alan81, 32b -> 64b, I think you are confusing some things:
Over the past few years Intel has been converting all the 64b folks over to 32b Xeon servers...
This has been a great cost savings as companies offload tasks that previously used expensive big iron. This isn't any rejection of 64-bit, rather it is an embracing of cheap servers for tasks that don't require 64-bits (yet).
These customers' software needs will grow as the software grows. Right now there are lots of applications that will work fine in 32-bit, but given the opportunity to utilize large memory resources they will work better.
I am concerned about customers converting to Opteron, but so far that does not appear to be a problem.
For good reason. Server architectures are not the fastest things to launch, and AMD has been steadily building manufacturing and market momentum for Opteron. Still, this year's Opteron output does not challenge Intel very much.
What happens next year? Two things:
1. Opteron manufacturing takes off; with volume and a growing track record, more design wins are released.
2. Intel endorses AMD64? That is the rumor this week, that Intel will announce support with product in 2005. That will make Opteron and Athlon64 instant successes and sell lots of them as the first available products for the new world of 64-bit x86 computing.
The real question is whether AMD can get a major foothold into the corporate world so that they are entrenched when Intel's product comes.
One other thing to ponder: AMD really cannot compete (in terms of units) with P4 desktops because of manufacturing contraints. However, if AMD were to focus much of their manufacturing on the Xeon market, they certainly produce enough product to displace quite a bit of Xeon production.
