Well in terms of cpu cost Apple is generally quite a bit more generous on the laptop end. They move in high volume, and the displays are most likely much less expensive compared to a 27" IPS although pricing adjustments from NEC and Dell suggest that panel costs have come down. They tend to cater to a market of needs stable color but cost remains a factor. Both tend to drop pricing pretty aggressively when they can.
Anyway I'm getting off topic. With something like the imac you would be using a different chipset as I mentioned. This would also keep them on Sandy Bridge with Sandy Bridge E as opposed to migrating that configuration to Ivy. Given a specialized board design for one model with the testing involved for operating temperatures in a small area (note custom HDD firmware for temperature control) combined with higher cost cpus pushing the price up, you'd be headed toward a niche product which would likely require higher margins to survive.
It could easily push the price to $3k balancing everything involved. It sounds high, but we're accounting for more than the difference in cpu cost here. The cheapest mac pro uses a $300 cpu and a cheap graphics card. The same thing was true when it debuted. An upgraded 15" macbook pro uses a more expensive gpu and a cpu that started with a suggested price of $568. Volume is a factor when setting pricing. If something doesn't move as many units, they're likely to look for a high margin per unit. Obviously the idevices are somewhat of an enigma here. Anyway even at a much higher cost, they wouldn't have some of the features currently available to a mac pro. If you were going to pull the mac pro option completely, how many of those customers would be retained by such an imac?
Beyond that if you're trying to hold some of the 8 to 12 core customer base, you'd want to implement such a solution on a year where intel makes a huge breakthrough to minimize the hit from the lack of a dual package option. Sandy Bridge E wouldn't be likely here, then Ivy Bridge E would use the same board design, so there's no real reason to migrate there. The earliest you'd see such a solution would be Haswell on the Xeon end, but I still don't think it's likely. If 8 core cpus become the new workstation standard, then 6 will probably filter to consumer models on the high end where they share some overlap with low power Xeons.
There are a few accusations that Intel is slipping due to lack of competition from AMD, but if anything the use of Teslas poses more of a threat than opterons in HPC implementations. Given the problems with SATA implementations and stepping in Sandy Bridge, it's likely that they're being cautious before releasing new hardware.
