wbmw,
good post. You said it better than I did, since I got sidetracked in my post.
I think at this time, the leading developers are split between mapping the K9 project, and those that are completing evaluation of improvements to K8 Hammer. At this time, they probably have enough data from simulations run on real Hammer silicon to determine which bottlenecks can be removed from current hammer and which ones will need to go to the next core. I don't really expect much to be changed in 90nm Hammer, since it is following so closely behind 130nm versions, except the obvious L2 increases. another project in early stages may be a dual core hammer on 90nm.
There are number of areas where R&D efforts exploded compared to last technological transition. While there was a small team of hackers who internally developed the 750 chipset, AMD now has a full fledged chipset division, and they started another development center in Boston from ex-API employees, who are working on Hypertransport. This group is IMO a group that will prepare the infrastructure for Hammer to become a mainframe class processor within the 8th generations lifetime and beyond.
Plus, AMD acquired the small Alchemy team apparenly already has some products ready for sale. This division seems to be a long shot, but I think they are in a segment that is viable.
In addition to that, AMD is funding the process development of number of proceses, including the current 170nm flash, 130nm flash, 130nm SOI, 90nm SOI, a team that is moving K7 to UMC process, the early development team working on 65nm process technology.
THen, we had the team that came up with x86-64, a team that is supporting 3rd party developers working on x86-64 code, an inhouse team writing device drivers for internally developed chipsets / CPUs.
AMD has moved from being a cloner to a full fledged architecture developers, and is fully funding a future reasonably independent of Intel. The current revenue level can't support these activities (which is the challenge that needs to be overcome quickly). Underinvestment in the future is not the problem.
Joe |
