Advanced Micro Devices Inc (AMD)

[mmoy]

Did you read my questions on Woodcrest from Thursday night
and Friday morning? It's pretty clear to me that the big
cache and aggressive prefetching account for the majority
of performance improvements in Conroe. That works really
well for desktops and laptops but servers typically have
far more processes running and trying to run aggressive
caching on that kind of a workload would be counter-productive
as processes flush out what other processes need.

I was a bit surprised that none of the Intel folks that are
more technically-oriented responded to my posts. CJ was the
only one to make any comments about it.

I think that the use of cache and aggressive prefetching
is a good way to decrease latency to the point that it
nullifies the integrated memory controller but you have
a limit as to how much that cache can be used. But as
the number of processes goes up, you wind up with more
and more cache-misses which is where the integrated
memory controller should reassert its advantages.

My gut feel tells me that the two-year low isn't going to
hold here. The AMD chart is also bearish and I posted that
on Friday. I need to calculate retracements on Intel. I
already did so for Intel. It looks like shorts may be
profitable on both companies though AMD probably has more
to lose percentagewise.