Conclusions
ATI is clearly the technology leader when it comes to implementing PCI Express for graphics, and that puts NVIDIA in the unenviable position having to justify its bridge chip. Fortunately, NVIDIA dodges ATI's jabs well and makes a reasonably convincing argument for why HSI is the right strategy for GeForce PCX.

Honestly, though, some of the posturing over PCI Express seems a little silly to me. For starters, we're only talking about PCI-E implementations of current generation graphics chips. Early adopters of PCI Express, at least in the enthusiast community, will likely be concerned about the PCI-E implementations of next-gen chips like NV40 and R420, not rehashed versions of last year's GeForces and Radeons.

All this talk of the relative merits of native and bridged PCI Express implementations also assumes that today's applications actually need the extra bandwidth that PCI-E can provide, but that's not necessarily the case. As we saw in the move from AGP 4X to 8X, doubling available graphics interface bandwidth doesn't necessarily translate to higher frame rates or better performance. Applications need to be designed to take advantage of PCI-E's abundant bandwidth before the interconnect can offer a tangible performance benefit over AGP 8X, and it could take games and desktop applications years to catch up.

Don't get me wrong. PCI Express has the potential to enable all sorts of exciting new applications and eye candy for PC graphics. However, I'd wager that the vast majority of enthusiasts will be using those applications with tomorrow's graphics chips, not yesterday's.