>> I here late '08, '09 as the when.
If this is indeed the info churning in the silicon rumourmill, then it likely is true.
As for Intel's answer to HyperTransport: Actually, I would not go into convulsions if Intel announced one day that they were joining the HyperTransport Consortium. But, the chance of that happening are at best 10%. About the only other *existing* microprocessor I/O interconnect that might have a chance at working for an X86 processor is RapidIO -- and that would be if Intel took a big interest in hijacking the spec and beefing it up. Then, there's the super fast, shiny, high-tech wunderkind data channel called InfiniBand, but nobody uses InfiniBand for processor interconnects.
However, InfiniBand is the data channel that *could* replace just about every other type of bus or channel except super-cheap USB. InfiniBand has the capability to replace ATA, SATA, SCSI, SAS, Fibre-Channel, Ethernet, and PCI. InfiniBand is currently the darling of the supercomputing crowd, but it should soon begin making inroads into the realm of the blade server, where the multi-core / multi-processor plug-in server blade has a very fast very-low-latency 5 Gb/s or 10 Gb/s InfiniBand connection to an InfiniBand fabric for all its I/O. Somewhere out on the InfiniBand fabric -- maybe even a half kilometer away -- are host channel adapters for direct connections to multi-port Ethernet switches, multi-port Fibre-Channel switches, multi-port SAS/SATA switches, PCI Express, general serial I/O, etc. OK, I'll quit babbling.
This leads me to wondering what the hell Intel has been cooking up in the labs to take on HyperTransport. Could there eventually be another PCI specification? Maybe something called Super PCI, or PCI Ultra, or PCI Grid, or the Mother-Of-All-PCI? Maybe a radical new PCI that does everything what PCI Express does now, but adds microprocessor interconnects for X86 and Itanium? This would be my guess if Intel doesn’t have a change of heart and join the HT Consortium (cold day in Hell) <cough>.