Gabe Newell: next-gen game engines will be ten times harder
I was at Valve last month to interview pretty much everyone I could find, and play one of the most exciting PC games on the horizon: Portal 2. The preview I wrote, and the profile on Valve themselves, is in the new issue of PC Gamer in the UK. But we're also putting up the interviews here on the site, one a day for a week. Today's is my conversation with Gabe Newell, Erik Johnson and Doug Lombardi about the difficult but exciting future of game engines, and why they hire who they hire.
Gabe doesn't spare me the technical terms, so if you're allergic to jargon skip to "So what's the interesting thing from a gamer's point of view?"
PC Gamer: You've mentioned one of the things you're thinking about is changes on the hardware front - what's changing?
Gabe Newell: I think we’re going to see a lot of features in CPUs that makes them work more like GPUs, in terms of throughput architectures. So the specific instructions are going to look a lot more like GPUs, and depending on how they get there - there’ll be a couple of different approaches that the different CPU companies use to get there – but essentially they’re trying to provide the what, in an architecture independent way, is called throughput architectures. So you’ll see each different company coming up with a different answer that they’re essentially solving that problem.
PC Gamer: I remember hearing a similar sort of thing a few years back where the cores on a graphics card were getting more like a CPU, and we’d be able to use those cores for things like AI and stuff like that. Is that still happening?
Gabe Newell: Well, you can kind of go both ways. You can make GPUs more general purpose or you can increase the parallelism of a CPU. And it’s a lot easier to do this than it is to do this. So... I think that I’d like to do it, but it’s harder to get people to write code and instructions for that. If you’re not Intel, if you’re not ARM.
PC Gamer: So what’s the interesting thing from a gamer's point of view, what practical impact might that have on your game?
Gabe Newell: I think we’ll continue to see accelerated performance, even faster than we’ve seen before, that’s applicable to a wider range of functions that you perform. Right now it’s easy to do scalable graphics, and I think that it’ll become similarly easy to do scalable AI, scalable physics and the top end will stretch a long ways. So in the same way that there’s a huge difference between embedded graphics and a high end GPU, game engines are going to have to accommodate a much wider range of potential performance for AI and physics. But the top end is going be awesome in terms of just the sheer amount of computation that we’re going to do.
The other thing that it’s going to do is, the best programmers are going to become even more valuable. This will affect everybody in the industry so all the programmers, anyone associated with programming. The amount of difference between a good programmer and a great programmer is going to get wider, in terms of the amount of value that you can create.
So that’s a function of the fact that it’s just hard to write the architecture that allows you to take advantage of these hardware approaches. You have to sort of redesign all of the code so that not only can you create batches of work to do, but you also have to make sure that it’s relatively bulletproof. So somebody can’t come along and accidentally break the entire system by not understanding what the trade-offs are. You don’t want somebody who’s sitting there making a rig for animation to accidentally blow through some constraint, so that by adding two more creatures on the screen your framerate drops by a factor of, say 50.
You have to create an architecture that not only allows you to take advantage of this kind of hardware, but you also have to do it in such a way that it’s resilient in the face of people writing leaf code (not core engine code) and people creating assets for that pipeline. So that’s just a hard problem. If there were 500 people who could write a good game engine in the last generation, you’re really talking 50 people who are going to be good enough to do it in the next generation. Which is good for those people.
PC Gamer: Do you have those people?
Gabe Newell: Well, we think we do, yeah. But a good programmer could create an engine to take one of these architectures, and a great programmer could come in and do the same kinds of things and do them ten times faster on the same kind of hardware. In other words, it's a non-linear return. And you just don’t see that today. Nobody walks in onto an Xbox or a PS3 team and can make something that runs ten times the amount of content on the screen. But in these emerging generations, that will definitely be the case.
So you know, guys like Carmack or Sweeney or the guys at Crytek can do it, but it’ll be harder and harder for other people to be able to program at that level. So that’s a side effect of it. The games will be great; I mean the opportunity is just immense for what’s going to be possible going forward.
PC Gamer: You guys seem to hire people who have good ideas on the back of the fact that they have good ideas. When you look at a bunch of indie guys, like a DigiPen team, what's the difference between a game that looks cool and has a nice idea in it, and a game that makes you want to hire everyone that works on it?
Gabe Newell: It's talking to the people, right? This is like a four hour side conversation. What it really comes down to is that you talk to the people, and you get a very clear sense, quickly, about what they can contribute, and how they're going to impact the people around them. Everybody here has a huge impact on everybody else, so when we bring in somebody like Bay Raitt (the Weta Digital engineer responsible for Gollum's facial animation in the Lord of the Rings films), or Kuda, it's weird how much impact, even at this point in our history, one person has on all of the people around them.
The decision to do that is all about the people, and not about the game. The game is something that happens to go with them. Hopefully we help them make something sooner and better than they would be able to otherwise, but it's always the people decision first. And we don't have reqs, it's not like somebody says, “Now we're going to hire three texture artists." We hire anybody who walks in the door who fits. We'll make them an offer, and we'll pursue them relentlessly. We have one guy who I think we're finally going to get to move here that we've been pursuing for twelve years now, and we finally have convinced him to join the horde. What do we call ourselves?
Doug Lombardi: Horde is not it!
Gabe Newell: Alliance? (Laughs) Red? Blu? Company?
Doug Lombardi: Studio would work.
PC Gamer: Who is that?
Gabe Newell: I can't say. I can tell you there are people out there that we would love to work with that we aren't working with yet. The guys at Media Molecule. We think those guys are awesome. There are a bunch of guys at RAD Game Tools that we think are awesome. Who else?
Erik Johnson: In the game space? There's a lot. There's people at id, there's people at Epic, there's people at DICE.
Gabe Newell: So yeah, for us it's always about [the people]. I see Erik more than I do my wife. (Looks at Erik) You're looking really nervous.
Erik Johnson: My awkward scalar just started going.
Gabe Newell: So, you know, we want people who are going to make us smarter, and make us excited to go on the move. I mean, I get to be the biggest fan boy of all. I get to see everything everybody does and all the different versions of it. I get to see it first, so it's a huge amount of fun to work here. It's exciting. Like I just reviewed 48 different box concepts for Portal 2 - 48! - and so I get to see 48, and customers only get to see 1. Actually, Valve's just a very clever way...
PC Gamer: To become the world's biggest Valve fan?
Erik Johnson: It's still a big event when we hire someone, everyone's still super excited about that. It's just like it was 10 years ago pretty much.
Gabe Newell: It's amazing how much of an impact the right person has when they come on board.
Tomorrow Valve tell us about their evil genius plans to measure how much we sweat, and make games about it. Seriously, you need to read this one.