Memory Latency hit if run Asnychronously?

If I run my memory at the same speed as CPU fsb, the two are in synch, no hit, but if I run my memory faster then my CPU fsb, is there a latency hit? and how much faster then the CPU fsb would I need to run the memory before I see benifits which overcome this latency hit?
I am looking for percentages, as then other people can use this posts findings to see if they are falling into the paradox (i.e. running you memory faster, but actually getting less out of it).

Anyone know actual figures or even real life "I tried it once.." figures they have noticed? All observations are welcome.

Ah, I think you got it a bit wrong. The divider lets the ram run SLOWER than the fsb.
I have an Abit ST6. 4-4-1 runs ram at full fsb speed and 4-3-1 would run the ram at 3/4 of the fsb. So if you run async then 133fsb would run memory at 100, 166fsb would run memory at 125, etc.
Running CAS 2-2-2 5/7 at the async speed is much slower than running CAS 3-3-3 7/9 at synch speed. I have run Sandra and 3DMark to verify this. I don't see the point of running async unless you have very poor quality ram limiting your oc.

rule of thumb! if youre memory is running faster than youre fsb you are wasting bandwith. the cpu only gets the bandwith that the front side bus will allow. it doesn't matter how fast the memory runs the cpu will only take in what the front side bus is feading it. it may help data transfer from componants. but % is minamal to system performance.

In theory there is a performance hit when you run memory asynchronously, albeit not a very big one. I wanted to prove this with a demostration, but the results are somewhat counter-intuitive.

Using a multiplier unlocked P4 on an 845PE based board I ran my memory at DDR400 with a 150fsb (async) and a 200fsb (sync). With the async bus my sandra score was 3043. With the 'sync' bus, that score dropped to 3004.

I guess the short of it is that running memory asynchonously won't cause much of a performance hit (if any). It really depends on the peculiarities of the chipset. If it is designed to be run async, it will probably be faster async than sync. On an Athlon mobo there may not be any benefit. But if you're talking DDR P4 systems, async memory is key.

I have my pc2700 samsung and CPU both running at 166 using CAS 2-2-2 5 is this the fastest wayt o run this, or would reducing the FSB/timings give me better results ??

Basically what I am asking is if you have a dual channel DDR system, you get twice the bandwidth from your ram. SO If you run your CPU FSB at twice that of your ram, your CPU bandwidth would match your ram bandwidth. Is this correct?

Would I be able to run a 200FSB Athlon and have dual channel DRR PC2100 at stock(266) and not see a performance hit from running the RAM asynch. as the ram bandwidth still exceeds the CPU bandwidth at those speeds.

Anyone????