Originally Posted by Cletas
Is there a point to where one should start trying to tighten the timings and stop increasing the ram speed?
Or better to increase ram speed as much as possible then see if it is possible to lower to timings at that speed?
It's always a combination of both, and neither can be neglected.
More MHz = more bandwidth, lower latency.
Tighter timings more bandwidth, lower latency.
One of the most important factors in the performance of your chipset/memory, with an Intel board, is tRD (performance level). The faster your memory runs relative to the FSB, and the lower the CAS timing, the tighter you can make tRD.
Generally, at around 400MHz FSB, you are looking for a tRD of 5-6; at ~450, 6-7; and at ~500, 7-8.
FSB and tRD being equal, a faster clock rate is almost always better than tighter timings.
For example with a 400MHz FSB, you can run at tRD of 5 with DDR-800 CL3, DDR-1000 CL4, and DDR 1200 CL5. The DDR-1200 CL5 will be faster, in bandwidth and latency, than 800 CL3, or 1000 CL4 (though the difference between 1000 CL4 and 1200CL5 is less than that between 800 CL3 and 1000 CL4).
If you were to run 1200 CL6, that would force you to a tRD of 6 at the lowest. In this case, you would be better off with 1000 CL4, so you could keep that tRD of 5.
If you have to sacrifice some MHz to get a CAS low enough to run a better tRD, do it. If you can sacrfice CAS for a jump in clock speed that gives you the same, or better tRD, do that.