I have put the Stepping 7 Intel Core 2 Extreme Q6700 on the ASUS P5WDH Deluxe
for some overclocking. I had her up to around 4.75GHz on this motherboard.
Then I went on to benchmark her at 4.6GHz in 3D Mark 06:
Of course, since the Processor has not been officially released, results can’t
be published as yet, and it is recognized as unknown CPU. However, it has it’s
quirks on this mainboard at this time still, with BIOS 1407. You see, when you
raise multiplier above the default 10x on this Processor, 2 cores rise in multiplier
but 2 other cores remain fixed at 10x:
2 Cores @ 16x, 2 Cores @ 10x
Apprarently, multiplier below 10x work flawlessly however.
4 Cores @ 8x
I’ve seen the ASUS P5B Deluxe to be free from this ailment, hopefully this
can be easily fixed with a BIOS Flash.
Next, I tested out the maximum FSB it can go stable with the QX6700. It passes
SuperPi 1M at 350MHz FSB on this motherboard, but not much
more. This is similar to what I’ve encountered on the Intel D975XBX1 Bad Axe
1. It seems that at this moment, the Intel
D975XBX2 Bad Axe 2 is the only 975 chipset motherboard that goes above 400MHz
FSB easily with Quad Core Kentsfield CPUs. So why is this so? A straightforward
guess is that the chipset is of a better batch. However, this reasoning is hard
to swallow when you see such a big jump in FSB tolerance. So I decided to take
a closer look comparing the Bad Axe 1 against the Bad Axe 2.
Above shows the Bad Axe 1 beside the Bad Axe 2, at the area between CPU socket
and the NorthBridge. You see a lot more traces running from the Bad Axe 2 to
the Northbridge than on the Bad Axe 1, especially from the LGA Land rows of
A1 to 10. Initially, I had thought that this was an enhancement (signal perhaps)
move to optimize the board for running Kentsfield. After asking around, it seems
that it is in fact a change in PCB layer sequence. The 4 data group running
from the Processor to the Chipset is arranged across the 6 layer PCB of the
motherboards differently on the BX2 as compared to the BX1. This is the main
enhancement made for Kentsfield. We have learnt that most of the motherboards
based on the i975x chipset is based on the data group trace layer sequence similar
to the BX1, while the new ones coming out should be based on the BX2 sequence.
We have heard faintly about one – based on the i975x chipset – soon to debut
from Gigabyte, featuring this optimization for Quad Core Processors. I don’t
pretend to understand why this arrangement would enhance operation of the Kentsfield
Processor, but I have seen the difference myself when it comes to FSB overclocking.