Dear Tench:
I believe that it is far more likely that server vendors after trying out Palomino at 384K exclusive in dual configs, asked AMD for a larger cache version sample either at 640K (512+128) or 1152K (1024+128). After looking at the performance increase on the 760MP for typical server loads, they decided that large cache Palomino makes sense for servers thus reconstituting Mustang. Now that there is demand for it, AMD will now start producing it. Since Tbird is fine for high end desktop and low to mid workstations, there is no need for Palomino at current cache. They think to use Palominos in high ASP areas, server and mobile. Once Intel gets desktops to near Tbird limits, desktop Palomino can use all the experience from the mobile and server versions and come to the rescue. The power users will jump to high end workstation grade Palominos on dual boards anyway.
All in all, AMD is trying to maximize profit by increasing the revenue for each processor's R&D investment. This is what we as shareholders want in the long run. Lots of profits per share of investment and a predictable rise in both revenue and most especially profits. Since we are capacity constrained (if we can supply all of the market for CPUs, going for the throat makes more sense but, if we can't, increasing profit per wafer makes more sense). AMD seems to have decided on the second for now. When infrastructure (fabs, assembly, chipsets, motherboards, and distribution) allows a complete takeover, they may change their tune.
Pete |
