Graphics for PCI Express to pose real cooling problems

Jeffrey Cheng from ATI, Michael Abel from INTC and David Reed from Nvidia all agree that the faster bus will give better bandwidth and scalability, allow large request size and pipeline depths, and give isochronous support. The bandwidth will deliver simultaneous 8GB/s concurrent peak bandwidth in X16 mode. PCI Express will also provide a pipeline depth of up to 256.

Graphics cards using PCI Express tech will, however, require 75 watts max, and that means a few things. First, the 12 volts on an ATX power supply won't do the trick. The answer will be to use a 2x12 connector with the same pinouts as server SSIs, which requires a 300W power supply.

The other problem is that machines will need to use side panel vents, ducting, and use large fans.

Cheng will claim that the entire graphics industry will be "reset", AGP will die, and new features will "open doors" for emerging graphics and multimedia apps. A PCI Express graphics card, however, will run at PCI compatible mode, with few software changes.

There are some clear differences between managing non-local memory in PCI Express, however. Developers have to check 64-bit compliance.

The kind of graphics applications include production grade video editing, massive monitor array architecture, "hot plugging" of graphics, and using graphics as a co-processor are on the cards, Cheng will say.

Nvidia's Reed will say that graphics performance will be increased by around 10 per cent.

All three agree that there will be new 3D user interfaces that take advantage of PCI Express improved graphics. We do so hope that Microsoft and its Longhorn team are listening. µ