Because USB 3 is, near as I understand, actually a different protocol than USB2. However, they have piggybacked the USB3 connectors (and wires) into the USB2 cable by using two different plugs in one connector. Depending on which USB standard the device supports only the USB2 or the USB3 wires will actually connect.
To the consumer it will appear that USB2 and USB3 plugs are the same.... but in fact they are very different .... and... most importantly makes it easy to use a USB2 cable to connect to a USB3 device. This creates complications and confusion.... and as we know Apple likes things simple.
IMO, Apple is avoiding the mess of USB3 (for regular consumers) and waiting for Light Peak. Light Peak, if implemented as designed, will allow a single type of cable to connect everything. Ethernet, video, audio, external HDs, keyboards and mice, speakers.... everything. You take your peripheral, take up to 100 ft of cable, plug one end into the device and the other end into any d*mn available port you want. No more trying to fit the USB cord into the FW port, the telephone cord into the ethernet port, etc etc. because you're on your knees, reaching around the back of the tower under the desk, trying to feel with your fingers where to plug the end of the cable.
Stay tuned. Probably announced next spring.
There are some technical issues with current USB 3.0 chips, but it's most likely to do with costs.
It would have meant an additional semiconductor, which means more money and work (PCB would need to be reworked to accomodate it, and it would have affected the lane configurations as well). Remember, the Tick Tock cycle is designed so that the same board (socket, chipset, and ICH) can work for 2 CPU generations by means of a firmware update (no hardware changes are necessary).
There's also the small issue that USB 3.0 is technically capable of exceeding the 500MB/s per lane on PCIe Gen. 2.0 standard (though realistic throughput is only about 400MB/s due to latency under ideal conditions = no bandwidth limitations between the USB 3.0 controller and PCIe lane configuration).
2009 boards were designed in a 16 + 16 + 4 configuration where the 4x lanes are shared via a PCIe switch for Slots 3 and 4. So to dedicate a lane per USB 3.0 port, the slots would have had to be reconfigured, though the PCIe switch could be eliminated if they had.
Intel has stalled USB 3.0 implementation in their parts to allow Light Peak a chance of gaining traction in the market, which means those not willing to add the separate semiconductors to their products (cheaper to stick with Intel's parts only), will wait. Exactly what Intel's hoping for, so they can get LP's parts into full production and into distribution channels.
But if a user needs USB 3.0, there is the possibility of adding a 3rd party PCIe card (CalDigit has one out, but I wouldn't trust their offering and wait for another vendor to create drivers for OS X and release a product usable in a MP).
Well yes, I was mearly saying that ICH11 or whatever it is will be coming with X68
I will, though I'm not sure if they'll stick with that naming scheme.
Given Intel's statement of "2012", it could mean the back end of the year, not beginning. Which means USB 3.0 may not be available for the consumer models of Sandy Bridge (LGA1155 parts). This seems likely to me from available information, and won't possibly surface until the LGA2011 parts ship and show up in systems (released for distribution in Q3/4 2011 <EP and EN parts releasing in different quarters>, and systems more likely to show in early 2012 due to final validation and assembly testing).