Interview with Greg Buchner

[06.06.01] » The main man behind the GameCube graphics chip talks about moving Nintendo past cartridges and sizing up the console competition.

The motherboard for the Nintendo GameCube.

    When multimedia company ATI bought chip designer ArtX over a year ago, they not only gained a new business, but they also became responsible for the graphics chip inside Nintendo's next console. The Nintendo 64 had been widely criticized as an expensive system to develop for, and ArtX was charged with easing the gaming behemoth into an era of disc media.

    Under the watchful eye of Greg Buchner, now Vice President of Engineering at ATI, ArtX had already completed the "Flipper" chip that will help bring the next Marios and Zeldas to life. Buchner recently sat down with the GIA to describe the process behind creating graphics for a new round of consoles.

GIA: What's the first thing you deal with when designing the graphics chip for a gaming system, particularly for the GameCube?

GB: We thought about the developers as our main customers. In particular for GameCube, we spent three years working with Nintendo of America and with all sorts of developers, trying to understand the challenges, needs, and problems they face.

First among these is the rising cost of development. The GameCube can see high performance without too much trouble; it isn't a quirky design, but a very clean one. It was important we didn't require jumping through hoops for high performance to be achieved. On top of that, it is rich in features, and we worked to include a dream group of technical features that developers requested.

Many of the engineers here worked on Nintendo 64, but this is a different challenge, and we've learned a lot. The Nintendo 64 was focused on driving the change from 2-D to 3-D, but now everyone develops primarily in 3-D, and we're trying to extend 3-D to give better image quality and a higher overall level of performance.

GIA: The published ATI specs for the Flipper chip do mention a 2-D engine. Exactly what role does 2-D play in chip development today?

GB: For the most part Flipper is a 3-D pipe, but there are important operations and applications that are still 2-D in nature. Much less than a product aimed at a PC, but the chip's capabilities allow the GameCube a full range of 2-D operation and emulation. Fundamentally, texturing is a 2-D operation, so a lot of crucial things are done there.

GIA: The N64 is regarded as a mixed bag by most – 3-D horsepower that was hard to take advantage of, especially when it came to textures. What kind of changes have you made with the GameCube to address these problems?

GB: The N64 had raw horsepower, but a few problems cropped into the design that kept developers from reaching peak performance. You can see that now in our competitors' consoles; the numbers advertised as peak are simply not reachable. We've made sure that isn't true with the GameCube.

The real fundamental difference between the Nintendo 64 and the GameCube in terms of texturing is a large on chip memory in the GameCube for texture storage. There is also a large bandwidth for texturing, and both of these should easily take care of previous problems.

GIA: What other lessons have you learned or knowledge have you gained from other consoles?

GB: The GameCube has been made from scratch, and is a very new architecture and design, but we certainly looked at what had gone well and what hadn't for the PlayStation. There is a long history of working with Nintendo and their legacy, plus we got an inkling of how PlayStation 2 was going to be.

We drew from all those different consoles in terms of what was right and the direction we should go, what directions we should avoid. With that said, it wasn't like we copied anything.

GIA: Did the near-concurrent development of the PS2 and Xbox have any affect?

GB: The Xbox had little, but the PS2 added some real pressure to finish, as they put down a guidepost very early. They went public with what the PS2 was going to be spec-wise and architecture-wise about a year before the product launched, which is a long time.

When we saw their design, it really validated to us that we had made the right choice and done something different and efficient. They made some mistakes in the architecture, and it made us feel good about what we had created.

GIA: What kind of mistakes did the PS2 make?

GB: From a very high level, going from the PlayStation to the PS2, they made it harder to develop for. With the GameCube, we made it much easier to work with than the Nintendo 64. From a development point of view, it looks like they went in the opposite direction we went, and that isn't good.

GIA: How did you deal with the change from cart to disc media? It's something many people have been wishing Nintendo did for years.

GB: It was a huge factor, and the number one thing that held the N64 back from being a larger success was the choice of ROM carts instead of discs. Now we are on equal footing in terms of publishing cost issue, of inventory risk, and in production lead time. Back in the N64/PS1 days, there was a huge gap between system costs, to the order of $13 on the average game.

That is a fundamental difference, and there was a built in cost difference between two otherwise identical titles.  If you wanted to pick up a football game for the N64 instead of the PS1, it cost more for both the consumer and the developer.

GIA: Third-party support is something widely seen in the market as key for the GameCube. You talked about designing with developers in mind, but what role does a company like ATI have in advising and recruiting developers for a new system?

GB: Early on we were involved in listening to what their needs were, and we still provide some support. We did demos and examples of how you would do certain features, specific weather tricks or other effects. But Nintendo is primarily responsible for the day-to-day relationships. 

GIA: Nintendo has been vocally promoting a “games-centered” philosophy for the GameCube. How does a stance like this practically affect the chip's design?

GB: It provides focus. What we did at ATI for Nintendo is make the best graphics technology possible for gaming. It isn't meant to be a general-purpose widget. This means there aren't a lot of extra costs from features and options not aimed at their marketplace, costs paid by everyone. By doing that, we've produced the best chip for them and for the gaming community as a whole.

GIA: How deeply is Nintendo involved?

GB: It is a very collaborative environment, but ATI created 90% of the technology. We didn't create it in a vacuum, but we weren't simply handed specs and told to make it. Many of the folks at Nintendo we had worked with in prior lives at Silicon Graphics or General Electric. There were both philosophical discussions and technical discussions, but the technology was our job to create.

GIA: Considering ATI's strong reputation for DVD playback on PCs, do you think it was a waste not to include DVD in the GameCube?

GB: As you saw, a DVD/GameCube hybrid by Matsushita was displayed at this past E3 both on the show floor and during the press conference. They are a very strong electronics brand [as Panasonic in the US] and the perfect company to market products that include features beyond gaming.

Which one they want to introduce first isn't something we can speak for them on, but not having the GameCube have DVD playback was definitely the right answer. There would have been added costs, a dramatic physical size difference, and if you look at the other two consoles, an equally dramatic cost difference – 100 dollars at retail. And we know the higher priced competing consoles subsidize their hardware quite a bit, which is something Nintendo likes to avoid.

GIA: Was Flipper designed with those hybrid products in mind?

GB: Yes.

GIA: So we can look forward to more hybrid products in the feature?

GB: The DVD player shown is one of several options Matsushita has for bringing products to market.

GIA: What has ATI learned from working on the GameCube hardware?

GB: The team that developed the Flipper chip is now working on other developments for ATI, so lot of the ideas and concepts we brought to life for Nintendo will be showing up in future PC projects for ATI. We look to be raising the bar on PC products in the future and bringing what we learned into that world as well.

GIA: Microsoft has been heavily promoting its partnership with Nvidia in regards to the Xbox, something that has made Nvidia more visible and even raised its stock price. Nintendo has been less vocal with ATI's role. Would ATI like that attention?

GB: At the end of the day, it's being successful in the market and making money there that matters. We would certainly love them to talk about our technology, but Nintendo is focusing on what's important, and that is software.

Microsoft is trying to make this into a tech war, which I think we would win hands down. But focusing on the content is what the end consumer is going to see and care about. I know we've done an incredible job and created some remarkable technology, and while I'd enjoy seeing Nintendo promote it, I don't think that would change our market success in the end one bit. 

GIA: Speaking of Microsoft, they've been touting the Xbox as having a PC-like development environment, at least in terms of simplicity. How would that be an advantage over something console-specific?

GB: If you are doing a graphics chip for the PC, there are lots of non-gaming applications and issues to deal with, which is one of the reasons you see such a large market for dedicated 3-D accelerator cards. The programming interface on a PC was not created by the same people who did the hardware, and the whole idea of PCs being so simple to develop for is overrated.

In a console the hardware and interface are developed together, meaning you can end up with something that has the same features and same capabilities as a PC but is much cheaper. There are tons of random issues to deal with in PC development, but with a smart console design, there is a lot more freedom to do things right.

Anybody with a PC can call themselves a game developer, and because the Xbox is basically a PC, you're going to see plenty of that. What is meant by PC-like development is that porting games between the two is easy, not that the development itself is easy. Overall, PCs are somewhere between the GameCube and PlayStation 2. 

GIA: A built-in hard drive is probably the Xbox feature newest to consoles. Will not having a hard drive be a problem for the GameCube?

GB: I look at successful games and I just don't think it is a factor. It will add to the complexity of games, and that leads you back to lots of issues – does a new program change the drivers? What happens when the drive is full? There's a hidden complexity and huge hidden cost there, and I can't say it'll change things in a positive way.

GIA: You began your career at General Electric, making graphic engines for technical products such as flight simulators. What kind of unique challenges are there in producing graphics for games instead?

GB: I've gone from 30 million dollar flight simulations to $200 GameCubes, and the challenge is across the board. With gaming, you're trying to get something that is widely manufacturable, requires almost no support, and is very, very cheap. At GE every single machines was custom. They took 50 engineers to make, and 10 shipped with it (laughs).

Now we want to ship millions of products a year without a single support call, millions of GameCubes.

GIA: Thanks for taking the time out to chat, and good luck with the GameCube.

GB: Not a problem; I hope everyone enjoys our work.