Kaiju said:
I've been trying to figure out the benefits of synchronising Conroe RAM and FSB (1:1) over using a divider, and have never got a good answer.
Can someone please explain why its better to run memory and FSB synced (if that is indeed the case?) Or link me to some benches perhaps.
If indeed it is better to sync', then that would at least validate buying higher frequency RAM (PC8000 over PC5300), without taking in account the chip (D9 Micron for example).
It hasn't been massively important since the AMD XP on nF2 motherboard days where running anything other than 1:1 decreased performance by a significant amount.
With regards to Conroe, I haven't seen anything too definitive yet, beyond the odd synthetic benchmarks which are not an especially useful reference for real world use.
Would imagine it won't be too long before more thorough results emerge though.
Some users have mentioned that their systems "feel faster" when using tighter timings (say 3-3-3 vs 4-4-4) to which I can only offer "hmmmmmmmmm".
From personal testing, one thing I should point out is that the chipset not only imposes an FSB limit but also a RAM clock limit, that is to say that i.e. if your board tops out at 400FSB then there is no guarantee that your RAM, even if capable, will be able to run a 2:3 divider for 600MHz clocks on it.
It may require a slower strap to do this (and not just for increasing FSB as is often discussed). The more sticks you use, the greater the likelyhood of this.
With the issue of strap changes, this is one area where the benchmarks definitely do indicate a decent enough decrease in performance to discourage moving to a slower strap if it can be avoided. (unless the strap change allows you to push the FSB so much further as to compensate for the decrease - typically it doesn't though).
Whether this loss is noticeable in everyday use remains to be seen...
