Re: A8N32 overclock settings help
The following is by Cisco from Nforceshq. It is taken from a thread about the Asus A8N32 SLI Deluxe board. Thanks for the help Cisco
1. HTT, LDT...all Hypertransport things.
In the A8N32 there are 2 hypertransports to think about--not just the one as is the case with normal SLI MBs - this is unique to this MB and other MBs based on the SLI-16 chipset. The main, conventional, HTT is calculated with the FSB X K8>NB multiplier X2. The defaullt is FSB=200 and K8>NB=5X. The second one is SB>NB Frequency (you get at this one by turning "auto" off in the SB>NB Overclock item in the Jumperfree page) X SB>NB Multiplier X2. The default is the same as the other 200 & 5X.
Conventional wisdom has it that when overclocking by increasing the FSB, the multiplier should be lowered to maintain the HTT at, or below, 1000 (X2). So for FSBs in the range of 201-250 the multiplier should be set to 4X and for 251-333, at 3X. Many people have found however that this MB can exceed 1000 by up to 20% with no problems (see the HardOCP review for example.)
All of that applies to the normal HTT. When it comes to the secondary SB>NB HTT we're on our own. Some claim that it is best left at 200 X5 and some others believe it should match whatever the primary is set to: i.e. match the SB>NB frequency to the FSB and keep the multipliers the same (I do it the second way and it works for me.)
I haven't seen the "linkwidth" setting for K8>NB and SB>NB discussed very much, other than the odd and unreliable claim that if you set either one to 8X from 16X "problems" arise...I just leave them at 16X and it works for me.
2. Memory timings
Memory clock frequency is derived by the FSB and Memory divider/multiplier settings. One obscure little problem with this is that the BIOS (up to 1009 anyways) does not report things properly. This is further confused by the memory divider being expressed as frequencies--not dividers.
For example: If the FSB is set at 250 and the memory divider is set at the default 200 the resulting frequency is 250--that is because the "200" really means 250 X 200/200 or 1:1. A setting of 166 there with the same 250 FSB would give you 250 X 166/200 or 207.5--the BIOS boot screen will report it at 166 but any utility (like CPUZ, CBID, etc. will report the correct frequency.)
One thing worth noting is that this MB also provides memory multipliers (settings above 200.) If you're overclocking the FSB, you might as well just forget about them. They would be of use only if you want to keep the FSB at 200 but want to run the memory faster than the FSB--it can be done with the right, premium (DDR450, DDR500, etc.) memory modules but the FSB on this MB can easily do anything this memory can do so it sort of begs the question, "why would you?" If my memory can do 250, you can bet that my FSB would immediately go there also even if I have to reduce the CPU multiplier to do so (this would be in the case of a dud CPU since just about every single AMD64 will gladly overclock by 25% with minor tweaks to the vCore voltage if required.)
3. Odds and sods
-- Peg Link: This is the ASUS BIOS utility for overclocking the video card by up to approximately 12%...It's probably just me but I am offended by BIOS settings labeled "Normal, Fast, Faster and Fastest" Use if you wish but "disabled" is my recomended setting. There are much better ways of overclocking the video card through "coolbits" or other programs that give you actual numbers to work with.
-- Auto overclocking options: A Ferrari with an automatic transmision? 'nough said.
-- Cool & quiet: Think of this as the "Underclocking" setting. To each their own but mine is always disabled...perhaps my CPU will only last 3 years instead of "one lifetime"
If you find other incomprehensible settings just ask away...unless it's one of the incredible number of memory timing options we have with this MB--beyond locking my memory at 1T command rate, I freely admit to being mostly clueless with settings such as "DDR Clock skew
Here's a link to the same board and some basic help on overclocking this board. Again I'd like to thank Cisco for taking the time to put this down. I've pm'd him to let him know we're using his info here.