Thursday, August 19, 1999

When 3dfx announced their new T-Buffer technology, I admit to being a little underwhelmed. After all, I was expecting to hear some news about the upcoming 3dfx Voodoo4/Napalm card, and here we are receiving info on a new method of processing 3D images. After my initial disappointment wore off, I traded a few emails back and forth with 3dfx PR Manager Brian Burke, and he kindly offered to clear up any questions or concerns I had about the T-Buffer announcement. Not being one to look a gift horse in the mouth, I sent off the 10 questions you will read below. I hope you will find them useful in understanding more about the T-Buffer technology, API support, as well as the 3dfx demonstration itself.

Extreme Hardware: What is the visual difference between using a T-Buffer card with current 3D games (like NFS3) compared to developer-supported T-Buffer 3D games that will come out later?

Brian Burke: As for the spatial AA, there will be none. All the APIs have support for full scene AA already. Because of the huge performance hit on today’s 3d accelerators, it is seldom implemented. As for the other features like motion blur, soft shadows and depth of field, they are not supported in any APIs. We will be faced with a task like we had when we introduced multi-texturing with the Voodoo2, in that it will take some evangelizing on our part. We control GLIDE, so they will all be supported in GLIDE. We are lobbying Microsoft to have them included in the next version of DirectX. We also can add them via the OpenGL extensions, or the DirectX extensions. I do not think it would be fair to characterize the fact that there is not support for these new technologies in current APIs. 3dfx is taking steps to advance 3d technology instead of doing the same things other 3d accelerator manufacturers are doing. It is good for the industry when someone pushes the envelope, it forces everyone to step up and deliver better products to the consumer.

EH: Is there a speed difference between the two types of T-Buffer usage (regular and developer-supported) in terms of 3D performance?

Brian: There is not a speed difference between full-speed AA when we "force" it on via a control panel option or when the developer turns it on in the game.

EH: Have you done any performance testing with T-Buffer in 16-bit colour? 32-bit rendering? If so, is there any way you could let us have some specs?

Brian: We have been working on this technology for sometime and have done extensive testing. Since this was a technology announcement and not a product announcement, we can not release any specs, yet. I will say that 3dfx is pushing to have 32-bit rendering with full scene AA for benchmarking the next generation of 3d accelerators. We also maintain that anything under 60 frames a second is unacceptable for gaming.

EH: How does the API affect T-Buffer performance? Is it different/better with D3D, OpenGL, Glide?

Brian: The T-Buffer is very API agnostic - it doesn't require much in terms of extensions to an API so it should basically be the same in all APIs.

 3dfx T-Buffer Screenshots

EH: Is there any chance that this T-Buffer technology will be licensed to other manufacturers? Have you approached Microsoft for inclusion into DirectX?

Brian: We are already working with Microsoft to have the 3d rendering techniques that the t-buffer provides into DirectX. We have not been approached by any potential licenses to date, but we would be willing to listen to offers if they made sense for our business model.

EH: When do you expect to have T-Buffer hardware available to gamers?

Brian:They will be on the shelf this fall for the Christmas buying season and will be priced in line with mainstream consumer 3D accelerators.

EH: Are you looking at enhancing your current product line (Voodoo3) with T-Buffer hardware?

Brian: No, the t-buffer will be a feature of our next piece of silicon.

EH: Even though Sharky had a pic of the 4 Dual-Voodoo2 boards, could you confirm that was the platform the demo was done on?

Brian: Yes, the demo was done on an emulation machine. We were very up front about that fact and did not try to fool anyone. The demo was done on the Heavy Metal machine that was put together by our partners at Quantum 3D. It is about a $10,000 solution for the full scene AA problem. We will be solving it this fall at prices that consumers can afford.

EH: Also, why was this platform used? Was it timing or did you not want to unveil the Voodoo4/Napalm at this time?

Brian: We have some education to do before we unveil our next product. It will be bringing some new features to 3D rendering. By getting the information to the public now, we can get more acceptance and understanding for these new features when the product ships. It is also just part of the product introduction we have planned for our next generation architecture.

We used the Quantum machine for a couple of reasons. First, we did not want to introduce the next generation product at this time, we wanted to start the education process for the new features. Second, prototype hardware is a very precious commodity. We did these demonstrations over a period of two weeks and had a good deal of preparation on top of that. You can do a lot of engineering over that time period with a prototype board. They more boards you have to work with, the more you get done and the faster you ship.

EH: At what level is the T-Buffer performance gained through hardware vs.software (IOW, is it a 100% hardware solution?).

Brian: It is a hardware solution, 100%.

Interview by: Vince Freeman from Extreme Hardware

Copyright 1999 - All rights reserved Hilbert Hagedoorn