface speed from Intel will be far greater than the 1600MHz that we have now which means that sync might be the way to go again. Hopefully the link will work properly - employees have already quit Intel over the dispute over which way to go with the memory controllers.