Making any automatic optimization based on a benchmark name is wrong. It subverts the purpose of benchmarking, which is to gauge how a similar class of applications will perform on a tested configuration, not just how the single application chosen as representative performs.Carmack discusses several facets of the issue, so I recommend reading the whole thing.
On a semi-related note, Kyle Bennett and I had a very interesting meeting with ATI during Comdex, just before they released their new Radeon 8500 drivers. Those drivers remove/fix the Quake III optimizations, according to ATI and according to Anand. I'll be testing them myself soon. Beyond that, I believe ATI is making a new commitment to the enthusiast community. Let's hope it works out well.