simon12 :
Thanks for the answers does anyone know if it would cost any more to make P67 boards? I mean for the components not does Intel charge Asus & others more?
For different criteria, you need beefier components than the "mainstream". In some cases this takes the form of better voltage regulation, in others it takes the form of better signal integrity on the I/Os.
More I/Os, for instance-- like the video port-- can drive more layers in the PCB, which raises cost. If you don't need video, you can go with a cheaper PCB and put in better inductors for OCing; vice versa if you go the other way. If your motherboard has to be built to accommodate both video and overclocking, you're going to have expensive motherboards across the entire product line. So you differentiate the featureset at the chipset level so every motherboard vendor doesn't overbuild and make your platform costs skyrocket across the board.
In doing this, you try to "package" features together that you think will go with a given market. This is more of a dark art than a science, IMO-- you have to make a judgement on whether, for instance, P67 customers will need RAID, and I will be honest and state I don't know the marketing guys' methodology. I am sure they have criteria for which features are activated and deactivated for a given segment of the market, I just don't know them.
As to why... keep in mind that validation is very intensive in both money and time (and complexity), and every feature needs to be validated on every platform, for every stepping, for every BIOS release, so if you shut off-- for instance-- VTd on one platform because your research has suggested no one who operates a Home Theater PC uses virtualization, you won't have to spend the resources validating this feature on platforms where it's been de-featured. You differentiate products at the CPU and chipset level based on market and and who is willing to pay for which feature-- designing, testing, and validating each of these features costs a finite amount of resources during development, and you want to be repaid for them if a customer is actually going to use them.
For something like QuickSync-- something which makes intensive use of graphics portions of the processor which are otherwise barely active during P67 operation-- I think you would have to validate lots of completely unrelated hardware on the CPU in order to ensure QuickSync worked on that platform. Hardware which could change the thermals and turbo characteristics since it's typically inactive during discrete graphics operation. This could be done, of course, but at costs of time and cash, which would drive up prices in general and possibly miss crucial product timeframes. If instead you postpone the dual featureset (which will only be sought out by a small subset of end users anyway) several months, you increase your time available for validation, get away from the driver thrash of an early platform, and can build momentum and OEM experience with your product before introducing a wildcard like the Z68.