A little more on next-gen graphics

— 11:20 PM on August 12, 2002

Thanks to all of you who have written nice e-mail messages about my next-gen graphics tech article. I've been overwhelmed by the response. This article was a bit of a depature for us, and we'll try start producing similar articles when the situation warrants.

Some of the mail coming in has been particularly interesting. For instance, I didn't know that Mark Peercy, author of the multipass rendering article I referenced, now works at ATI. In fact, he was one of the architects of the Radeon 9700. That explains a few things, eh?

Also, one of Peercy's co-authors on the multipass rendering paper, John Airey, wrote and offered his two cents on the future of graphics chips. Now that graphics chips are "datatype rich," and not just feature-laden, he's predicting a radical simplification of GPUs in coming generations. General programmability, not specific function units, is the future according to Airey, and I can see why he says so. Once developers are writing high-level shader code and feeding it into a compiler, there's less reason to have so much complexity at the ASIC level, and such complexity only slows chip development. Interesting stuff.

Oh, and there's an intriguing problem I hadn't anticipated with rendering "production" images on an AGP card that we're examining, but I don't want give too much away just yet. Stay tuned.

Tip: You can use the A/Z keys to walk threads.
View options

This discussion is now closed.