Obviously the industry is in transition from DX9 to DX10. How much did that impact the design, and how would you expect those tradeoffs to manifest over the lifetime of the R6 family?

What a question! DX10 basically decided everything in the R600. We kept DX9 in mind, and certainly wanted to make a great DX9 part and did (basically, we are generally 1.5 to 2x the performance of our previous high end part in DX9). But we focused all the new hardware on the DX10 feature set, and we believe we have a great part here. Regretfully, not only do we need a great part (which we have), we also need a whole new driver for DX10. This means that our system is a little green yet still on DX10, but I expect only greatness over the coming months and years for this part, in DX10.

Did any other industry trends make a noticeable impact on the technical design? 1920x1200 monitors (and larger) becoming more mainstream? HD digital media? Etc (feel free to add here)? If so, how did those considerations manifest in the silicon?

Changes occur all the time in our industry. Memory technology, interface protocols, new manufacturing technology, default resolutions, CPU performance and memory, HD-DVD/Blu-Ray, etc… We certainly focus on the appropriate ones for every chip. In the specific cases you mentioned, HD media and resolution increases, both of these were targeted in the HD 2000 series. The sweet spot resolution was re-targeted, and this led us to virtualizing all the compression and hierarchical memories used in the chips: On R5xx, we targeted 1600x1200 for compression and hierarchical Z; for R600, these are large caches now, capable of dealing with any resolution.

That's just an example. We also developed the UVD solutions for HD 2400 and HD 2600, so that they could run in very low power mode and display a full resolution HD movie on a notebook, in one battery. The higher end has a shader based solution, which leads to higher quality (but has more power consumption). Another example was HDCP keys allowing for encrypted data to come over as well as audio encoding and control into the HDCP port; basically, we have all the key critical features required for HD, and we designed that in 2004. There are myriads of other examples...

Many of us remember long ago that Dave Orton was very excited about what was then called R400. Obviously, a great deal of roadmap transitioning has happened in the years since. How similar or dissimilar is R600 to that original vision for R400?

Pretty different – Though some of the best aspects of that design were leveraged for Xbox360 and now for R600.

With Xenos being embedded in a closed platform, how much help was there, if any, on the driver side leveraging Xenos driver code for R600?

Regretfully, very little on the driver side. Xenos is not a DX10 part, and so nothing could be leveraged in terms of that driver development (the Xenos driver was written by Microsoft). As well, Xenos had no OpenGL, so we had to have a new OpenGL driver for HD 2000. And finally, the DX9 driver was mostly inspired from the R5xx DX driver, and so little influence came over from Xenos again. However, from a SW porting standpoint, we can truly make a difference. The HD2000 family's architecture leverages so much from Xenos, that porting Xenos games to the R600 should be relatively easy. The games should have similar performance characteristics. There's a very good synergy there.

How would you describe the maturity of R6 drivers at this point and how "target rich" the optimization opportunities still are going forward?

I think we have a very dedicated and excellent set of engineers working on our drivers. However, given a new operating system, a new DX, a new OpenGL driver and a whole new architecture, we got the perfect storm! I think that the drivers are very stable and give users a good experience. But when it comes to performance, there's still a lot to optimize for. And it will take time to exploit the full performance potential of the hardware. They have done what was required, which is give the best $399 part, but there's so much more to do… So I would say that a lot of progress has been made, but things are still “green”, especially on the DX10 front.