7:27:  They want to make the game great for everyone that Quake Live is something anyone can play, and we can get people who'll never pay $60 for a game. He's interested to see how QuakeCon changes next year. Will newbies come down to Texas next year?

RAGE + DOOM!

He's still locking himself in hotel rooms and working for new platform work. Most of what he does is on stuff like id tech 5.

7:30: All the guys who are making games on the BREW and Java platforms are making the classic mistakes, with clipping problems, cracks, and other weird shit happening. John and Michael Abrash made highly-polished rasterizers that looked good. Hardware accelerators saved you from having to worry about that stuff. Only Mesa devs had that stuff. 

Had two days one time and two days another time to make the BREW rasterizer. Wow. It's sounds like really scary assembly code. 

7:32: He spent a few days looking at vision of the future. 90%+ of his time is spent working on Rage, and soon Doom. He gets to choose what he works on. Robert Duffy is the lead programmer now, John just does the engine stuff. Now  they have more programmers at id than they had employees when Robert Duffy started. John leaves the programmers alone so that he doesn't get "promoted past his point of competence".

Pulling the switch on a major change to animation, and one significant thing that needs to happen to make Megatexturing stuff working. They did some stuff that are two separate parts of the wasteland to make the video look awesome, when they're really going to be separate areas.

With Doom, they want to exploit the full potential of the current generation. They're going to have a working set of tools to do the most they can with it from teh very beginning. They'll be able to provide one strong graphical hook, but the designers should be able to do the best job they can with it from the beginning. While programming was the long pole of the tent for a long time, it's always the thing that gates the whole process. Even though there's a massive amount of data to create, programming is still a gate. This kind of goes against what he said at the GDC keynote in 05 or 04.

 7:45: He's always worried that no matter how carefully polished he's made his code, that someone trashes a pointer and breaks everything.He's talking about defect management, and people are leaving. "We're not on a quest to make perfect code, we're on a quest to make good enouhg code" 

I'm really very, very hungry. Haven't eaten since west coast breakfast. 

He seems pretty tied to C++ for AAA games, simply because the performance isn't there for the more managed languages like C#. Could id tech 5 have been something that's a stable dev tool, like Renderman? It seemed like you could do that, but if you want to be a fast, 60Hz game, you can't be flexible and Renderman esque or you can't take advantage of the incredible power that you have now in hardware. 1,000,000 times as powerful as when he started.

He doesn't think that id tech 6 and other next-gen titles (Xbox 720 or PS4) will be in Java or a protected language. 

7:50: There will be a next-gen engine that uses hardware that doesn't exist right now in id tech 6. There's Larrabee, Fusion, and a lot of other players. There are people who think the PC will come back and will make consoles obsolete, but "I think that's a naieve view" PC space is still relevant.

Nintendo kicked everyone's butt this generation. Nintendo has always been id's least friendly first-party, and it's great to see them win big. What happens wiht all these vendors on the next-generation? We don't know a huge amount more about the next steps. Honestly, it will be great if this generatoin of console lasts twice as long as the last one. He doesn't think it's going to turn out that way, and he thinks that different console manufacturers are going to jockey for early mover advantage. He hopes that Doom will come out on this generation of hardware. Power PC across the board is nice.

7:55: The stakes are really high with the console wars, and a monoculture isn't a good probably. Larabee vs. CUDA derivatives is interesting. He's concerned about massive parallelization, everything is 16 or 32 vectors wide. There will be massive amounts of stuff running on these massively parallel hardware apps that will use 1/10th of the available power. He's talking about the voxel stuff. Parallelisation is sometimes really hard. The real difference is memory scatter and gather, which lets different elements pull from different areas of memory. 

Does this mean that lots of other algorithms can't benefit from parallelization? He doesn't know, but thnks probably not. People have been working on parallelizatoin a lot for a long time, but it doesn't work that well. VErtexes nad fragments iwht graphics processing are the two best, most parallelizable.

 No one has the hardware to work on. CUDA gives Nvidia a strong lead right now, and it's influencing their directions. ATI and Apple have general purpose computing stuff, which is interesting, but he thinks Nvidia has the lead. Does Microsoft have a standard that will drive sutf? We don't know. There's no background,like there was like Direct3D and OpenGL. These are all toy research projects, and no real apps. Billion dollar bets happening with this. We're going enter a huge period of flux again, like mid-90s. Yo ucan't spend yeras on speculative architectures.

8:01 - Carmack sent a couple of days fixing a deadlock on Microsoft's Xbox 360 background video loading code. He tried it, it works, and then it hung 1 minute into a setup on loading in their code. He never figured out what the problem was, but he stopped doing one thing that just happened to fix it. You don't have the source code, and sometimes stuff just is fixed. You don't have time and mental bandwidth to evaluate everything in the codebase.

He knew everything about Quake3. He had global knowledge until the end, when he added Jean-Paul's bot code, which he knew nothing about. In Doom 3 he knew about very little of the game. That's only going to get worse, and parallelism makes everything worse and harder to make work. This is something to worry about, and we need to make architectures work better with this. He's usually an optimist, but on this he's really concerned. 

Q&A next -

One of the reasons that Rage is published by EA is that Rigatello had cogent points to make on actual gameplay when he came to Texas to look at it .No falling out with Activision, there's every chance that Doom 4 will be published by Activision. They don't sign long-term contracts though.

When you were working on Quake, when did you realize it was great? With Quake, there was a moment early on in the game's development, and he was on ledge looking down, and the shambler was walking below. He was plodding, and had weight and mass and that was one of his defining moments for Quake.

How do you manage creative types? id used to be a very Darwinian environment. There used to be owners and employees. With Doom3, there were leads (designer, artist, programmer). We might add one more thin layer. Is it because we're getting older and are more mature instead of egotistical loudmouth 20 year olds? Maybe, but no one wants to put $20M-30M into a game when people behave like children. He has distanced himself, because he's bad at dealing with people. That's cool that he can see that. John makes tactical decisions now about the games, "The gun should react like this"

 ATI has its first Cinema 2.0 with voxel rendering. Is that something that's generic, or could Nvidia do that? Carmack can't speak to specific implementation details. Everything's going throgh triangle rasterization pipelines. The underpinnings aren't built for voxels, but this is real programming now. You can be much more clever with these general purpose architectures now. 

How's stuff going with Valve and Steam? What do you think about digital distribution and your relationship with valve? He's happy about the id megapack, because he doesn't want old games to just disappear.For the current slate of products, they're probably going to be boxed goods, without a novel online distribution. Big bet titles will remain conventional (Rage and Doom4, no digital distribution.

He plays a lot of Wii and DS games these days. Looks at all the serious titles, and everything looks really great. There's a lot of great talent in the industry, and you don't see a lot of people releasing games that have amateurish mistakes or flaws.

He doesn't like software patents. He can put a dozen developers in a room, and half will come up with a similar solutoin. It's about an equally smart person who followed a similar development path and he feels strongly about the fact that that's bad. Gaming industry is hyper-competitve. In aerospace, it's a lot different. In two decades, not much may change, but everyone shares info.

More Quake Live info. The baseline isn't out of the box Q3, but lots of what the community evolved to support competitve play. 

Will there be closed-captioning or subtitles in id's games for deaf and hard of hearing? There was closed-captioning on Doom 3, from an end user mod.

Are there any suprises from id Tech 5? The neat thing about id Tech 5 is that artists can build stuff that looks really amazing in a really short period of time. The stamping approach makes them able to do something using Megatexture that's really unique. The first game never has all the capabilities right, but the engine gets better and better as ti evolves. 

 Quake2 was an accident, so they finally just said screw it, we own the trademark, we'll use Quake.

Sequelitis isn't necessarily a bad thing. Call of Duty 4 kicked ass, and it was different than Call of Duty prior. We can't make games fast enough for people who want to buy them. People know what they're getting with franchises and they know what they're going to get for their $60.

Maybe would have been better off to just make the next game twice as good, rather than tear up all that work and do a new IP and need all new art. 

What's your feeling on Linux? I can't say that I feel really good about Linux.  Mac is the next target for Quake Live, after that maybe Linux. The AAA titles won't have day-and-date :Linux ports. If someone wants to do it internally, on their time, they can, but not on company time. PC problems are an order of magnitude worse on Linux.

He doesn't see conditions that are going to force that to change, unfortunately. Linux might not be good as a console OS, but open source would be good. Basic OS of current consoles are pretty competent. Linux is not even on our radar right now for our current projects.

And that's it! Check back this weekend for more Quakecon 2008 live reports! Happy fragging!