INTERVIEW: Unlocking PS3′s Cell Power
Unlocking the power of the PlayStation 3’s Cell processor is a difficult task, but how much so? High Moon Studios’ (Darkwatch) technical boss Clinton Keith tells Next-Gen how much the Cell has to offer and how his studio plans to bust the processor wide open.
Late last week, Vivendi Games and its game studios wrapped up a workshop with Cell microprocessor co-creator IBM. The point of the “Cell Summit” was to help Vivendi-owned studios such as High Moon get to know the PlayStation 3’s complicated CPU.
In the midst of that workshop, Next-Gen caught up with High Moon chief technical officer Clinton Keith, who spearheaded the two-day gathering. Here, he talks about the complicated nature of PS3 development, when developers will fully realize the PS3’s power, the trade-off between ease-of-development and sheer power and working with IBM further down the line.
Next-Gen: Now you describe [the PS3] as “radically different," but can you elaborate on that more? Just how radical is the PlayStation 3 compared to even its contemporaries like Xbox 360 [note: High Moon is also working on an Xbox 360 title] or other systems?
Keith: In comparing it directly to the Xbox 360, you know the Xbox 360 has three general purpose processors in it. But they’re more like the typical processors that you might see in a PC or Macintosh… With the big general purpose processors, we can write the software traditionally the way we’ve done it in the past, so we don’t have to change things so much.
What IBM did with the Cell processor is it really embeds about seven processors, one of those being the general purpose core and the other six being these real dedicated specific-use type of processors that are extremely fast. But seeing that they’re not general purpose, they’re a little bit more challenging for programmers to get under control and to write software for.
With these Cell processors and these small processors called the SPEs, we really have to not only write software different but we have to think about how we’re solving problems in a completely different light.
Once those complexities are unwound, how dramatically will the PS3 development environment change?
This is what we’re looking forward to in years down the road, as Sony said they want this machine to be around for the remainder of the decade… Right now, the games you’re seeing come out are using engines that are more in the traditional way of creating games, which is that your engine architecture has access to all the other parts of the engine itself. With the PlayStation 3, we’re going to have to figure out how to divide up these things up so that they’re much more separate.
[We’ll have to] explore things such as "procedural synthesis," which really has exciting potential on the PS3. Rather than creating all these environments and all these behaviors by hand, now we’ve got a lot of this power, [so] we can come up with ways that the processor can create environments, and create artificial intelligence rules that kind of emerge with gameplay and adjust to the gamers’ input, so we can have a lot more variety. That could interest somebody with the concerns of the rising cost of development.
When will developers be able to fully realize the PlayStation 3’s power?
That’s something that we’re trying to discover right now. I think that there are games out there that no one’s ever seen before. I call them sandbox games where—take one of my favorite games, which is the Battlefield series—where you get to play with dozens of people online in a large environment with lots of explosives. The thought I have is that every time you play those levels, those levels are the same. They stay the same and they never change.
What I’d love to do is I’d love to play in an environment that changes over time, that if there’s a building where the snipers are hiding in, you can make a big hole in that building and it stays that way for awhile. To do that believably without creating a ton of assets, we’re going to have to mimic real life and real physics. I think that’s the potential for what the Cell processor can do. It can crunch a huge number of calculations if those environments are built correctly [and] if we figure out how these SPEs work. I think [we’re] in the generation to start figuring that stuff out, so we’re trying to bootstrap that and trying to experiment with those things and see what’s possible.
Will it be something like two or three, four more years before we see just truly mind-blowing games and results from the PS3?
I think somebody might surprise us with a few things here and there. We’ve actually got a prototype on the PS3 that simulates liquid like no one’s seen before and we’ve actually built a little minigame around that to take advantage of that… The goal of this small game was that [we] might be able to put out a small downloadable game that somebody might buy for five dollars, play it on their 1080p television set with their PlayStation 3 and just really show people an experience that no other console can give them on their $3000 TV set.
As a developer, would you rather be working on an extremely complicated system with lots of power to unlock or an easier-to-develop-for system with a little bit less power?
It would depend on whether I have a schedule or not that I’m trying to keep. This kind of reflects what we’re doing right now [with the Cell workshop]. We’re focusing on an R&D effort, a heavy R&D effort on the PS3. The focus of what we’re doing this week is to have a small team—called Beachhead team—really kind of explore and make small games that can’t possibly be made on any other platform. They don’t have a schedule and basically they’re kind of discovering as they go along.
Now if I have a title [due out] in 18 months, we’ll have to basically narrow what we’re creating within six months. Then obviously I want tools and to know [those tools] right away.
So what exactly happens after this workshop’s done and how will the relationship between High Moon, IBM and Vivendi continue after this?
I think that we’re exploring continuing this on a more one-on-one basis. Right now, they’re coming down here and they’re working with about thirty engineers across Vivendi in a workshop and giving [Vivendi] hands-on experience with some of their tools that they’ve developed. Going further, I think that the work we do could benefit both IBM and us in looking at specific problems that we’re trying to solve like procedural synthesis… It’d benefit IBM to see the actual application challenges of the Cell processor.
In addition to that, we want to look at utilizing some of these blade servers, which are arrays of Cell processors. We’re doing things like right now, if we had a huge environment and we wanted to pre-light it, it can take half a day for a PC to cook that lighting to the level. The promise of these blades is that we can do that in a few minutes. We always say that the quality of whatever you’re working on is based on how many times you can iterate on it, so we’ll certainly come up with far better looking levels if we can rapidly iterate on things like that.