Which does not require Sandy Bridge...
It does though.
Graphic cards already in existance cannot pass a graphics signal to the TB chip. It either means the board has to be redesigned (integrated GPU on the main board), or graphics card vendors have to make some modifications at the very least (cheapest and easiest way to go, would be use a flexible PBC <just traces> with a DP chip to get it to the TB card). I didn't mention this however, as these do not yet exist (board would need a notch for a flexible PCB, such as what's used for Crossfire or SLI), and there's things to be ironed out between vendors (standard as to layout, dimensions, ...), as well an answer to basic question, "What's in it for GPU card vendors?" (they could be hesitant over unknown adoption).
If it takes off in the mobile market, they'd be much more willing to do so, as not having this could mean a loss in card sales. But if they choose not to take this route, the previously mentioned bridge chip would be needed or add the TB chip to the card. Ultimately, the method taken will be based on lowest cost (flexible PCB connector, bridge chip, or add a TB chip to the graphics card).
Regardless of the implementation however, current Graphics cards wouldn't be capable of providing a graphics signal to the current TB chip. Hence any PCIe card based on the existing TB chip, would only be capable of transferring PCIe data (not all that usefull in systems that have PCIe slots, as TB is only 4x lanes).
Faster storage is possible by using a PCIe slot (4x or larger), as it can exceed what's possible with TB running at 10Gb/s.
Then why ignore it?
Now if you meant the current chip is fine for DATA ONLY transfers (no MDP signals), then it would have been prudent to qualify your statement with that small bit of information.
It's also possible for Apple to add integrated graphics to the existing Mac Pro to provide output on the internal TB port. However, whether having video output is an actual requirement is still something that was up in the air, and I'm not sure we've gotten an answer on that yet.
Integrated graphics would defeat the purpose of a workstation (but integrated graphics is fine for mobile and low-end desktops).
GPU data isn't a requirement of making PCIe work over TB (electrical or optical once the cables are available), but Intel's intended TB to have both, which existing systems and graphics cards cannot provide (no way to connect a DP output signal to the TB chip's DP input signal). This is why it's aimed at the laptop market right now (not only possible to do NOW, but simplifies things like docking stations or allows for fast/large storage pools that haven't been available previously <faster than eSATA, assuming the model had an ExpressSlot>).
Even looking at Apple's current history... Apple likes to do yearly releases, this plays right into their yearly release schedule... And it lines up with a few other things I've heard.
They can do faster releases with consumer products (Intel based), as their update cycles are a bit faster. But they're still at Intel's mercy wherever they use Intel CPU's and chipsets. For their devices (iPod, iPhone, iPad), they have control over the entire process (design their own ARM based chips via the PA Semi acquisition), so have a greater capability of making sure they can meet their intended upgrade cycles (ARM has some influence here, but their products aren't as complex as Intel's, so can get revisions out faster).
The enterprise market OTOH, isn't as quick. Apple follows Intel's cycles here, and intermediate updates are too expensive (anything that requires a PCB redesign), as there's fewer systems to divide the costs over.
Intel has so far made TB out to be an open standard, which means they wouldn't be able to prevent anything... We'll see if that holds true.
They make and sell the parts, and could limit who they sell them to (based on intended use of said parts, not a "we don't like you" basis). I suspect there's compliance language in any sales contract that would prevent some uses.
I think Intel is saying a lot of conflicting things (such as that they wouldn't rule out a laptop TB card, which makes no sense if a desktop card is impossible.)
It would depend on the specifics, such as a few additional pins added to the ExpressSlot to transfer the MDP data (possible, nor difficult in the grand scheme of things; just update the ExpressSlot connectors needed and add a few extra traces). No additional semiconductors needed.
Assuming such conditions were needed, it would again only apply to new systems (card may/may not fit existing ExpressSlots for DATA ONLY operation - it would depend on how the slot connectors are redesigned).
