Do you have a question? Post it now! No Registration Necessary. Now with pictures!
- Posted on
- New Intel Board for Overclocking
- Ed Light
November 30, 2006, 3:25 am
rate this thread
Re: New Intel Board for Overclocking
Ed Light wrote:
I think the X6800 (top of the line), the multiplier goes up or
down. On the other members of the family, I think you may be
able to adjust the multiplier down (for those who like to
optimize for memory bandwidth). But if you research the
strap options for Intel chipset Northbridges, you may want to
leave the multiplier as is, and just raise the FSB instead.
I'm still waiting to hear whether alternate chipsets, like
Nvidia, have that same strap design or not. In other words,
the multiplier setting for the processor, has an equivalent
effect on the Northbridge, and can cause the Northbridge
to be overclocked. It is one of the "invisible" features of
overclocking Core2 Duo (a feature you cannot see in the BIOS),
that the cutting edge people have been spending so much time
exploring (you set the clock low in the BIOS, and do your
real overclocking in Windows).
When overclocking Core2 Duo, you keep a copy of SuperPI or
equiv handy, because you need to evaluate the results of each
overclocking experiment, to see if it is giving more performance
or not. When the Northbridge is operated at high settings via
the BIOS, a high latency choice apparently goes hand in hand with
that choice. Which is why you want a low clock setting in the BIOS,
to get a good strap selected by the BIOS, and then overclock
in Windows, to the final desired clock. Then test for performance
with a benchmark, making careful notes of whether you are getting
ahead or not. Simply blind overclocking may miss the performance
peak by 10%, if you don't do it right. I'm hoping to read,
that the non-Intel chipsets (Nvidia etc) make this a bit easier to do.
If you want to overclock Core2 Duo, I recommend going to the
overclocking private forums, and read up on their research.
Since the things they are investigating are not documented
in the Intel datasheets, it is hard for me to repeat what
I've read, without corrupting the message. It is a tribute
to their thoroughness, that they were able to detect that