I'd just like to make sure I understand this by giving a personal example. By trial and error, I stumbled upon a so-far stable setup shown in my sig.
So, if I am clear, the reason for its stability is that 13.5/11=1.227, 1.227x333x4=1634.72 - this is very shortly after the 1600 strap so that although NB latency may be high it is more stable due to being just over the nearest strap. It also explains the drop in some benchmarks.
[I'm a little fuzzy on this bit] If that's true, given the memory is rated for 200MHz and is running at 333, I'm wondering if at this point I'm best off increasing the FSB incrementally, say to see how far I can get to the next strap and remain stable?[EDIT
It suddenly occurred to me I wasn't asking the right question. In my BIOS, my only choices for RAM settings are to set the bus at 2.00, 2.66, 3.33, 4.00, and auto. This leads me to believe that in the case of the RAM, I just need to test the impact of running ram asynchronously. Thoughts on this are especially appreciated.
Here are my calculations with figures chosen to go just slightly over the strap in order to see what incremental increases in FSB can be applied. I apologize if I have this all wrong.
Current, repeated for easy reference 11x333=3.66GHz, NBCC 13.5/11x333x4=1634 (I'm assuming a strap at 1600)
If I provide more room with this,
12.5x310=3.875GHz, NBCC 1.08x310x4=1339.2 (assuming strap 1333)
do you think I'd be able to drive up the FSB incrementally by, say 10 at a time? That is, would this lower starting point allow me to push it further?
The $64,000 question, of course: would 11x363=~4Ghz (NBCC = 1782) be better or worse than 12.5x320=4GHz (NB = 1382)? My gut reaction is that 11x363 is the obvious choice but I now am concerned about stability and the faster memory timings at 12.5x320 must allow for even higher bus speed increases.
I believe I will undertake some testing of my own but I would appreciate your thoughts on the above to let me know if I am on the right track and whether I can narrow my choices. I will report back what I learn but it may take me some time due to the large number of data sets to acquire.
Incidentally, just in case it helps, I suggest an easy way to quickly track the change in NB FSB as the CPU FSB changes.
1. Determine ratio,r e.g. 13.5/11 = 1.227=r
2. Multiply by 4, 4 x r=4.91*=C
3. CPU FSB increase x C = increase in NB FSB
So it's one simple multiplication each time you want to raise the FSB and check the impact to NB FSB.
*You only ever have to do this once unless you change the multiplier. I'm sure there's a kind soul with a talent for scripting who could come up with a little tool for this.
Along the way it partially answers my question above in that starting at 266 with a higher multiplier offers much higher granularity, perhaps more than is supported by my mobo. I'll check.
In closing, I hope this helps but I also hope for your advice as to what is a more or less good limit above or below the strap?
Thanks. Sorry for the long post, particularly from a noob.Edited by Rebound - 7/17/10 at 10:47pm