Carmack Talks Rage, Other Stuff

Polish games mag CD-Action (who, according to these videos, seem to have their own drink?) have posted an enormous John Carmack interview to the web, and we’ve reposted it below. In it Carmack chats about Rage, stepping away from graphics to work on player experience, the problems of contemporary developments, the challenges of modding in the “post-Doom 3 era”, and so forth. Thanks to Pat at VG247 for the heads up on this.


  1. Baka says:

    Dude – did you really kill that cat?

  2. kai says:

    No, they don’t have their own drink, it’s just that the brand was hidden so it wouldn’t be product placement. (I still wonder whether that was a joke from you or not D:)

  3. Dubbill says:

    Play all three videos at once for instant Being John Carmack.

  4. Stupoider says:

    @Dubbill: Man, that’s insane! xD

  5. Sam says:

    @Dubbill: Yes, you’ve hit on something wonderful here.

  6. Nelson says:

    Ugh. Can’t someone just post the good bits so we don’t have to watch one of the world’s extreme über-nerds talk for a HALF HOUR!

  7. Schmung says:

    Multiple nerdasms there. Always interesting to listen to talk about this sort of stuff. Sad to hear him effectively sound the death knell for the mod scene, but I suspect that he’s right.

  8. Senethro says:

    Good god, the man can talk!
    Fascinating stuff.

  9. Turin Turambar says:

    Heresy Nelson! Teh Carmack’s voice is pure sex. Pure geeky technical sex.

  10. Ozzie says:

    They don’t even have to ask question!
    He will talk half an hour just by himself…

  11. Theory says:

    As far as I can tell they’ve got a system that creates unique textures for everything by baking decals, then streams all that data off the disc as needed. I guess they aren’t into digital distribution that much!

  12. rivalin says:

    So his attitude to modders is that they’re too stupid to use the new engine so why bother giving them the chance…what an enlightened approach.

  13. Schmung says:

    He didn’t say that at all..

    The simple fact is that the production techniques of modern games are so much more complicated and time consuming than in the times of Quake etc that it rapidly escalates to a point where it’s out of reach of the hobbyist. Time was you could work on a mod as a hobby and everything would be fine and dandy. It’s increasingly the case now that the only people working on mods are people who are already in the games industry or who are making a concerted effort to get there. You can’t just be an amateur now because the players expect professional level visual quality and the commitment required to achieve that is pretty much beyond what most people with jobs and lives etc can manage.

  14. Theory says:

    (An lo, it came to pass that iD weren’t into digital distribution that much.)

  15. skizelo says:

    In the second video, he slams day/night cycles in games: proog that Carmack listens to the RPS podcast! (does anyone else remember that one podcast where John Walker (possibly) slams them because he kept crashing in Burnout Paradise when it turned dark)

  16. skizelo says:

    ugh, “proof”, and I know synonyms for “slam”. Like “derides”.
    Bring back the edit button.

  17. Vasagi says:

    He sounds like a trannie i met once, where he had such a deep voice when tried to speak like a woman well he sounded like carmack!

  18. Paul Moloney says:

    “Sad to hear him effectively sound the death knell for the mod scene”

    Does Rage come with duct tape then?


  19. Richard Clayton says:

    Still not much info about the actual game itself though. Has there been much information released about this?

    It’s just too easy to make “Fallout combined with Madmax” assumptions just by the look. Any id facts?

  20. DarthInsinuate says:

    I’m trying to decide which is better: Randy Pitchford being very talkative about shooting people’s legs off, or John Carmack being very talkative about sparse voxel octrees.

    I find them both infectiously enthusiastic, and I imagine if they both met at a dinner party the galaxy would collapse into the gravitation pull of their conversation.

  21. Tei says:

    What! a video withouth subs? how I have suppose to see what is it about??. Also, why you guys talk about the dead of mods?, can you guys help me there, have carmack commented something about that?

  22. Simon says:

    I personally found it very interesting, I look forward to seeing what Rage brings to the table in terms of graphics and gameplay.

  23. Schmung says:

    Tei : In video 3, about 30 secs in he talks about lack of mod support in rage and how since the Doom 3 sort of generation it’s become far harder for amateurs to generate the sort of work produced by the large teams that modern games require.

  24. Tom says:

    LOL Dubbil.
    I love this kinda stuff.
    Whenever i see this kinda stuff though I always end up wondering what Sony were thinking.
    What was the logic behind their architectural decisions?!

  25. lethu says:

    C’mon no dynamic day time? and why… because supposedly some places wouldn’t look as nice as the devs meant it to at some other time of the day, isn’t it already the case with reality? Some landscapes wouldn’t look as beautiful at any other time of the day than at dawn or sunrise, isn’t it even the case in stalker games and many others? And it seems not to pose a problem with anybody, it’s even a plus, I like wandering in lands and noticing day time and lighting change, and feeling it’s too late or to early at some given hour. What we will get instead is a game fixed at 12 o’clock or dunno what other hour of the day…

  26. Jad says:

    I can’t watch the video (at work), but as for mods:

    I can see why it is much harder for amateurs to make total conversions or long SP campaigns or even professional-quality multiplayer maps (although I’ve seen a number of UT3 maps that put the lie to that statement) — but much of the power of modding was not in the big stuff, it was in the little stuff.

    The duct tape mod for Doom 3 is the perfect example — I doubt it was particularly difficult to make and it didn’t really add anything new, but had a significant change in the gameplay. Some people thought the game was better with it, and they downloaded it; others thought it took something away from the game, and they did not. That’s the power of the open PC platform.

    How many recent, console-port games have you wished had one little thing changed? How many people would have enjoyed Farcry 2 more if you could have downloaded a 500k “Checkpoint Respawn Timer Fix” mod? How much would the conversation around the new Prince of Persia have changed if you could get a “You Can Die” mod? (I would not have used it, but I’m sure some would)

    Anyway: when Carmack says there is a lack of mod support in Rage, does it sound like he means no big mods, or no modding whatsoever?

  27. lumpi says:

    @Schmung: I respectfully disagree.

    What you describe (and I don’t blame you) is mainly a comfortable myth game developers/publishers love to explain their lack of a proper SDK with. The amount of time to create custom content doesn’t scale linearly with the amount of polygons or texels. I would even go as far as to describe Crysis’ level-editor as easier to use than the Quake 3 one, for example. A lot of technical barriers (compiling, hour-long pre-render sequences for lighting, poor documentation and usability standards, lack of re-usable/procedural assets…) have improved over the years, streamlining certain aspects of development so it more than counter-weights polygoncount and texture resolution.

    The only reason some games are less fun to mod for today, is the dependence on expensive third-party 3D modelling programs such as Maya. If editors, for example, would let modders make simple walls and handrails in-editor instead of having to export them from a third program or provide a versatile scripting system (see Source) development gets easier in an instant. There have always been games which do not have mod-support. It has, however, little to do with “hyper-advanced future technology” (as it has always been claimed). It’s a usability issue.

  28. Schmung says:

    @Jad : I took it to mean lack of a proper SDK/devkit thing with level editors and so forth. It being id I imagine they’ll at least give you something, which is more than can be said for a lot of companies.

    @lumpi :
    I actually agree with part of what you’re saying – the usability and documentation supplied as part of most devkits and stuff that are you’re given are beyond terrible. Often lacking in content, or worse still supplying you with information that is entirely wrong.

    I have an excellent example of this. A mod I’m working on at present used the CSS player models as placeholder content so that they could get on with testing gameplay, maps and so forth. The time came to replace it, but we figured we could still use the animations and just make a new model. Went off and made said model that conformed to the example supplied by valve in their SDK folder. Only problem is that said model and skeleton is not the one actually used by CSS. Months of effort wasted as we tried to work out why we were getting so many horrendous errors and problems. There’s plenty more examples – parameters and features that don’t work as documented etc etc

    A lot of your experience of course depends on your engine of choice – source still requires pre-compiling of lighting etc for maps. Crysis and UT are certainly a lot more useful in this respect as they don’t need too much in the way of building, but they each have their own foibles.

    In any case, unless you choose to set your mod in the world established and provided by the game then you’re going to need to create some additional content and it’s this that is becoming increasingly time-consuming and complicated. Crysis is lovely if you want to make something set on a tropical island, but once you go beyond that you have to start creating content.

    As you so rightly point out, once you get to this stage you need some very specific tools and a skillset. This IMO is where the problem is. Back in ye olde Half Life 1 days a gun would be something like 2.5k triangles at the upper end of things, maybe with a 512×512 texture. Modern source mods are something like 5k and at least a 1024×1024. For UT3 and Crysis you’ll be using a specular map and a normal map as well. With Half Life you built everything in hammer, a pretty easy to use tool. Source still uses a good deal of BSP geometry, but there’s a great deal of prop usage with varying levels of detail etc. UT3 and Crysis rely even more on geometry that is created outside of the level editor and again you’re working with normal, specular and diffuse maps instead of just creating one texture.

    It all adds up to far more work to create something roughly equivalent. You need much more content, far more man hours and a more diverse skillset to work with a modern game engine and create something that matches the standards that players will accept if you branch out from the games universe.

    While there’s still a place for compelling experiences created within the gameworld or with minimal recourse to additional content the mod as I think of it – a totally new experience using the game engines basic features and maybe some of it’s content – is an endangered species. Somewhat perversely, Rage might actually be a bit more mod friendly (based on what little I know of it) than some current gen engines, but that’s sort of a moot point.

    So, as a brief sort of summary, while I think you have an excellent point RE usability there are enormous technical barriers as far as mod creation is concerned nowadays.

  29. l1ddl3monkey says:

    Is the fact that he has exactly the same voice as the mad scientist guy from The Simspons purely coincidental?

    Fascinating talk from someone who probably understands the subject better than anyone else in the world. Bit long but captivating nonetheless.

  30. TheSombreroKid says:

    i agree with him about the ray casting voxel tech, it sounds like a good idea, especially for simd processing.

  31. pkt-zer0 says:

    “C’mon no dynamic day time? and why…”

    Because the engine can’t handle it.

  32. Radiant says:

    The issue with Sony is that unlike Microsoft, where they were free to just [approximately] make a mid range pc in a box with structured online play, Sony has to consider their other departments and partners.
    Bluray, Cell, the styling all linked into other departments and business interests.

    It’s a wonder it came out half as well as it did.

  33. Schrodinger's Lolcat says:

    “C’mon no dynamic day time? and why… because supposedly some places wouldn’t look as nice as the devs meant it to at some other time of the day”

    No no. That’s the retroactive justification of the issue. The why comes from the limitations on dynamic elements within any given scene in this engine. If I had to take a guess, I’d say cooking all of those decals into a landscape is so work-intensive that they simply couldn’t do it for all possible lighting conditions presented by a true day/night cycle in the outdoor areas.

    If you’ve got thousands upon thousands upon thousands of decals in any given area, many of them overlapping one another, there’s probably a whole buttload of calculation that goes into just getting one lighting pass on them for global illumination.

    I’m pretty sure Carmack was trying to be “glass-half-full” here in pointing out that even though a dynamic lighting system outdoors was not in place, the flipside is that the artists are able to choose the exact spectrum range of color, the exact angles of shadows and so on – in other words able to be more precise about the look of a scene.

    No big loss in my opinion. It’s not like Van Gogh’s Starry Night needed to have a daytime view to fully appreciate what he was doing with the night, no?

  34. MrBejeebus says:

    Im commenting before i actually watch these videos, just because of how funny his expressions are..

  35. tapanister says:

    Man, Carmack is a genious, but the interview was pretty borring even if you can understand what he’s talking about – couldn’t go through all that. Can anyone post a summary on the “death of the mod scene” at least? -_-

  36. Mo says:

    I believe the decals get baked into the texture before the lighting pass is done, which would be the entire point of the engine … the number of decals placed at build-time do not affect performance.

    The move away from the unified lighting system of Doom3 does make day/night cycles harder to do, but not impossible. It really does just come down to Carmack prototyping it, the team saying it wasn’t worth it, and Carmack not pursuing the work on it. I’m not sure why that’s so hard to believe.

  37. l1ddl3monkey says:

    @ Tapanister: I think Shmung pretty much summed it up in his comment above. Short version: Modern engines require so much work to create custom content for that creating said content requires a more man hours that most mod teams can commit to.

    I think it might go back to a more “Quake” scene where Total Conversions/Mods were less common than custom maps (but the quality of some of the maps was far, far in excess of what the game creators turned out).

    Which reminds me: Anyone think Black Mesa Source is ever going to be released?

  38. lumpi says:

    The thing about real-time lighting is that, unless you do radiosity or proper global illumination which would probably destroy most of today’s hardware, it will always look artificial. If surfaces don’t reflect lighting properly, you can never simulate the glow of bright sunlight correctly. Which would be bad for a desert-based game such as Rage.

    It’s funny because Valve used pre-rendered light for the same reason, back when Carmack rigorously defended full real-time stencil shadows a la Doom 3. Guess even he was beaten by the aesthetic side here, settling for a technological compromise so seldom found in id engines.

    @Schmung: Yes, all quality points. What I meant was mainly the difference between “graphics” and “graphics technology”. Or “Game design” and “game technology”. An artist that could do an acceptable looking 500 poly weapon 8 years ago, would now be able to do a 50000 poly one as well. Most of those polygons are automatic curve-smoothing anyway. What makes it look good are correct proportions. In fact, there is less to worry about optimization, less paintery in Photoshop, since you can actually model those little ripplets instead of hand-painting them. Playing with shaders is hardly the reason for delays here (and let’s not forget the unintuitive, slow, buggy stuff and the lack of online tutorials we had to deal with 10 years ago). The amount of work, in total, might have increased by 15%, but not by 1500%, as you might hear Carmack suggest in this piece. Doing a mod properly would maybe take 3 months more, not 3 years.

    Doom 3 and Half-Life 2 came out the same year. ModDB lists 1,358 mods for HL2 and only 82 for Doom 3. Not to mention the quality of said mods. Why is that? John Carmack complained about Doom 3’s mod scene being so small, but quickly found an explanation: The Doom 3 engine is too advanced! Gee, I bet he could live with that.

    But seriously, is Doom 3 really more complicated than HL2? I doubt it. More efficient, sure, but probably even less “complicated”. Does it rely more on custom model work? Maybe a little, but there’s a lot of model props in HL2 as well. I hope you see what I’m hinting at. I can’t tell you exactly why HL2 is so much more popular for modding than id’s last major creation, Doom 3, but I’m pretty sure it has nothing to do with the amount of models in-game. The fact that there are only so many ways of doing Martian hell-labs (or tropic islands, underwater metropolises, dystopian wastelands,…) might have more to do with it. But look at HL2’s default set of content: Gray/brown Eastern-European war torn cityscapes? Not exactly the freshest and most versatile content to work upon either.

    I tend to reject Carmack’s excuse for the poor modding scene of their later games. It’s not technology that killed the id modding scene. The story, IMO, already started with the poorly hacked together Quake 3 (“OMG, we created a new FPS genre by doing Quake 1 with better graphics while cutting singleplayer, story and stylistic direction!”).

    If Valve will ever drop the artificially-kept-alive oldschool part of their Source engine (L4D2 already looks a bit outdated) and creates a proper “Source 2”, I bet it will have good mod support. While id will have stopped trying altogether, long time before…

  39. Schrodinger's Lolcat says:


    I was under the impression that they did not effect playing performance but they were severely linked to time it would take to design an area. I thought that’s what he was implying with his comments about how it could take 20 to 30 seconds to lay a single decal in certain sections – in other words an issue of performance for designers, not players.

    So I extrapolated from there that some sort of lighting pass across all the decals in a scene for every shade and position possible from a dynamic day/night system was unfeasible or not worth the extra calculation time per decal in the outdoor areas? Once again, I’m talking about the design time – the amount of time it would take to place a single stamp under these conditions.

    Not saying you’re wrong at all though – did I misunderstand that bit?

    Where did you get the notion that the decals go into the level without ever needing to calculate color and shadow from the static lights that are present? I assumed that was part of the process of “baking in” he referred to.

  40. lethu says:

    @Schrodinger’s Lolcat
    Oh I see, so it’s decided more against a pure technical barrier basis. I thought it was more of an artistic point of view based decision. Well, based on this I think I wouldn’t have much problem with playing the game the way they want it to look, given that the fixed conditions you will have to play through look good enough.

    I think I should practice more on my English listening skills, I totally misunderstood this point when watching the vids,
    Carmack got a pretty easily understandable accent though, it was nice listening to him talking :)

  41. Martin Edelius says:

    Carmack might be a genius but he’s a boring one…

    +1 on the transcript idea!

  42. Aubrey says:

    I got my start in modding, but I have to agree with Carmack. If modding Quake/2 wasn’t so easy, I never would have taught myself to code. If I had started today, it would have been a false start. Better to start experimenting in something like Game Maker. Even XNA would have been a push for me as a beginner.

    I think there’s still aspects of engines you can use to try things out – you might be able to teach yourself how to make a small level, or assets, characters, a weapon etc. etc. but a full blown total conversion is out of the reach of so many when compared to how it was in the Golden Age.

  43. vader says:

    Why doesn’t RPS have their own drink! This is an outrage! I demand you make your own drink and put lots of alcohol in it!

  44. int says:

    Am I exaggerating if I say John Carmack is the Leonardo da Vinci of games?

  45. Theory says:

    ModDB lists 1,358 mods for HL2 and only 82 for Doom 3.

    Yeah. Now filter for released mods. Now you’re down to 242 and 43. Considering that Source has a far far larger playerbase, has been receiving incremental upgrades for five years, and will continue to receive them for the foreseeable future, 5:1 isn’t too shabby.

    The falloff from Source’s figure, incidentally, largely represents all the excited people who tried to create a whizz-bang amazing “professional” mod and failed.

  46. Schrodinger's Lolcat says:


    I would suggest reading lumpi’s points on this as I think he makes a compelling case that the height of that bar is directly proportional to the quality of the tools presented to the community and the documentation available as to their use.

    @lumpi, Schmung, etc.

    As an aside somewhat, I think it would be interesting to see a developer specifically making general-purpose “generic-brand” content not for use in their title but to seed content-creation for the modding community – sort of the equivalent of a starter pack in Magic the Gathering (pardon the nerdaphor.)

    I don’t know if this would be feasible or even worthwhile for modding teams, but even if only to use for as stand-ins until the final model comes along… I guess my feeling is that every bit of help you can get really goes a long way. Maybe someone with more experience in the mod scene can speak to this?

    Aside 2: Given the supposedly vast flexibility of the editor in Starcraft 2, what do you think their tools will look like? Blizzard is notorious for creating looooong-lasting titles and supporting their titles and fanbase well beyond launch – perhaps they might be a good metric by which to judge these sorts of things?

  47. Dworgi says:

    Schrodinger’s right. The biggest problem with making mods is that game developers these days do extensive building and exporting and baking of textures, geometry, shadows and decals. The developers themselves amortize these costs with amazingly complex build systems that utilize racks of servers and cache all the assets that someone else built a few minutes before you. Amateurs just don’t have access to that, and getting a single level into the game could take upwards of an hour for the first time.

    The other problem is, as someone mentioned above, that Maya or 3DS Max or XSI are requirements for modern game development. The tools are built for internal processes, and the internal processes rely on these products, so there’s no way to split the SDK off from the modelling/animation package. Well, that’s not strictly true: there is a way, but it involves reinventing the wheel and duplicating functionality that modelling packages already do well.

    Also, most developers don’t invest heavily in tools unless they foresee it being used by mod makers or licensing the engine (both, in reality). Releasing an SDK isn’t just a case of pushing out the tools that the game was built with, because in many cases those tools are largely ad-hoc (read articles on Gamasutra about Turbine’s tools process, for example) and would get absolutely slated by someone who wasn’t being paid for using the tools.

    That means more polish to get the SDK to a stable, usable state, because there’s no in-house developer to give support or slap in a quick bugfix when something unexpected happens. Releasing an SDK essentially involves creating two products – the game and the SDK – where developers invest heavily in creating the game and the tools are only there to support the creation of the game (and rightly so).

    I’d much prefer to play an excellent game than fiddle with an SDK, but perhaps that’s just personal preference.

  48. Schrodinger's Lolcat says:


    Put perhaps a more interesting statistic than either of these is: how many of those released drastically change game mechanics/aesthetics to the point of qualifying for “Total Conversion” status. I’m talking about your Natural Selections and Counter-Strikes here.

    (pardon my double-post.)

  49. reaper47 says:

    @Theory: You have a point. I can trump that with an argument for quality, though. Look at the best 10 mods for Half-Life 2, compare them to the best 10 for Doom 3. Most D3 ones are small code fixes. I do not see a single total-conversion mod either.

    Granted, I’d love to see more experimental/gameplay-driven mods (Source Forts, Garry’s mod, Research&Development etc) and less compete-with-professional-quality ones that never see the light of day. But that strikes me more of a philosophical aspect than a technology-driven one. Look at the indie-games scene. It survives without pixel shaders. If you accept basic place-holder graphics, you can experiment just as much with HL2 as with HL1. And even the other way round, if you look at HL1 mods such as Natural Selection, I can guarantee you, with experience in both engines, it was as much work to do as Insurgency for HL2. It’s a stylistic trend in development philosophy that already started before Doom 3 & Co. And, with support of the gamer community, it can be replaced with a more originality-driven one, eventually.

  50. Biz says:

    graphical modifications were always hard, so yes that scene is pretty much dead

    but gameplay mods aren’t. people just need access to some source files for game-changing mods or scripts. and they need an editor for making levels.

    editors becoming more sophisticated will eliminate the need for a game to be modification-friendly for 99 percent of the user-created content to exist.