The year in graphics
2001 has shaped up to be one of the more interesting years in recent memory. While NVIDIA was initially believed to be moving away from its six-month product cycle, ultimately the company responded with a number of product releases in both the desktop and mobile segments. Once Microsoft released DirectX 8, NVIDIA was the first graphics company to release a product that took advantage of its capabilities with its GeForce3 graphics accelerator.
Earlier this fall, NVIDIA released its successor to GeForce 3 with the GeForce3 Titanium 500 and 200. In an attempt to learn more about these products and the technology that lies underneath, we conducted an interview with NVIDIA's David Kirk, Chief Scientist at the company. As part of this role David spearheads the technology development that occurs within NVIDIA. This makes him an excellent source of information for us, and hopefully an entertaining for you, the reader. For this interview, our questions are in bold and Kirk's answers are in normal font.
FiringSquad: The GeForce3 Titanium boards utilize an "enhanced" 0.15-micron manufacturing process. Exactly what has been enhanced in this new process from TSMC?
David Kirk: I don't know that I can really go into the details, but I can give you an answer in broad strokes. When you're building chips, there's a "recipe" - a little bit of this, a little bit of that. As we gain experience in a particular process (as in 0.15-micron), we can play with some of the details, and choose to make some minor changes in the manufacturing process that will provide enhanced speed, and reliability. One reason that we are able to do this is that our semiconductor manufacturing partner, TSMC, works very closely with us to meet our needs. TSMC and NVIDIA have a great relationship that benefits us both.
FiringSquad: Titanium launch documents mentions "vertex program optimizations" and "lightspeed memory architecture algorithm optimizations". Exactly what types of optimizations have been made and what kind of performance improvement will these optimizations bring?
David Kirk: When we produce a new product, such as GeForce3, often the first driver release does not take advantage of all of the new hardware features, sometime including performance enhancing features. When we release new drivers, and when we release a new hardware version, we are able to fine-tune the performance, giving our customers a continuously better experience. Specifically, for this product release, we have optimized commonly used vertex program fragments. What this means is that programs that are often used by developers will simply "get faster" with no extra work.