Processing The Truth: An Interview With David Shippy

A recent, sensational Wall Street Journal article about a book called 'The Race For A New Game Machine' raised a number of new questions about IBM's co-creation of the Cell chip for the PlayStation 3 -- near-simultaneous to its creation of the PowerPC Xenon CPU used in the Xbox 360.

The co-authors of 'The Race For A New Game Machine', David Shippy and Mickie Phipps, were two leading figures in the design of the Cell at the Sony-Toshiba-IBM design center, which jointly spent an estimated $400 million to develop the technology.

Shippy was the chief architect of the power processing unit for the Cell, and overall technical leader and architect for the team that created the Power Architecture-related microprocessors that ended up in both the Xbox 360 and the PlayStation 3.

The idea that the Xbox 360's processor shared some of the same design lessons as the PlayStation 3's Cell -- and the apparent claim that Microsoft's Xenon CPU might have taken any of IBM, Sony and Toshiba's PS3 learnings and applied them to the Xbox 360 -- has been controversial. So Gamasutra decided to talk to David Shippy, co-author of the book in question, to find out more details from his perspective.

Although Shippy was a distinguished IBM engineer for many years, his story in relation to the Cell chip began in 2000 when Sony first contacted IBM asking for an "order of magnitude increase in processing performance" to use in their next game machine. IBM demonstrated for the company all of its in-house tech, and other chips Shippy himself had worked on -- "and Sony said, 'That's really not what we're looking for,'" he recalls.

"They said, 'We want you to start from scratch, throw out everything you've ever done before, and come up with this groundbreaking new architecture with more performance than anything out there, any game machine, any PC,'" Shippy says. "Ken Kutaragi had this bold vision -- 'Don't come to me with anything you've got; go and invent something new.'"

The goal, Shippy says, was to create a "supercomputer on a chip." At the time, most PCs' processing power peaked at about 1.5 gigahertz, and dual cores were still a relatively new idea. The Cell, on the other hand, featured eight parallel processing engines. "We came along with Cell and we were double the frequency of any PC out there," Shippy says.

The larger feat, however, was to provide that level of processing power on as little an energy budget as possible. "Game machines can't really afford a huge fan and a heatsink," Shippy says -- so the plan was to create a high-performance chip with a smaller footprint.

The Cell project kicked off with a full team in March 2001. The Sony-Toshiba-IBM Design Center was established in Austin to build the chip, and for some two years, Shippy said the team was "heads down" designing the next-gen CPU chip for PlayStation 3.

According to Shippy: "And then, along comes Microsoft in 2003, and says, 'Hey, IBM -- can you design a chip for us?" Just like its rival, Microsoft was not easily impressed by IBM's existing offerings.

But the company did take an interest in IBM's PowerPC chipset, the latest versions of technology first implemented in 1992 by the firm, and previously used in simpler form by Nintendo for the Gamecube. Where Microsoft differed, says Shippy, was that it wanted multiple cores on the same chip.

"But what caught their eye, I would say, was the R&D effort that we had put into the Cell technology," he says, apparently referencing some of the high-level PowerPC advances made to help build elements such as the power processing unit of the Cell, which he was lead architect on. (The Cell architecture consists of a central PPU, and a set of parallel processing SPUs.)

Shippy doesn't believe that Microsoft yet knew that Sony had the PlayStation 3 in the works -- but liked what it saw in the PowerPC technology that was now possible thanks to design principles partly researched for Cell. "The initial tech that we built -- yes, it was paid for through the Sony-Toshiba-IBM Design Center, and was developed for the Cell chip," says Shippy. 

"All three companies… legally all had rights to go and put any of that technology, any of those processor cores into other spaces. All of them talked about doing that… so it was every bit IBM's right to sell any of that tech to other design spaces," Shippy explains. "It is very common to develop an interesting, leading-edge new technology and then utilize that technology across multiple platforms."

"I guess what everyone didn't anticipate was -- before we even got done with the Cell chip and PS3 product -- we weren't anticipating that we'd be showing this off specifically to a competitor."

Does that mean Microsoft got a look at the Cell itself? "No, we didn't show them the Cell chip," Shippy clarifies. "The Cell itself and the fundamental architecture that went into that, actually not -- that was all proprietary for PS3. What was shown to Microsoft was just a technology road map that said, 'hey, we can go do these high-performance PowerPC cores at very high frequency and low power'."

"In no way did IBM say anything about the fundamental architecture of the Cell chip; it was more about introducing Microsoft to the circuit design technology that enabled us to create these really fast processors."

So despite some higher-level conceptual ideas in common, Shippy stresses that both consoles' processors are very different, from architecture to software models. "They differentiated themselves in their own unique ways," says Shippy. "What's interesting is that they did that with this common building block that was designed initially for the PS3."

Shippy says things got more interesting once the two companies became aware of one another -- the Xbox 360's Xenon CPU in development solely by IBM, and the PS3's Cell chip by IBM, Sony and Toshiba.

"It really sort of changed the whole dynamic of the design center," Shippy recalls. "I had Sony and Toshiba engineers working side by side on this project as part of my team, and when this whole Xbox thing came in, there were certainly a lot of guys -- even on my team -- that were fairly upset about it."

He added: "The initial reaction was it sort of felt like a betrayal -- 'Hey, we've been designing this really cool Cell chip, and now you want us to do this Xbox 360 thing on the same timeframe? Seems like we're aiding a competitor here.'  At that time, only the IBM engineers knew about it. They struggled with betraying their partners -- and that was my initial reaction, too."

But with two major products now on the to-do list, Shippy says there was no time for moping. "I had to get over it pretty quick. While I was upset about it, I had to put on my PS3 hat or my Xbox 360 hat and basically just tell my team, 'Hey guys, this is no different than when Intel and AMD create a microprocessor and it goes into a Dell PC and an HP PC. Just get over it, and let's move on."

Once news leaked out that IBM had been awarded the next-gen Xbox game chip, Sony and Toshiba began to speculate, Shippy said. "That led to some distress there from our partners, because they weren't really sure what was going on. But they started putting all the pieces together, and saying, 'Wow, we're actually now designing this superprocessor core not just for PS3, but for our competitors.'"

"At the end of the day it didn't matter," he says. "At the end of the day, I had these awesome engineers and they just wanted to create the best leading-edge technology that they could, and it didn't matter whose box it was going into."

The broad commonality, then, between the two consoles is the significant amount of parallel processing in the CPU enabling high power at low energy budgets -- compensating for the minimal space game hardware offers for cooling measures like large fans and heatsinks.

But whereas the PlayStation 3 seems to manage its temperature issues well, the launch of the Xbox 360 as a whole was plagued in its early days by serious heat problems that caused widespread console failure -- though this was not necessarily a CPU-specific problem. But why might that have happened?

"I think, like any company, they stretched the limit on the power budget," Shippy suggests. "They probably wanted to get absolutely as much CPU chip performance in this product as well as GPU, and cram as much as they could in there, and I think they were just a little too aggressive on how much they could package in there. Like any company, they probably took it right to the edge -- so you're playing with fire a little bit, there."

"I wasn't that involved with some of the console decisions and what the overall budget was, so I think from a system standpoint, they did their best due diligence to say, 'We believe we can build this and meet… the power budget.'"

But can Shippy's insight on both console's processors finally answer the age-old debate about which console is actually more powerful?

"I'm going to have to answer with an 'it depends,'" laughs Shippy, after a pause. "Again, they're completely different models. So in the PS3, you've got this Cell chip which has massive parallel processing power, the PowerPC core, multiple SPU cores… it's got a GPU that is, in the model here, processing more in the Cell chip and less in the GPU. So that's one processing paradigm -- a heterogeneous paradigm."

"With the Xbox 360, you've got more of a traditional multi-core system, and you've got three PowerPC cores, each of them having dual threads -- so you've got six threads running there, at least in the CPU. Six threads in Xbox 360, and eight or nine threads in the PS3 -- but then you've got to factor in the GPU," Shippy explains. "The GPU is highly sophisticated in the Xbox 360."

He concludes: "At the end of the day, when you put them all together, depending on the software, I think they're pretty equal, even though they're completely different processing models."

What about the familiar refrain that the PS3's architecture is overcomplicated, challenging developers to program effectively? "The Cell architecture, from a software programming standpoint, is definitely a new paradigm," Shippy concedes.

"And I think what the game community would argue is that, since it is different, initially it is harder to program the Cell chip -- it's not a traditional multiprocessor environment," he says.

"The real hardcore coders would argue that, once you do understand it and can program to it, you absolutely get the most out of the hardware, and really write some fairly low-level code that's really high performance," he adds.

"I think some of the bigger game houses that will write more high-level code would really prefer an Xbox 360 -- right out of the chute, it's easier to write code for. I think you can really leverage the Cell hardware technology -- but it is harder to get your head around."

Shippy's recounting of his role in the console wars has drawn some skepticism, with discussion taking place across the web suggesting the book has over-sensationalized the idea that "Sony funded the Xbox 360."

But he maintains he never intended sensationalism. "The Wall Street Journal article came out, and then all over the internet, there were these debates," he says. "I read a few of them -- and none of these people had even read the book yet. So everyone was speculating on it, and what happened and what was going on -- the reality of it is that isn't why we wrote the book."

"If you read the book, it's just an interesting story about a high-performance design team that delivered an incredible product on an incredible schedule, and the leadership skills that went into that. It's really a pretty uplifting story," he says.

"It really wasn't meant to be sensationalized as 'Sony funded the Xbox [360]'. That was never the intent of the story. It's an interesting angle that that's how it sort of played out, but the story stands on its own without any sensationalism."