Notch vs Unlimited Detail

That's not unlimited. Why, it's only 600 pixels wide

Mr Minecraft himself, Markus ‘Notch’ Persson, isn’t at all impressed by Euclideon’s bold, bewildering claims of an ‘unlimited detail’ graphics engine. And he’s not talking about that voiceover guy – he’s talking about the engine itself, and alleging that it isn’t really what it claims to be. So he’s broken down why in great detail. Caution: science. Also, videos.

Here’s the main thrust of his assessment:

“They made a voxel renderer, probably based on sparse voxel octrees. That’s cool and all, but.. To quote the video, the island in the video is one km^2. Let’s assume a modest island height of just eight meters, and we end up with 0.008 km^3. At 64 atoms per cubic millimeter (four per millimeter), that is a total of 512 000 000 000 000 000 atoms. If each voxel is made up of one byte of data, that is a total of 512 petabytes of information, or about 170 000 three-terrabyte harddrives full of information. In reality, you will need way more than just one byte of data per voxel to do colors and lighting, and the island is probably way taller than just eight meters, so that estimate is very optimistic.”

He also notes that the video shows “repeated structure[s], all roughly the same size”, claiming sparse voxel octrees are very good at referencing the same data over and over. He also reckons this system wouldn’t be much cop for animation.

Nonetheless, “It’s a very pretty and very impressive piece of technology, but they’re carefully avoiding to mention any of the drawbacks, and they’re pretending like what they’re doing is something new and impressive. In reality, it’s been done several times before.” Which he claims is demonstrated to some degree by these projects.

The Atomontage Engine:

The engine for Voxelstein, created by the guy who made Duke Nukem 3D’s engine:

SVO raycasting from 2009:

Obviously these aren’t exactly the same as Unlimited Detail – and I’m probably the wrong guy to ask such techery of anyway – but they are, to Notch’s mind, using the same technology. For that reason, he alleges that “They’re hyping this as something new and revolutionary because they want funding. Don’t get excited. Or, more correctly, get excited about voxels, but not about the snake oil salesmen.”

More here. Will there be a riposte from Mr Euclideon? We can but hope.


  1. Jeremy says:

    While I am definitely skeptical about this infinite detail claim, I am excited that people are out there experimenting and breaking new ground on what could, in years to come, be the next “big thing” in technology.

    • Phoshi says:

      Except this isn’t new ground, it’s a voxel engine, they’ve been in the “Really interesting but ultimately not as viable as polygons” bucket for decades. This is nothing new.

    • Jeremy says:

      I realize that this is nothing new. But I support all parties that look to advance the future of games. Even in failure, new ideas can come to light and push boundaries.

    • LionsPhil says:

      Yup. And that was evident at the time.

      But I’m glad Notch has taken the time to throw his weight behind this. It is all marketing fluff and does not deserve promotion from RPS.

    • Groove says:

      Anything that pushes towards a Horace-filled future requires RPS support.

      Infinite bear requires INFINITE DETAIL.

    • Malibu Stacey says:

      Horace is endless not infinite.

    • JFS says:

      But Skyrim’s dragons are, and just imagine how great they would look then!

    • Groove says:


      That’s pretty much a tautogoly.

      If he’s truely endless than that implies infinite. The other options involve his end being merely imperceptable, in which case it still exists; or displaced, in which case he would still end, just not at a regular bear’s end.

  2. Unaco says:

    Well said Notch.

    • Faxmachinen says:

      I do like Notch, but it should be obvious to everyone that he’s not really one to think out of the box. From what I gather, the Unlimited Detail engine is to a voxel engine what a vector graphic is to a bitmap. To prove that there is no “unlimited detail” through math is rather naïve and/or missing the point.

      I may very well be wrong, and the Unlimited Detail engine may just as easily be a glorified voxel engine, but that still does not mean that the idea has no merit.

  3. IDtenT says:

    THIS IS INTERNET WAR! Aka crap nobody cares about. Give me a game to play instead. :D

    I’ve also pointed out in the comment box of the previous article that there are information theoretic limits that imply that simulating individual particles physically would require a machine greater in scale than the simulation itself. (If the complexity of the particles stay the same.)

  4. wccrawford says:

    Had they not used the word ‘infinite’ I could think maybe they have something. But ‘infinite’ isn’t possible on a finite computer. It’s not even desirable!

    They also have yet to show that they have anything that can’t be done with current tech. As everyone is noting, they aren’t showing animation or real games.

    But! They are -comparing- to real games.

    When they have a real game that out-details everything out there, then I’ll be interested.

    But only if it also has a good plot and control scheme.

  5. deanb says:

    Via the sleuths of reddit it appears they’ve being shilling this stuff for quite a few years: link to

    And as Notch pointed out they’ve not really made a point of the many drawbacks it has and why this seemingly marvellous “ultimate detail” rendering method isn’t used today.

    And not just Notch whose spoken of this but John Carmack too: link to goes a fair bit in-depth on it’s advantages and limitations.

    It seems like it would be great for environments, but since we’re getting in the mood now for dynamic and destructible environments it might not be worthwhile in the end. (Though a fair few genres still do fine with static environments)

  6. skinlo says:

    The Atomontage Engine is the more impressive on there, if you look at some of the other videos he’s done, you see how the sand is actually being moved out of the way as he drives. He drove back and forwards over a rock, and after a while he actually bogged down the vehicle so it couldn’t move.

    • James says:

      RPS needs to listen to me right now!

      Atomontage is quite a bit different than the other examples given, and even at this early-ish stage of development that comes through.

      Not that I’m deferring to Notch exactly (as he’s by his own admission making a guess about what he’s seeing with unlimited detail), but he’s already expressed enthusiasm for Atomontage and for good reason. It’s is an attempt at a game engine above all else, so if you have the interest and take a bit of time to find out about the project you’ll be glad you did. Unless you’re a dullard.

      Notch lumped it in with the other examples above, which was unfortunate. I’m just glad it’s getting some attention, even if it’s a footnote to another story.

  7. thestjohn says:

    I love the look of the Atomontage engine, not seen that before. I was not similarly impressed with Infinite Detail. I mean compare the Euclideon video’s use of, as Notch pointed out, uninteresting and unanimated repeated objects, to the APC carving out deep tracks in the sand in one of the Atomontage video blogs. Especially as whoever did the voiceover for the Infinite Detail video managed to find an approach and level of bullshit that annoyed me to the point of fury.

  8. SirKicksalot says:

    John Carmack:

    “Re Euclideon, no chance of a game on current gen systems, but maybe several years from now. Production issues will be challenging.”

    link to

    Carmack > Notch

    • 1stGear says:

      Which is notably less optimistic than Insufferable Accented Narrator. There’s is a vast, vast, vast gulf between “Could be interesting a few hardware generations down the line” and what Euclideon is trying to claim.

    • diebroken says:

      Carmack takes him down a notch… again.

    • karry says:

      “Carmack > Notch”

      Eh, i wouldnt trust Carmack so much, he’s too often getting beaten at his own game.

    • The Innocent says:

      “Eh, i wouldnt trust Carmack so much, he’s too often getting beaten at his own game.”

      The game of being the principal designer/programmer of boundary-pushing engines? Consistently?

      Uh… who’s beating him at that game?

      I’m not some sort of Carmack apologist or really a fan of id’s most recent games, but I don’t see how you can make that claim.

    • Atic Atac says:

      John Carmack is pretty much beyond criticism at this point. The single most important programmer for gaming. Period.

  9. pkt-zer0 says:

    Oh, I figured out why I remember Carmack talking about sparse voxel octrees – QuakeCon 08, this is the tech he’s planning to explore in id Tech 6.

  10. molten_tofu says:

    While I’m certain Unlimited Detail is over-hyping themselves, you don’t have to switch formulas to do something radically different – sometimes you just have to tweak. I’m always suspicious of claims that this or that is the same because it has identical foundations – after all the core i7 and the 286 are both x86 and not only that are both built using transistors.

    • jalf says:

      Yep, and there’s very little *new* about i7 from that point of view. It was just developed at a time when it was possible to cram thousands of times more transistors onto a single die.

      That’s not even “tweak”, that’s a plain and simple brute-force approach to solving a problem.

  11. LeJosh says:

    ID tech 6 will be interesting to say the least.

  12. Xighor says:

    Well, here, in the previous movie [ link to ] our beloved Aussie states (05:50) that it is not a typical graphical engine as it uses a search engine and displays only the visible ‘atoms’, therefore, the computer is not busy displaing those which cannot be seen from the current camera view. That explains the unlimiteness of this engine.

    • BurningPet says:

      but…. first of all, this really isnt something new (rendering only the visible screen) and second, isnt the unseen data still require space on the hard drive?

    • aldo_14 says:

      Determining what is actually visible involves using computational resources of its own, does it not?

    • Xighor says:

      Does google display the pages about sex toys to you if you searched for a quantium physics? :}
      I think that they look at the graphics from the different way: you’ve got 1600×900 pixels on the screen, so we search for 1440000 atoms visible. And not for more. Not the only way around: we generate all the atoms and then eliminate the ones you can’t see. That is how I see this engine.

    • Amun says:

      @Xighor: That doesn’t change or make up for the drawbacks discussed by Mr. Notch. Namely, storage of all the atom data and how to make animations work without looking horrible.

    • Samuel Bass says:

      Yep, that’s occlusion…it’s a pretty common feature in most (polygon based) rendering tech these days.

    • hexapodium says:

      @Samuel Bass: Occlusion still requires you to search the whole scene, then cull things which are occluded in the z-plane. The speed improvements seen in raster-polygon graphics tech are associated with how expensive it is to fill/map/texturise/shade polygons, and how occlusion-culling lets the engine not bother to render those polys, rather than the search cost (which is what would be incurred by the Unlimited Detail concept)

    • Lusit says:

      So it supports culling?

    • Baines says:

      The secret of storage is probably repetition. The video shows a *lot* of object repetition. When the video asks viewers to excuse the repetition because they aren’t artists, I think that is more a matter of excusing how visible it is rather than to realistically expect a large island to have no repetition.

      Instead of building the entire island from individual atoms, build the island from specific objects which in turn are built from atoms. Think of it as a 3D version of tile based graphics, where a world is created from a limited set of tiles and the tiles are created from pixels. Or even build the island from combinations of objects themselves built from atoms. (Think of a tile based system where the world is composed of blocks of tiles, themselves composed of individual tiles, in turn composed of pixels.)

      The video shows only a few different trees, and the trees themselves appear to be composed of recycled parts. (All the leaves of one are identical, the palm tree trunk might have been made of a repeated cylinder, etc.) The same statues get repeated a large number of times. Though I didn’t try to analyze it, the rocks are probably repeated, and possibly sets of them are. The sky view of the island shows extremely obvious repetition of blocks of trees and ground.

      If they wanted to show their engine could realistically handle a lot of variety, they could have done it fairly easily. They could have just gone to some of the websites that host free 3D models, downloaded a ton of them, converted them, and just tossed them all onto the same landmass. Instead, they went with a rather simple world map composed of repeating the same objects over and over, pretty much using a basic tiling (cubing?) method.

  13. Jibb Smart says:

    “If each voxel is made up of one byte of data, that is a total of 512 petabytes of information, or about 170 000 three-terrabyte harddrives full of information. In reality, you will need way more than just one byte of data per voxel to do colors and lighting, and the island is probably way taller than just eight meters, so that estimate is very optimistic.”

    Atomontage is looking at around half a bit per voxel. Before you say “You can’t store less than 1 bit!” it’s got a groovy compression algorithm, apparently.

    Also, any game has a lot of repeating stuff. That’s not the issue.

    The issue with Unlimited Detail, as mentioned, is that nothing is animated, and the shader options seem extremely limited. And by ignoring tessellation as a solution to the flatness of worlds, the narrator comes across as either intentionally deceptive, or oblivious to actual gaming technology:

    Tessellation example with Unigine.


  14. Sagan says:

    In their last video they claimed that they weren’t doing voxels. Also they claim that they are doing their thing on the CPU. The SVO stuff is done on the GPU. So they probably aren’t doing voxels.

    As for the amount of data: Notch is talking about uncompressed stuff. If there is repeated data there is obvious room for compression. Also they are only storing the surface of the objects, which Notch didn’t include in his calculations.

    And when they are talking about “unlimited detail,” then it should be obvious that they mean “we aren’t limited by rendering power anymore, but rather by how much data you can fit on a disc.”

    As for whether it’s a scam: I kinda believe that they are for real. But even if they are a scam and you can’t actually do this on the CPU, then within a couple years we will be doing this on the GPU anyways with voxels.

    • Khann says:

      Yes, the fact that only the surface is rendered reduces these numbers by a quite considerable amount I imagine.

    • othello says:

      No. Even with compression we are still at O(n^3) asymptotic space requirements. This technology is not new, not innovative, and is stuff that we already know how to do. It has not made its way into games because it isn’t practical. In computer graphics academia it is well known that voxels are essentially useless (bar a few exceptions), both because they have limited resolution and because of the ridiculous memory requirements. They are nice for cool and creative projects but aren’t practical in the way these guys are marketing it.

      In addition, we are not limited by geometry. We can push polygons out of modern cards like no tomorrow. Rather we are limited by fragment processing such as lighting and shading. If they tried to do shading similar to that found in modern games in software their fps would drop way below the meager 20. Notice how there are no real shadows (they are all projected from a light source straight above)? Even the image they showed with the “new” shadows appeared to be more ambient occlusion than actual shadows, which is simple to do in a voxel engine (hell minecraft does it and it’s not considered a graphical powerhouse).

    • Dave Mongoose says:

      I’m generalising a bit, but compression typically works by finding patterns in the data and representing them in a ‘shorthand’ form.

      If you need your data to be compressable, then that imposes limits on what the data can look like: The Complete Works of Shakespeare will compress much better than a string of random letters of the same length because there are certain sequences of letters that are common in English (like ‘the’ or ‘ing’).

      It isn’t too hard to believe that there are patterns in natural landscapes, but it does limit you if you want to have an artificial landscape like a city.

  15. RF says:

    Because, as everyone knows, Notch is an expert on engine programming. *cougheveryonesayshiscodingisshitcough*

    • diebroken says:

      *coughinfiniminercough* ;)

    • Anthile says:

      *citation needed*

    • RF says:

      Just ask any Minecraft modders you know. Notch’s code is a mess.

    • qwiggalo says:


    • zbeeblebrox says:

      I’m not going to take some modder’s word for it. They’re so hit and miss, for all we know half of them code in Pig Latin. Obviously, Notch wouldn’t have been commenting out a lot of the more obscure stuff seeing as how he started this project on his own purely for fun – which is probably the majority of what the mod community’s complaints are about – but Jeb was clearly able to figure it out and there are a million mods even *without* official support, so there’s no way it’s truly as unreadable as all that.

    • Dave Mongoose says:

      I was also going to make this point. Even without looking at the mod tools, etc. I can guess that minecraft’s code is a mess because he didn’t program it with a client-server architecture, then tried to add multiplayer.

      Normally for a multiplayer game (even playing it in singleplayer mode), the ‘server’ process would do most of the calculations and keep track of variables like health, while the ‘client’ process would handle rendering and communicate player input to the server. With this design, if you add a new feature, you mostly just update the server-side code.

      Minecraft wasn’t originally planned to have multiplayer, so at first the main bulk of the code was client-side and the server just relayed what the other clients said. Even now, things are more or less duplicated between client and server with some wrangling if they ever disagree.

      That’s the main reason why multiplayer was so limited compared to singleplayer for so long, and while it will always be a nightmare to maintain unless he does a pretty comprehensive re-write.

      In short: Notch isn’t really one to comment on other people’s programming.

  16. tyrsius says:

    Notch makes a pretty silly assumption that no sort of compression or algorithmic generation is being done to store/create the atoms. This seems like a very silly assumption to make, since almost nothing is stored uncompressed anywhere.

    • Quirk says:

      Notch isn’t making the assumption that they’re not compressing their atoms. His reference to the amount of data you’d need to store uncompressed makes that pretty clear. He goes from there to assuming sparse voxel octrees, which would have the compression oomph to show off what they’re showing, and mentions some of the fairly major disadvantages of working with voxels with regard to animation. Given they aren’t showing off anything that couldn’t be rigged up for demo purposes with known technology, there doesn’t seem any compelling reason to believe they have anything that isn’t simply snake oil.

  17. Galcius says:

    We don’t need to talk technology to know that this is all bollocks. All we need to ask is “what sort of company makes a massively hyped announcement and then disappears completely for most of a year with no explanation, returning only to show another massively hyped video?”

    The answer is, of course, one which has nothing to show for itself, and knows it.*

    If there was anything practical to this, then Euclidean would get buckets of money thrown at it by investors or bought out by a large firm. Instead I think this is either the modern equivalent of a traveling snake-oil peddler, looking for investment money that they can line their pockets with before mysteriously going bust, or is being spearheaded by one of these people who are convinced that they’re years ahead of everyone else, but are in reality years behind**.

    *(Well, games developers do this sometimes too, but usually their claims aren’t so grandiose)

    **For example, a particularly sad case I came across was somebody who wrote his own “Revolutionary” OS. Which couldn’t run on more than 4mb of ram. Yes, MB. Yes, recently.

    • Xighor says:

      Oh well, the world is not so simple. Good projects not always get financial support for various reasons and every company has to fight for the money.
      [ Did you know that blue laser (used now for blu-ray discs) was invented in Poland in 90s (!), but did’t get the proper funding so the Sony bought the technology and made use of it. ]
      If they do not have the money, they do not have means to push the project further in a proper pace. And then they are releasing a short promotion video. Not a surprise really.
      Whether it is going to change graphics in games or not, I cheer for all new/original/clever ideas as every single one pushes us forward. And if this or other engine had greatly improved the graphics well… I would’t mind.

  18. Soon says:

    I use point clouds a lot but in a different industry and don’t really come across the other terms that often. So, can anybody explain the difference (if any) between a point and a voxel? Because nobody here (in the office) ever refers to points as voxels and these are supposed to be point clouds.

    • Soon says:

      Okay, so a voxel is 1x1x1? A point is not. It’s 2D. If these are point clouds, they’re just clouds that are dense enough to appear solid. And, of course, hollow in the centre.

    • Xighor says:

      Someting that is 2D can’t be a point. The trick with points is that they don’t have dimentions at all.

    • Quirk says:

      Mmm, the point is in matter of fact infinitely small. Rendering it can be done in a number of ways, most treating the point as effectively a small sphere. Carmack suggests ray tracing and splatting as the most likely methods used – these methods are both typically used in voxel rendering.

      Anyway, another important distinction between voxels and point clouds is that voxels are regularly spaced, breaking space up into 3D voxels in the manner that bitmaps break up 2D images into 2D pixels. This gives you some fairly tasty compression opportunities that less regular point clouds lack.

    • Soon says:

      Well. I know what I mean. They have no volume, is the point. (Ha). (Edit: which seems to only be true in how we use them anyway, so, hm)

      I guess it’s just a terminology thing, and we probably use it incorrectly. A point to us is visible, marking the coordinate location. So a point cloud came to mean the visible cloud and not a giant list of coordinates.

      Thanks for your responses.

  19. Mike says:

    I dunno, irritating as the spokesguy is, ‘scam’ and ‘snake oil’ seem pretty strong words to describe what is basically a middleware hopeful. Ponzi it ain’t.

    • Shuck says:

      Well, given that the tech appears to be something well known but presented as revolutionary, that their demos are carefully put together (instancing a couple of objects to fill the screen) to give the misleading impression that it scales in ways that it most likely can’t, and that they’re comparing static, un-animated, un-shaded, repeating objects to actual functioning game footage… yeah, snake-oil seems about right.

    • Pointless Puppies says:

      Can you honestly not catch the whole “I’m selling snake oil” vibe on the video? Curtis Stone over there was practically yelling it at you with his “AYE MATE. CRIKEY! THESE BRIDGES ARE SO USEFUL AND CHEEP!” routine. All they needed was a series of fake “before/after” ‘shops and they could’ve easily landed an infomercial spot at 2 AM on CNBC.

  20. wisnoskij says:

    As someone at least moderately educated in the right field, I am really not sure about the accurateness of Notches claims (but I admit he is likely to know more about every detail of this stuff then me).

    But size concerns? obviously no graphics technology can reasonably store all the details for a entire island filled with completely unique trees/rocks etc.
    Also to animation limitation, well how often do you have animations on the map in any game, and what prevents you from using normal polygon graphics for the animated objects and people?

    And well everyone marketing something are always overly enthusiastic about there product and talk about how it is the only solution to whatever problem it is designed to solve.

    That is not to say this is not vaporware or a scam, but notch’s examples of other engines seem to prove that a engine utilizing this idea is possible and not at all necessarily a scam (not that UD has not always seemed a little iffy).

    • diebroken says:

      Absolutely. Novalogic have been using voxels and polygons in their games for years. Crytek also use voxels to a degree in their interations of the CryEngine.

      link to

    • LeJosh says:

      I think Notch is condemning them purely because of the marketing crap and purveying it as new.

      My prediction is that we will see polygonal/”atom” hybrids. Such as static scenery like mountains/rocks will be used using voxels or whatever but character animations etc will still use polygons.

  21. Zogtee says:

    I don’t quite understand why we are so eager to shoot the idea out of the sky or make snarky comments about the narrator, but whatever.

    I do vaguely remember a french developer way back in the day, who released a fighting game where all the characters were 3D, but created out of pixels and not polygons. It looked amazing at the time, but was so slow that it was bordeline unplayable. That sounds a bit like the claims in the video of building objects out of atoms.

  22. tanith says:

    I was promised science. >:(

  23. Squishpoke says:

    Even so, I’d imagine you could use this technology just for “repeated” things such as the ground.

    I wonder if id Tech 5’s megatextures would work with this tech…. but that’s just my ignorant brain coming up with something that might not work.

    • LeJosh says:

      ID tech 6 has been rumoured to use some form of raytracing/voxels.

  24. vash47 says:

    I don’t buy this, but notch isn’t the right person to comment on this since he can’t program for ****.

  25. Hoaxfish says:

    There’s something terribly depressing about youtube videos without any sound

  26. Bart Stewart says:

    Miguel Cepero — he of the astonishing “Procedural World” work-in-progress — has also commented that the claims made for Unlimited Detail are excessive: link to .

    • SirKicksalot says:

      It’s hilarious how many people say the claims are excessive because of huge power requirements. The Euclideon guys say that themselves…
      What’s the big difference between showing this and Samaritan or the ray-traced id Tech demos? None, except that this one comes from an enthusiast 9 people developer and the others from established developers. It’s a glimpse of the future, and people invoking processing power as why this can’t possibly work should shut up.

      It feels like back when people hated on polygons and kept playing their fast and beautiful 2D games.

    • othello says:

      @SirKicksalot Well, I think people are bothered more by the fact that they claim to have something 100000x better when they don’t have anything out of the ordinary. Raytracing in real time is certainly possible, at quite high framerates. Voxel based solutions are proven mathematically not to work well because of their cubic asymptotic memory cost. (I’m saying voxel because that’s probably what it is; point clouds are even more ridiculous).

    • qwiggalo says:

      There wasn’t a seamless jump from 2d to 3d polygons, if you might remember we had games like Doom which had no polygons but were 3d.

  27. patricij says:

    Ken Silverman’s Voxel engine is pretty solid, but sadly discontinued as fair as I know..Still, there are people working with it, tho

  28. Shadrach says:

    The Atomontage engine looks very good – maybe one day soon we can have *proper* sand or snow in games, not just textures that make “dust”. I would imagine destroyable environment would be (partly) made with voxels.

  29. CharlO says:

    What Notch says is right and all, but I dislike his point of repeating dots as a drawback. We have 118 dots, and use roughly 90 most of the time, and quite an impressive world. I think that should be the focus of deployment.

  30. Vagrant says:

    While I am dubious about this as an actual game engine, and even made the same points Notch made in the previous post, I’m going to play devil’s advocate.

    Has Notch never studied file compression? Seems pretty easy to compress that data.

    Also, complaining about them using the same models over and over isn’t any different from how things currently work in the industry.

  31. Vaughn says:

    Was I the only one that watched their older video where the guy says their engine is NOT voxels, not polygons and not ray-tracing. Of course, they could be lieing, but why? Considering they are adamant about putting the tech out for trial relatively soon. Notch makes some pretty big assumptions here considering they say the opposite.

    Atomontage looks pretty incredible though.

    • Quirk says:

      I find it interesting that most of their image gallery shows point clouds of the type that would be reasonably trivial to render on a modern PC instanced multiple times: look out for repetitions of items in the scene, especially trees, tufts of grass, etc.
      link to
      They may be doing something a lot less impressive and more limited than sparse voxel octrees.

      Also, they’ve been promising the moon for years, so I wouldn’t hold my breath waiting for them to release an API.

  32. wererogue says:

    Ultimately, I’m reserving judgement until we get our hands on a demo/code, but here’s some of my initial thoughts:

    1. Assuming that it’s a standard voxel engine seems premature. They talk about a point cloud system, and it seems to me, from the screenshots and as a programmer, that what we’re talking about here is a point cloud that describes an object, but is not necessarily the entire surface of the object. I think that the object is regenerated from the point cloud data in real time and rendered, which would bring down the storage costs a lot.

    2. We’ve only seen one brief shot of animation (and unless I miss my guess, it wasn’t actually footage of the UD engine) and no dynamic lighting – both systems that I’d expect to be difficult with this kind of tech. I want to hear about both, but I don’t think that they’re deal-breakers – I don’t mind the idea of a static world with baked lighting for my polygonal characters to move around in. If it’s as efficient as they say, maybe I can even retain a reasonably large poly budget to render my characters with (so long as I have a good depth buffer).

    3. Yeah, the guy sounds pretty arrogant – I think that’s why most people responding are being fairly pissy about this. However, I don’t see any reason to think that his claims are exaggerated – he doesn’t even claim that this system is new, as in the latest video he talks about it being used in medicine and other applications. Lacking in humility maybe, but “snake-oil salesman” is pretty harsh.

    • SirKicksalot says:

      They’re using voxels. This post made in 2008 explains why they decided to avoid calling them voxels and that it’s not yet a practical solution for replacing polygons: link to

      Looks like not calling them voxels eventually bit them in the arse, not that other people are working with them…

  33. ericks says:

    Notch sounds jealous.

    This reminds me of a very interesting video on how FUEL did it’s geometry for it’s multiple kilometers of land that I unfortunately can no longer find.

  34. Lusit says:

    I talked about this with a coworker of mine (we’re both in IT) a couple months ago, maybe a year now. We figured that they were obviously not focusing on the downsides, mainly that they were static environments that wouldn’t lend well to animations and gameplay.

  35. ScubaMonster says:

    That screen cap of the euclidean engine looks pretty bad. Hope graphics can look better than that.

  36. GallonOfAlan says:


    “It feels like back when people hated on polygons and kept playing their fast and beautiful 2D games. ”

    The difference is that back then it was fairly obvious that a hardware-assisted polygon approach was entirely possible within a short timeframe.

  37. Davie says:

    I KNEW IT WAS TOO GOOD TO BE TRUE. You tell them, Notch.

  38. Detrian says:

    I don’t think Notch gets an opinion when his own block based game runs like absolute garbage.

    • skinlo says:

      Runs fine for me, maybe its your computer??

    • Snuffy the Evil says:

      I’ve never had a problem with it. Defrag your hard drive, or something.

      If anything, I would say Minecraft’s block-based world is exactly why Notch is allowed to comment on it. Minecraft isn’t rendered with voxels, per se, but I’d say many of the same concerns and limitations apply.

    • CMaster says:

      For the level of graphical fidelity, Minecraft performs very poorly on given hardware. On a friend’s laptop that can play TF2 and Portal happily enough, MC is barely playable for example.

      But thats a result of Java, the flexibility of the world and being in development.

      The “maybe it’s your computer” line by the way is always one of the stupidest responses ever. It’s not cool to pick on people just because they don’t splurge money on their PC every 6-12 months..

    • pepper says:

      cmaster, the comment that may be its your computer has nothing to do with picking on people but making sure the obvious gets ruled out. And I doubt in this day and age many people throw money at their computer every year, its not really needed anymore.

  39. MarkN says:

    The thing that bugs me most about this is that it looks bloody awful. They’re trying to sell a graphics engine that “looks 10, 000 times better” than what we have now and yet what they’re showing looks very much worse. There’s no denying it’s higher detail. It still looks rubbish though.

    They admit they’re not artists though. Which is good. So why haven’t they hired one? A rookie artist could stick a single yellow light in that scene and make it look about six times better immediately*.

    *assuming their engine incorporates single yellow light technology.

  40. PeopleLikeFrank says:

    Hm. So the last time this was in the public spotlight, it was identically debunked. (A clever RPS commenter pointed out that the memory requirements would be several orders of magnitude beyond what is currently available; animation would be impossible without CPU power of a similarly outlandish capacity; ultimately it’s nothing new.) The exact same arguments were made by apologists. (“It has a search algorithm!”, “Use compression!”, “Maybe they really do have some awesome secret sauce that lets them do the physically impossible!”). And those arguments were shown to not solve any of the problems presented. (“Yeah, so does any graphics engine in existence.”, “No compression could possibly make up the difference.”, “Sure, I’ve got a bridge you might be interested in buying as well.”)

    Oh well, guess we get to look forward to it all again whenever they need a new infusion of cash.

  41. Phandaal says:

    Wow, what a dumb move by Notch.

    He either comes off as a haughty asshole if he’s right, or as a clueless stickybeak if he’s wrong.

    Either way, Notch should stick to what he’s good at: ripping off other people’s ideas for money.

    • Dominic White says:

      So he’s an asshole if he’s right, and he’s an asshole if he’s wrong, and even if neither is the case, then you think he’s an asshole anyway?

      Why, that sounds completely reasonable and not at all biased! Thank you for your contribution!

    • Snuffy the Evil says:

      “Notch, your game sucks! Put this mod in Minecraft!”

      *puts in minecraft*

      “GOD, you unimaginative PIG”

      Damned if you do, damned if you don’t. I don’t envy Notch, that’s for sure.

    • Pointless Puppies says:

      I’m sure if he hadn’t commented on this you would’ve had SUCH a better opinion of the man.

      You know, considering how you think all he’s done is “rip off someone else’s idea”. Sounds to me like you had the insult at Notch at the tip of your tongue and just used this as the vehicle for your diatribe.

    • Sheve says:

      “I don’t envy Notch.”
      Screw that, I envy the hell out of him. Bastard got voted the best video game dev over Bioware. HE’S MORE POPULAR THAN BIOWARE. Let that one sink in. Then realize he’s made a crapton of money off of Minecraft.
      Yeah, I would be willing to put up with a few internet blowhards for that.

  42. ziusudra says:

    Remember Outcast. Great game.

  43. cday130 says:

    Update: Notch just farted

  44. teh Gibson says:

    I read some comments (if not said here, it was definitely said in the linked post about this tech) saying that the graphical improvements probably aren’t of much worth, be it because of hardware limitations, increased development cost, or whatever.

    I personally think that if one could apply proper physics to these particles and use that in conjunction with the current poly rendering model could allow for some amazing game play mechanics. I’m thinking Minecraft-esq scenario where one can dig actual holes by displacing certain atoms, as an initial example. Or some nice destruction effects where rubble piles up as it should, perhaps the bond strength between atoms could be defined for different materials, requiring different forces to break them.

    These examples are just off the top of my head and, for destruction at least, there are good, less resource intensive systems for representing such things. However, I still reckon game play is where the real benefit of this technology lies for gamers.

  45. Hematite says:

    Just wanted to drop in a quick note – only glanced at the various discussions.

    There’s a cool trick you can do with texture generation where you procedurally generate more detail as you zoom in – quite likely what they’re doing. Also, they seem to just have a rendering engine not a voxel/physics engine which would fit this idea.

    Think of it as being like Minecraft’s map generation, except instead of generating more terrain as you move laterally, it generates more detail as you zoom in. You can use some fairly specific parameters at normal zoom level to make the game world, then effectively use ‘noise’ to add detail as far as you care to zoom in – like that bit where they look at some dirt. All the zoom stuff is effectively randomly procedurally generated, but you can re-generate the same thing each time you look at an area because you use the same random seed.

    Or also like “Inside a Star Filled Sky” if you’ve played it.

    This sidesteps the memory requirements for storing the ‘whole island’, because you just store the high-level description and then use an algorithm to generate the specific detail as it’s required. It also makes it damn hard to do any animation or permanent modification to the world since everything is transient and disappears when you stop looking at it (even more than is usual in video games).

    I don’t really know about the issues around rendering this stuff though, I’m more on the information theory and dynamic computation end of it.

    P.S. I think there’s also a cool result of this procedural detail generation where the rendering effort is proportional to the number of pixels on the screen, not the number of polygons in the scene. If you have a hundred entities a long way away and one up close, I think a normal rendering engine will have to render 101 x entity_polys polygons, whereas a procedural generation can be lazy about the far away enemies and only generate detail for the very few pixels they cover, then spend most of the rendering budget on the identical but much closer entity which takes up half the screen. This means you could have heaps of enemies on screen, but also have them look good up close.

  46. Chikiwikie says:

    Notch discussing graphic engines…. yeah … right…

  47. propjoe says:

    Seems like an engine like that, if it were possible, would put a lot of extra burden on the art team. Now every tree, rock and bit of ground has to be individually modeled and textured? Yeah, that’s going to happen. Also, is there a reason why their demo map essentially looks like Minecraft? If I want to be impressed by some world graphics, I want some rolling hills, jagged cliffs and stone arches. I want to see glaciers.

  48. Schwolop says:

    WTF Notch? How did you fail to miss the most obvious point. Oct-trees… You know, those things that start with one big cell containing everything, then subdivide into eight sub-cells and keep doing so until no cell has more than one entity in it. Free space takes up zero memory.

    Add in someone else’s point about not storing the internals of objects and you’ve reduced the memory requirements again.

    • Doggy says:

      Oct-Trees would greatly reduce the number notch said, but that number would still be greatly too big.

      Another effect of introducing oct-trees is that they could reduce the speed of the search algorithm, since the hierarchy would need to be traversed.

  49. Pointless Puppies says:


    The only people jealous here are the Notch haters. The man is entitled to his opinion, and he has formed it through what he believes is accurate and comes out with the conclusion that the men simply want funding, hence the “THESE ARE INDIVIDUAL GRAINZ OF SAND!!!” Billy-Mays-speak. And he’s posting it on his own personal blog.

    It’s not his fault if the entire internet is in love with the man, so naturally the story’s going to run in every gaming news site. His opinions and his conclusions. Don’t like them? Go make your own blog and post it there. Don’t try to dismiss the man’s opinion simply because he has a bigger audience than you.

    Jesus. There’s a whole BUNCH of other people on this and all the other sites that are covering this story that say the exact same thing: this tech reeks of a scam and it doesn’t sound practical in the least bit, and they’ve all most likely accomplished far less than what Notch has. Are all the Notch haters going around saying these people have “no room to talk” as well? Are they saying these people are “jealous” too? Or does that only apply to Notch because he’s Notch?

    • qwiggalo says:

      Nobody is disagreeing with Notch we’re just stating the irony that he is talking down to this, but he is a shit programmer.

    • Splynter says:

      I’ve heard this repeated over and over again, but I’ve yet to see anyone provide evidence beyond “someone else says so, so it must be true”. His engine apparently runs badly, but I’ve seen people argue that this is all down to using Java. Why exactly is Notch a ‘shit’ programmer?

    • Batolemaeus says:

      Notch is far from a bad programmer. People simply have never coded in their life and assume perfectly optimized code and design just appear out of nothing in an organically grown codebase.
      Minecraft was always built in a rush and it shows. He’d have to refactor literally everything and tidy up the design from the ground up. Can’t fault him for not doing that, it’s tedious and absolutely not fun.

  50. qwiggalo says:

    Euclideon has a 2,000,000$ AUD grant from the Australian government. Discuss.

    link to