As a follow up to our post from a couple of days ago (before the site went down), Anandtech & Tech-Report have posted their own Half-Life 2 benchmarks that show pretty much the same kind of performance difference seen in those numbers released by Valve earlier.
NVIDIA has officially responded to the early benchmark release (given it's an unreleased game/platform for the time being) saying they will have a new set of Dets 50.x that will improve Shader implementation which should reflect directly on performance gains... For some reason however Valve didn't allow webmasters to use unreleased drivers on its unreleased game, so we are stuck with 45.x performance numbers for the time being. Afraid of cheating perhaps? We must support Valve's decision however & we are sorry NVIDIA but you did it big last time.
AMDmb released some DX8 benchmarks using new Detonator drivers however that makes little sense at this time. Guru3D, in the other hand, has posted some 'unofficial' numbers using the new set of drivers running Aquamark 3 which is meant to be a straight DX9 testbed. The gains are there but we won't be satisfied until drivers can be widely tested under different conditions.