We now know several things about this new mobile part, including the fact that it’s scheduled for release some time next year. Given that schedule, it will almost certainly be manufactured at 65 nanometers. AMD plans on making two versions of this chip, one of which will be aimed at ultra-low-voltage applications.
With AMD’s stated intention to use more modularity in its chip designs, however, it’s difficult to pin down how much different (and thus mobile-specific) this processor will really be. Many of its features are familiar from current AMD processors, including a dual-core design with an internal SRI, a crossbar switch, and an integrated dual-channel DDR2 memory controller.
Other features, including those AMD cites as mobile-specific ones, are familiar from AMD’s stated plans for its next-gen quad-core desktop and server CPUs. Those include separate power planes for the two execution cores and the memory controller, the ability to halt one CPU core entirely if it’s not needed, and the dynamic link power scaling that is a part of HyperTransport 3.
That sounds an awful lot like the Turion 64 all over againnot that there’s anything wrong the Turion or with using a common CPU microarchitecture, as Intel seems to have learned through painful experience. But this talk of mobile-specific designs can get a little thick, so I pressed Hester on this point. He insisted this is a real and separate engineering effort, more so than the Turion. The distinctive mobile optimizations seem to boil down to two elements, one specific and one less so. First, AMD has redesigned the memory controller for lower power use. The Opteron memory controller has a server heritage, he said, and this reworked one will better serve the mobile market (and the desktop market, too, incidentally.) Second, Hester alluded (and this is the less specific bit) to changes in the way power is delivered throughout the chip.
One way the new mobile processor will differ from AMD’s forthcoming desktop and server parts will be its use of the current K8 execution core, not the new one in development. That may be an indication that development of these chips really did branch off from AMD’s desktop and server efforts some time ago, and perhaps that we could see products sooner rather than later in 2007.
Future efforts on this front from AMD could be even more interesting. Hester confirmed without hesitation that AMD will continue mobile-specific engineering efforts when this one is complete, and that those future efforts would include mobility-tailored modifications to the processor execution core itself. AMD intends to push into ever lower power envelopes with its mobile processors going forward, because it sees smaller mobile devices becoming increasingly important over time.