IBM, AMD, Nvidia, Intel Talk The Future Of Gaming Processors
Last Thursday in San Francisco, Cadence -- a company not well-known to anyone but chip makers, but extremely crucial to them for the software and hardware solutions it provides for the design of chips -- held a roundtable discussion with high-level staff responsible for some of the most crucial processors in the gaming space.
Attendees included AMD's vice president of engineering, Robert Feldstein, Cadence Design Systems' CTO Ted Vucurevich, IBM fellow and director of processor technology James A. Kahle, Intel director of gaming program office Randy Stude, and NVIDIA vice president of graphics processing unit engineering Jonah M. Alben.
The discussion was launched by moderator, technology consultant Jon Peddie, with a rumination on Moore's Law, which is defined by Wikipedia as the the truism that "the number of transistors that can be inexpensively placed on an integrated circuit is increasing exponentially, doubling approximately every two years."
According to Peddie, "We are outstripping our engineering resources and we cannot produce all the things we are theoretically capable of producing, because of Moore's Law." Peddie suggested that software solutions could help with this problem -- and reduce engineering costs. This is, of course, a premise that plays to Cadence's strengths.
"The Great Thing About Game Console Chips"
According to AMD's Feldstein, who led the design for the Xbox 360 GPU, "The great thing about game console chips is that you're not bound by the linear path the PC has to take -- the dictators who say where the PC is going, like Microsoft." Feldstein even said Microsoft behaves differently when it comes to the 360. "You're also looking at a longer time. You're looking at something that has to live for years rather than months. It's a closed environment, you don't have the open factors of PC. It's a great place to prototype and design."
PS3 GPU supplier NVIDIA's Jonah Alban delved further into the issue of Moore's Law. "I think there's two factors. First, Moore's Law giving us the ability to create products with more and more transistors. The second is that the graphics area is an area that has an insatiable demand for those transistors. Moore's Law is your friend or your foe. Graphics has been a great combination of this strong desire for computing power and Moore's Law."
However, Feldstein dissented a bit. "It's not clear to me that game developers want us to go that fast. For some, the continual statement is 'Hey, I need stability, I need to know that this platform is going to stay.' I'm not sure that [our effort] keeps the average person playing games in mind."
Setting the tone for his comments throughout the session, Intel's Randy Stude, whose company has no current console products since the cease in production of the original Xbox, argued that consoles do not help with innovation. "Throw Moore's Law out of the window when you talk about consoles -- except every five or ten years when one comes out. Moore's Law applies to the PC industry."
However, IBM's Jim Kahle, chief architect of the Cell project which resulted in the processor for the PS3, disagreed totally. "Moore's Law is about density, and density applies to cost. On consoles you're trying to get the cost down."
Discussion of the Cell led to a snipe at the PS3's power -- specifically that Activision and Infinity Ward's upcoming Call of Duty
4 looks no better on the PS3 than the 360 despite the Cell's supposed performance enhancement over the 360. Kahle responded, saying "I think we've got to get back to where is this edge of graphics going to be. I'm not sure that pushing the graphics all the way to the extreme edge is where people are going to go. It's going to be the simulation part, the AI.""There are many characteristics going into the game," Kahle continued, adding "you haven't seen the full power of the PlayStation [3] exploited yet. That's what we worked on... we still have a ways to go." Referring to the developers gradually becoming more proficient with the system, he said, "I'm not sure we're necessarily on a different track than we were in the past."
Integrating Systems
The discussion moved to the possibility of integrating several systems into one chip, with the end of discrete CPU and GPUs posited by the panel. Kahle said, "There's a scalability problem going on -- the [space] around the transistors is getting to 10 atoms thick."
To solve this problem, "We have to advance a lot of techniques on power, how to reduce power, be more efficient. Specialized hardware is more efficient. We have more ability to put more memory on board. IBM's working on embedded DRAM. There are things in the future that will help those limitations."
Cadence's Vucurevich carried discussion of integrated chips further. "With that density comes the question, 'What am I integrating together into that visible area? Am I putting just more pixel processors and memory?' You're beginning to see systems being designed together. [If] I can start to bring more functionality into a single chip I will see from, an overall systems dynamic perspective, a power savings and [reduced] network latency."
When it comes to one of the most crucial issues for consoles, cost reduction, Vucurevich said, "I have to have some picture of how I'm going to reduce that cost over the four or five year cycle of that console being on the marketplace."
AMD's Feldstein -- whose company has no current inroads into this market -- honestly admitted, "When we're talking about game consoles, you have to remember the handheld game consoles are maybe the most significant out there -- they've seen problems like this already, power, how you get the cost down by reducing IO."
"Everything in the world in the next ten years may not want to be plugged into a wall -- it might be light or portable," he continued, "they are crossing a lot of the barriers first about how to live in the environments that are bounding all of us."
Alban noted that, "On the fab side it's really turned around... if you go from 90 to 65 to 45 [nanometers]," but Stude countered with, "If we can save wattage on the processor it's quickly consumed by the graphics processor."
Do Consoles Drive Innovation?
Peddie then posed the question that Stude argued earlier -- can consoles be considered a place for innovation in processing? Of course, IBM's Kahle argued in favor, with the PS3's Cell as the example. "I think in the future it will continue to drive us. It's not just that games are standing by themselves; there's a lot of synergy with other activites. [Processors] have to be good at physics. I think there are a lot of synergistic things [that] will drive us."
Feldstein said, "360 is where we got our 'go forward' architecture. You can make interesting tradeoffs in these convergence devices. If you don't oversize your system you have a system that attacks the problem -- especially in the game consoles, more entertainment."
Alban ruminated that the PC is the backbone to the whole enterprise. "It's important to not forget the PC side of the business to innovation. The PC provides the revenue stream every year for the team. If it was only a console business, we [AMD and NVIDIA] would not be here." With consoles, "every few years you can reach out and do an interesting thing, but if you didn't have development every six months on the PC, you couldn't sustain your team."
Feldstein suggests that since game development has made a major shift to consoles, so too does chip innovation. "If you talk to [software] developers today, their prime target is the console first. This is almost universal. Then they spread it out to the PC. I don't think it was this way five years ago. This year they're starting to go to the consoles first, because it's a more consistent place to play games."
"I think the more important distinction is between parallel versus serial tasks. We started out bringing out these parallel processors for game graphics... it turns out there are a lot of parallel tasks in the world, and that's very exciting," Alban offered.
Kahle agreed. "We're moving more general purpose," he said, which makes sense as the Cell processor has to be utilized by IBM and Toshiba as well as Sony (with Sony using it for more products than the PS3 as well.)
Feldstein talked practicals. "Cost downs are a real pain if you were Intel or AMD doing new processors, because you can't change anything." As processors shrink, "significant parts change. If you're doing an Xbox or a Sony, everything has to be consistent, the last game has to [run] the same as the first game that was bought. NVIDIA and AMD, we're so hell-bent to do the performance and the density and you have to restrain yourself because things have to stay the same." However, when questioned about the AMD GPU's role in the Xbox 360's billion- dollar unreliability issues, Feldstein demurred comment.
Regarding the Cell, Kahle said, "When we first started we didn't want to create another mild increase in architecture stuff -- Sony pushed us to the edge. We started in 2000 and we had to guess what games were going to be like in 2007. We had to keep it fairly flexible."
Relevant aspects of the design included input, output, and physics simulation, as well as speed. "We're being more innovative, we're not just riding the frequency curve anymore. We're putting new architecture in."
IBM is using the Cell for medical imaging applications: "We have some great results with the Mayo Clinic." When it comes to multipurpose, collaborative chip design like the Cell, "It was about putting the right things in there and the right collaboration."
Stude laid out the Intel war plan. "One task that we're undertaking right now is to address how we compete more effectively against consoles -- and how to make sure that in mature markets, when someone buys their desktop or notebook computer, that they're not seeing it as a competitor to a console, that they see it as the best..." solution for a variety of computing needs, including gaming.
According to Stude, "We did an evaluation of consumer satisfaction with PC as a gaming platform -- the feedback came back pretty bad. Customers are taking 53% of [publishers'] support calls on the PC where only 16% of where their sales are coming from. In the last two years, the top 50 selling PC games had 42 different [sets of] minimum system requirements."
But what does innovation in the processor space mean to consumers? Peddie rightly pointed out that most people don't care about these innovations in anything beyond a practical way: what the box can do.
Kahle put it like this: "I think we're in the early stages of what I call immersive computing. We're going to get closer to the computer -- the way people interact with the computer is going to improve. I think gaming is a good playground for how we interact with computers... I think the Wii is a good example. Maybe there's a better way to interact than a keyboard and mouse."
In Kahle's estimation, we can output many more types of data to computers than they can usefully input. "I think we're in the early stages how this is going to transform how people are interacting."
Driving Towards The Future
Vucurevich talked about a "plague" that hit World of Warcraft, extrapolating that "We're on the verge of some very profound changes on what we do and how we perceive technology in general. The gaming industry is very interesting because I look at it as a virtualization of reality. There are things the designers never thought of as part of the experience that begin to be possible."
The drive for realism in game graphics was also touched on. In response to Alban touching on the uncanny valley, Feldstein said, "We're still a long way from where the movie [CG] stuff is... but the paradigm for graphics shifted not that long ago, with shaders... I think you'll see hundreds of constructions per pixel to bring out that realism. What people want to do is to go beyond realism, to bring you into a state like Blade Runner, which isn't quite realism but reminds you of realism."
When asked if the PS3's current market performance will affect the spread of the Cell processor, Kahle simply stated, "I think the companies are still dedicated to the architecture and seeing other places we can take it. It's just the first year, and I'm not concerned." Peddie alluded to some new Toshiba devices, soon to be unveiled, that will also take advantage of it.
In Summation
At this point, Peddie asked each panelist to summarize. Kahle began. "I think we're still in the early phases of learning where we're going with some of this technology. From a collaboration point of view with gaming, we'll push it in different directions. As the price of the [PS3] console goes down, some of this is going to shift. I think [gaming applications are] going to drive innovation starting with game consoles, and then move to computers and affect how people do business and transactions. I think it'll be the driving force in technology."
Feldstein expounded further on the collaboration issue. "I actually think it's going to be about collaboration. We're talking about game consoles now. We're talking about unifying experiences that are tactile, visual, hearing. It's about collaborating in the ecosystem to bring that all together." He left us with the question, "What are the system tradeoffs to bring that all together, to bring a device people want in their living room?"
Stude said, "Innovation starts on the PC. The PC is the most collaborative device in the world. I think online gaming will totally hit the consoles from left field and completely erode their market to the place that they can't compete in the next wave."
"I think they are going to suffer the same fate as the music and movie industries if they think they can continue those models," he continued. "Steam has evolved into a scenario to where Valve is the self-proclaimed largest distributor of PC games in the world. The digital fulfillment engines that are out there are going to challenge the traditional notions."
Alban decided not to continue this tack, instead talking about innovation in processing. "This is just an extraordinarily exciting time. We're just at the beginning at tapping the capabilities that can come out of [parallel processing] not just in games. Thinking about what other things people can do with processors, that has nothing to do with games. It just creates an opportunity for people all around the world."
Vucurevich summed up Cadence's role to the four preceding speakers. "We help them to make sure that they can make the best possible designs they can make in the shortest period of time and make darn sure they work when they get out into the marketplace."