Replies to post #2610 on AMD strictly moderated (AMD)

[chipdesigner]

I think Intel chose benchmarks that:

(1) Took advantage of the bandwidth limitation of 939 vs. AM2 (it doubles)

(2) Took advantage of the ability to use 4MB of L2 for one core.

(3) Took advantage of who knows what else, given the special software, and closed boxes, and lack of current P4 performance reference point.

Number 1 will be remedied across the line by AM2.

Number 2 will be addressed by Rev F FX parts, which, per the Inq will come with L3 cache, probably 4MB.

Number 3 will be addressed in the fall, when independent reviews are conducted.