“When Cam and J and Jon Thomason came on board, the project hadn’t gotten anywhere yet. I mean, it’s also completely fair to say, those guys—Otto, Ted, Kevin, Seamus and Nat—were all on board before the project was approved. So, when you ask when did the Xbox begin, you know, you could say it was a gleam in Otto Berkes’ eye. That’s a fair statement. You could say Bill approved it in July because he said hey, there’s some good product ideas here. You could say Bill and Steve approved it in December, which was a different outcome than what Bill was looking at in July, but still ok. And you could say, well it didn’t really start until February because that’s when they approved what we actually decided to do.”
-Robbie Bach
When word of the console project began to spread, most people were dubious at best. A typical reaction was described by engineer Dave McCoy. “I remember when a lot of us first heard that Microsoft was going to build a console, our first reaction was to laugh because we thought it was such an untenable thing. I mean even inside the organization, it was like, ‘Hey guess what? We’re going to build a console.’ ‘What? Are you kidding? That’s never going to work. That’s preposterous. Microsoft doesn’t know anything about this.’”
Even after McCoy had been absorbed into the team, he and many of his colleagues were still having a hard time believing in the project. “We thought it was preposterous, but here I was working in a small little group and it was all over the map in terms of what was this thing going to be. Was it going to be a device that just runs Windows and is it going to have a keyboard. Is it going to be this or is it going to be that? I remember seeing a version with a little C: prompt on it. Ok. That’s cool… I guess.”
Jon Thomason, who would lead the Xbox operating system team, said, “There was all this posturing. No code had been written; no hardware had been built—on either side. All the people posturing, wanting to do this. And all these sort of demos that Otto and Seamus and everybody did, there’s nothing but smoke and mirrors there. And that’s ok. That’s what makes it a great story, because there were all these forces arrayed against each other with no substance and yet somebody won. And then a whole other team came in to build it. By the time I got on, my team… we didn’t spend one second doing anything but real code. No demos. Any demos that happened came from the ATG team.”
The Original Xbox Team When Approved
Corcoran, Elizabeth A.
Booth, Jennifer A.
Wilcox, Jon L.
O’Rourke, John S.
Allard, J.
Ferroni, Cameron J. A.
Spector, Barry J.
Del Castillo, Leonardo G.
Gibson, Greg S.
Kwoka, Paul J.C.
Liu, Dick C.K.
Reents, Jeffery M.
Blackley, Jonathan L.
Dernis, Mitchell S.
Engevik, Deborah L.
Ng, Brenda J. T.
Schnepf, Brett A.
Coyner, Donald R.
Daugherty, Gregg R.
Holmdahl, Todd E.
Stewart, James R.
Walker, Robert S.
Hebenthal, Douglas C.
McNulty, Mark J.
Mooney, Dennis
Vingerelli, Enrico G.
Friedrich, Bernd
Roshak, Todd G.
Thomason, Jonathan G.
Sharpe, Tracy C.
de Leon, Juan Carlos
Henshaw, Jeffrey D.
Bach, Robert J.
Angeloff, Drew A.
Hufford, David E.
Chips for Xbox
With Bill Gates weeks away from taking the stage to announce the Xbox to the world at GDC, two of the most important Xbox components had still not been determined. The story behind the Xbox CPU involves two major players and some last-minute surprises while the Xbox GPU story involves one clear industry leader, an unlikely upstart, and some hardcore negotiating.
The major contenders for Xbox CPU were Intel and AMD, although Rick Thompson says that Philips was also trying to pitch Microsoft on a custom CPU. Initially, Microsoft chose to use an AMD chip, in part because it had better performance than the current Intel chip and also because they couldn’t come to agreement on the price with Intel. Both the original fast boot demos for Bill Gates and the infamous Silver X had been on modified Windows machines using AMD chips. According to Rob Wyatt, they had become comfortable working with AMD’s ATX boards. They provided better performance for games than the Intel Pentium chips that were available at the time. At the hardware level, there was a lot of support for AMD. In addition, while Blackley and others were traveling around making their pitch for Xbox to developers, they were saying that Xbox would use AMD’s CPU and NVidia’s graphics processor, so that was the expectation.
Management wasn’t quite as loyal to AMD as the engineers were, and Wyatt remembers getting an email from Steve Ballmer. “All it said was, ‘Don’t fuck up the Intel account.’”
Bob McBreen was directly involved in the negotiations with various chip manufacturers, both CPU and GPU chips. He visited both Intel and AMD to see who would offer the best terms. His first experience negotiating with AMD went very well. “They were great partners, fabulous negotiating.” However, when he approached Intel, it was quite different. “Rick and I had a call with (Intel CEO) Pat Gelsinger and told him that we were doing this Xbox thing and that we had some very aggressive pricing that we wanted them to hit on chips, and in addition to that we needed them to develop chips with a faster front-side bus, and they basically told us, not interested. Not our business.”
So with Intel off the table, discussions with AMD got serious. They met in Las Vegas with AMD’s CEO, Jerry Sanders, and “a bunch of mucky mucks,” but now AMD was asking for a very much more aggressive deal. They wanted Microsoft to take a $200 million equity position in the company and also pay aggressive rates for the chips. “We kind of choked. All we were trying to do was buy chips, and they were trying to leverage it. ‘Wow, AMD is now a partner with Microsoft, and not only are they a partner, but they’ve taken an equity investment.’ And we told them they were crazy. They thought they had us over a barrel because it had been quietly leaked to everyone that AMD was our partner. And they threw this thing, at the last minute, and stared us down.”
McBreen says that it’s possible that AMD knew that Intel had turned them down and figured they had the upper hand, but the Microsoft negotiators told AMD to take this proposal off the table and come up with a new proposal without the equity requirement, and they said no. The best they would offer is some adjustment to the pricing. As soon as they returned to Redmond, they called Intel again, and this time was different… they were interested in dealing. “And so, for a period of time, I was going back and forth with Intel and AMD, and AMD I don’t think figured out that we had a backup plan. They thought that we were stuck, and so the negotiations with Intel picked up really dramatically.”
Meanwhile, it was almost time for Bill Gates’ big Xbox reveal at GDC, and they still didn’t know what processor would be in the box. According to McBreen, Rick Thompson had already left for Japan to make the announcement there, and his slide deck said that the processor was going to made by XXX. McBreen was still going back and forth between the two companies and keeping Bill Gates updated.
The issue with Intel was that Microsoft wanted them to promise a more powerful chip, one that would equal or exceed the performance of AMD’s chip. It came down to a dozen Intel negotiators with McBreen on the other side of the table. “And at the end, at the last minute they agreed. I sent the deal sheet over to Bill. Bill said, go with Intel. Got the information to Rick in Japan, and then all hell broke loose with AMD.”
CEO Sanders called McBreen right after the announcement, “and just went crazy. Absolutely crazy. We tried to tell them what was going on. We tried negotiating with your people, and they kept coming back with this equity thing, and we kept telling them we don’t want to deal with equity, and they kept coming back with it. And he was like, ‘Why didn’t you call me directly?’ It was crazy. Then he ended up calling Bill, but by then it was too late. We’d already done the deal with Intel.”
The Xbox GPU
From the beginning of the project, NVidia was expected to supply the Graphics Processor Unit (GPU) for Xbox. In part, the decision was based on NVidia’s quality and reputation as well as the relationship that some people at Microsoft had already established with them. Another reason was that it would lend credibility among third-party game developers that Xbox graphics would be powered by NVidia. And they needed those developers on board.
Negotiations began long before Microsoft had planned to reveal the project to the public, but word had leaked out somehow, and it made negotiations more difficult. McBreen was once again one of the principal negotiators for Microsoft. He says, “NVidia believed that there was no way we could ever choose anyone else for a graphics processor because it had already been leaked. They were at the top of the heap for gaming companies, and so if we came out and said we’re going to do S3 or somebody, we would lose the 3rd party gaming companies. So trying to negotiate with (NVidia CEO) Jen-Hsun was next to impossible. He was very proud of his technology; he knew he had us over a barrel.”
To complicate matters even more, the recently purchased WebTV team got involved. McBreen notes that the “honeymoon phase was still very strong… they were the darlings. They knew hardware. They were technology mavens, and so every decision we would make, Bill would say, ‘Hey. What do the WebTV guys think about this?’ Well, the WebTV guys did not think we were smart at all. They thought of themselves as silicon vendors, and they had this company that they had been talking to for doing graphics stuff on future versions of WebTV. It was called GigaPixel.”
Gigapixel
George T. Haber is certainly not a major character in the Microsoft saga, but his story is worth telling, if only because Microsoft almost made him a multi-billionaire. Haber was a Romanian immigrant who came to America to follow his technological dreams. He came from the Transylvanian region on the border with Hungary. “All the Dracula movies and things that you read and hear about vampires, for some reason based in this mountainous place.” Haber’s history is remarkable. His mother was a survivor of Auschwitz, and his father had fought and been captured first by the Russians, then by the Germans, and then by the Russians again. He says, “I was brought up with this thinking that the only thing that matters in life is the thing that you can carry with you wherever you go, which is basically your brain and your education, and the persona you become by learning.”
Haber was encouraged by his parents to explore art and learning, and at one point he picked up the guitar. “Back in Communist Europe, you could not build amplifiers or buy amplifiers because there were no materials. There was no guitar shop. There was no music shop. So me being passionate about music, I had to study electronics to figure out how do you actually build an electric guitar out of regular guitar by putting the coils to pick up the sound under the strings and then connecting them to transistors that will amplify, and then in the end, building a speaker out of paper coils and rubber from bicycle wheels. So, as it turns out, I probably was better at building these things than playing the guitar.”
Because of some political negotiations between the communist government and the United States (that had nothing to do with Haber personally), he was able to go to Israel and study electrical engineering at Technion—the Israel Institute of Technology. In 1988 he made the decision to move to Silicon Valley and pursue his dreams of becoming a technology entrepreneur. He worked at a company called Daisy Systems, where he met Vinod Khosla, who would later become a famous and powerful venture capitalist. Following Khosla’s lead, he took a job at Sun Microsystems. “I was employee number 100.” Haber worked on floating point mathematics for Sun’s RISC (Reduced Instruction Set Computing) processor and got a lot of experience in processor design. “Once I finished that job, I started looking around, and I saw all these colleagues that were working at Sun leaving and starting their own startups. So that bug bit me as well, and I decided to leave Sun and go and join a startup.”
Haber heard through friends that Silicon Graphics Incorporated (SGI) was looking for an expert in floating-point processing to help in building a secret graphics engine. Working there, he met some of his future partners. Next, he created his first startup, CompCore, which specialized in video compression and decompression algorithms for use in both software and hardware.
CompCore released some very successful products, including SoftPEG and SoftDVD, which were licensed by PC manufacturers as part of the standard OS designed for playing back DVDs. They also started working closely with Intel, and Haber was invited to sit in the front row at a keynote by George Grove with Bill Gates and Michael Dell. “And I’m an inky-dinky guy from ‘Eastern Europe nowhere’ sitting here with the giants of Silicon Valley, in the same row, because I had the software to play back DVDs.” Microsoft even invited Haber down and made what he called an obscene offer, “obscene in the sense that it was so low that it made me laugh, to buy the company, to the point where our annual sales were almost ten times bigger than what I was offered to sell the whole company.” He says that Microsoft offered about a quarter of a million dollars when he was bringing in eight to nine million a year in revenues. Eventually he sold the company to Zoran for $80 million.
After selling CompCore, Haber was approached by two of his ex-SGI colleagues who had developed a 3D technology that was faster and more efficient than anything on the market. They had a small amount of funding from Andy Bechtolsheim, one of the founders of Sun Microsystem’s, who was also one of the first investors in Google. In 1997 they agreed to create a new company. They called it GigaPixel, and very quickly found more investors and raised millions of dollars. Haber’s business model was based on what he called “Haber’s Law,” which stated that “If it can be done in software, it will.” He believed that, given enough horsepower on your system, you could do anything in software without having to build special-purpose hardware for the same purpose. And the unofficial corollary to Haber’s Law was that if you could build it in software, you could license it instead of selling a product. This was the model he has used successfully at CompCore, and it was the model he envisioned for GigaPixel.
The GigaPixel technology was based on the idea of tiling. “Tiling was basically dividing the screen into rectangular tiles and rendering only the visible objects inside those tiles, and then you have enough memory and cache to bring all the textures and do the best rendering you can. After that, each and every tile is rendered separately. You reassemble the image like a puzzle, and it looks perfect. It requires a lot less memory; it’s a lot more powerful.” In many ways, GigaPixel’s tiling technology was very similar to what Talisman did (see Game of X v.2 pages 357 and 358), but it was independently conceived, designed, and developed and turned into a real product, whereas Talisman was never officially released and, more significantly, it would have required an ancillary graphics card.
Eventually Haber came up with the idea of offering his technology as an embedded system in computer chips, and he made his first deal with a graphics chip maker called NeoMagic. He was also in discussions with Intel and others at the time. When he heard about PlayStation, he decided to offer his technology to Sony. “I went to one of these conferences where Kutaragi-san was then the head of the PlayStation, gave a speech. I more or less followed Kutaragi-san to everywhere he went, including the bathroom, to show him what we had, and in the end he yielded and came to one of the side rooms. We put up a small demo and we showed him GigaPixel technology.” Kutaragi said that the technology was interesting, but it was too late to incorporate in their next project. He also said something about other companies possibly jumping on the bandwagon in response to Sony’s success with PlayStation. “And that kind of gave me the hint that he knows that others are working on graphics technology, as well. So we started poking around. I called up friends at Microsoft and Intel, asking them who can be the big player, and very quickly it became clear to us that it is Microsoft, and that they are working with Intel and also working with AMD.”
The next logical step was to go up to Microsoft and offer to do some presentations. “Microsoft really, really loved this concept of tiling and saving power, saving memory, making something that can be both hardware and software implemented, and they signed a deal with us, gave us some up-front payment and said, ‘You guys are it. You’re going to be part of the Xbox team, and they even moved us into the Microsoft building in Mt. View.”
This sounds promising and simple. As it turned it, it was anything but.
Deals and Manipulations
In addition to the WebTV group, several chip makers also mentioned GigaPixel, and with NVidia’s Jen-Hsun continuing to drive a hard bargain, interest in an alternative option grew. McBreen admits that they were bargaining hard with NVidia, too. “We were asking for things like, in addition to just buying chips from them, we needed to be able to say, ‘What if we wanted to start our own fab <fabrication facility>’ That we had the right to build our own chips and just pay them a royalty. Which is crazy, right? There’s no way we’re going to build our own fab, but the guys down in San Jose, the WebTV guys, were thinking that their hardware was going to take over the world, and that the next step was that we were going to get into the chip business. And we told Jen-Hsun that we could just pay him a royalty. So basically he had to transfer all of his IP to us, and GigaPixel was obviously willing to do all that because they had never built a chip before. And so when Jen-Hsun was being stubborn, we went and signed a deal with GigaPixel. We gave them five million dollars. And what that did was really bring Jen-Hsun to the bargaining table.”
The GigaPixel deal forced Jen-Hsun’s hand. He had already leaked information linking NVidia with Microsoft’s new project, which had an immediate impact on NVidia’s stock value. Also, word had gotten around Silicon Valley that Microsoft had rejected NVidia in favor of the upstart GigaPixel. “Why would Microsoft, that wants to make the best gaming platform, walk away from NVidia. What’s wrong with NVidia? So when we signed the deal with GigaPixel, Jen-Hsun, as you can imagine, went ballistic. I remember this perfectly. It was a Saturday morning and my family was going to Sun Peaks for vacation, and almost the whole way there—my wife was driving—Jen-Hsun and I were talking about what we were going to have to do. He was concerned about his stock price, and so I told him, ‘Real simple. Go buy GigaPixel. George will sell you the company. If you think the bottom is going to fall out on NVidia because Microsoft is going to do the deal with GigaPixel, go buy GigaPixel.’ And for about two weeks all I told him was, ‘Buy GigaPixel. Buy GigaPixel. Buy GigaPixel.’ And he kind of went through the motions on that, but nothing seemed to be happening, “so I sent him a message. ‘OK, Jen-Hsun, we’re done dancing. We’re getting ready to make the announcement here soon.’ And then he came sent back, ‘Send me a deal sheet. Send me a sheet that says all the things that I’d have to meet to get back in the game.’”
So they sent a deal sheet that stipulated everything they had previously demanded, as well as the right to extend their price of chips to OEMs, thinking at the time that other companies might put out entertainment boxes with built-in Xboxes—thinking of Xbox as a platform.
After both Thompson and McBreen had signed off on the contract, they informed Bill Gates that the deal had been signed. Later, while they were both coaching a Little League game, Thompson got a phone call from Gates saying, “All right. Go with NVidia if they’ll really match the terms.” The next morning Thompson and McBreen were on a plane. “We went down to NVidia, and we took the GigaPixel contract and everywhere where it said GigaPixel, we crossed it out and wrote NVidia. We were down there all day. We went through it. We signed it. We initialed it. And then it was done. No one knew about it, but we got everything that we wanted from it.”
Jen-Hsun later initiated two lawsuits against Microsoft. One suit claimed that the contracts included sales forecasts, and that he wasn’t obligated to meet the original pricing past those forecasted numbers. Then he tried to sue, saying that the original contract was for 15 million units and that after that Microsoft would have to renegotiate. These suits came later, however. At the time, the deal was done, and NVidia would power Xbox graphics.
So what about GigaPixel? Haber says they had a deal with five million up front and a fifteen million dollar investment in the company. They had come up to Microsoft and demonstrated their technology to several major 3rd party developers, including Square Soft, and everybody had been excited about using it in their games. And meanwhile, GDC and the public reveal of Xbox was just around the corner. They had given embargoed interviews to several news organizations. There were discussions with bankers to take the company public with a valuation of two billion dollars.
Expecting to do an IPO, Haber began discussions with S3, a high-volume, mid-to low-range chip producer, and 3Dfx, who were in the high range, but sold in low volume. So his pitch was that, with his $2 billion IPO, he would buy both companies for somewhere between $400 and $500 million. This was Haber’s grand plan. “I actually get the world’s best brand, which is 3Dfx with the best known and most dedicated developers group, and the world’s most efficient high-volume 3D graphics engine provider, S3. And with our technology that is cheaper than what S3 had, and the significantly better performance than what 3Dfx had, we could take over the whole market.” In his mind, NVidia would stay in the business, as well as ATI and other graphics card companies, “but I didn’t see them having any chance of winning in the long run because we would have covered the high end and we would have had covered the low end, and come in with a technology that was cheaper and better than both.”
Haber saw how it all was supposed to happen. “So the scenario was that Bill Gates talks about Microsoft’s intention to become a leading player in gaming, and he introduces this unknown company that has the world’s best 3D technology, because Microsoft always brings the world’s best technology to consumers… that was the story. I go and shake his hand, and then a demo starts about our technology, I disappear and Bill continues his story. So it’s all buttoned down. The script exists. The story exists. The PR agencies are in line.”
And then he gets a call. Haber remembers that Robbie Bach called personally, but McBreen says categorically that it was not Bach. “On the day-to-day stuff, Robbie didn’t really pay a lot of attention. He probably didn’t even remember the term GigaPixel. There’s no way Robbie was there.” According to McBreen, once the NVidia contract was signed and official, he went directly over to GigaPixel to deliver Microsoft’s decision. “It was a very quick, ‘We decided to go in a different direction’ type meeting. Not a lot of details, other than hey, we decided to go in a different direction. There was a little bit of him trying to understand it. We weren’t really in a position to disclose a lot of stuff.”
Haber remembers that he was told the reason for the “different direction” was due to a lack of confidence in their ability to deliver the chips on time. There was some story about the chip maker, Flextronics, needing the components a couple of months earlier than originally thought. Stuff like that. McBreen doesn’t remember precisely what was said, but says, “That’s probably the party line that we gave him. Something like, ‘We have an extremely tight schedule. If we miss the schedule we’re not going to have credibility with the game developers that we’re so highly dependent on.” But McBreen says the real urgency was that they had already begun leaking the NVidia story to their game developers because the NVidia name gave them confidence and didn’t require them to adapt to any new technologies.
Years later, you would think that Haber would be bitter, but as an entrepreneur, he’s philosophical about it. “You can imagine I see the $2 billion disappearing. Today, after all these years of being an executive at running companies, I can tell you that I do understand them. From their perspective, like with any team, the weakest link is the one that breaks the chain. So they invested the billion dollars in Xbox on the software side, on the hardware side, on the PR, on supporting game developers, and they perceived us as potentially being the risky part, and if there is no graphics engine ready, then the whole launch is useless. How can you have an Xbox that doesn’t have a graphics engine? And based on timelines, they felt that it was just too risky.” Remember, his personal perspective on life was: “The only thing that matters in life is the thing that you can carry with you wherever you go, which is basically your brain and your education, and the persona you become by learning.”