Carmack on "MegaTextures
by Ian Winter · in Torque Game Engine Advanced · 05/15/2006 (8:15 am) · 20 replies
I'm not sure if this is the best forum for this, but I'm just wondering how Carmacks new (?) technique for large terrains differs from the upcoming Atlas implementation of texturing landscapes and what the advantages/disadvantages to both methods are?
The article is here:
http://www.gamerwithin.com/?view=article&article=1319&p=1
Perhaps I've missed something but Carmack suggests this as something new but it doesn't really sound effectively any different to what we were due to see implemented in Atlas in practice, is it just the underlying theory that's different?
Thanks!
The article is here:
http://www.gamerwithin.com/?view=article&article=1319&p=1
Perhaps I've missed something but Carmack suggests this as something new but it doesn't really sound effectively any different to what we were due to see implemented in Atlas in practice, is it just the underlying theory that's different?
Thanks!
#2
05/15/2006 (10:59 am)
Nothing of this is new, not Atlas, not Carmack's implementation. There's designs and papers available freely which you can study and adapt, I think chunkLOD is one of them. 
#3
05/15/2006 (3:54 pm)
Not sure where you guys get the impression he's the only one doing this or claiming that he is, I sure didn't get that impression from reading that article : after all, most of the techniques we use in real time rendering have been around for a long time, if not very long, long, long time (e.g bumpmapping arks back to the 70s, as a lot of other techniques)
#4
05/16/2006 (2:57 am)
Well the article calls it "new MegaTexture technology", new infers that it's something new no? I'm not trying to slate Carmack here or anything, it's just that reading the article gave me the impression it wasn't actually anything new at all despite the article claiming that it was. If there is something new here I was just intrigued to see what it is and how it differs from Atlas. Partly what had me intrigued is Carmack's claim that mod developers wont be able to make much use of this because of the workload, whilst I agree mod building has become much harder I think it's still possible, likewise game development with Torque is possible by indies. 
#5
05/28/2006 (12:40 pm)
"new MegaTexture techonolgy" is like saying "new Atlas technology" or "new TSE technology". They're not saying that the theory behind the technology is new, just that the implementation is new. i.e. it's new to id Software and the Quake engine, not some groundbreaking new world-changing algorithm.
#6
05/31/2006 (8:22 am)
The times where ID software made groundbreaking games are kinda over. That was in the wolf3d/doom time.
#7
06/12/2006 (1:22 am)
The new direction isn't graphics or rendering, it's interactivity. this is why the wii is poised to take over video gaming soon.
#8
06/12/2006 (6:21 am)
Aye, I have to say I REALLY want to get my hands on a Wii (no seriously, no jokes please :p) it does look good fun without the need for high end graphics ;) Hopefully there'll be an XBox 360 price drop around the PS3/Wii release also as I think I'll grab one of those two! Wont touch the PS3 with a barge pole though, never been a Playstation fan.
#9
Still to be respected but with a grain of salt..
06/13/2006 (9:37 pm)
This is the same guy that said DirectX was and would be a complete failure....Still to be respected but with a grain of salt..
#10
06/14/2006 (7:35 am)
At the time he said it, DirectX kinda was. But when the parent company has as much money as Microsoft, you can afford to burn the first 5 or so versions. 
#11
Most of the coverage I've seen of this "new technology" doesn't mention at all previous research. Just about every technique used in real-time graphics now, had it's start as a SigGraph paper by someone you'll never hear of like 10 years ago. Personally, I like to attribute credit to research, and make it known when I am using that research because I think it supports the people who do the research.
I don't think Ben has made any secret that Atlas is based on research done by someone else. This doesn't at all diminish the task of taking a paper and/or source code for a research project and turning it into a working implementation for a game engine. It is a massive undertaking to do research, understand research, and apply research. I just don't think it's very fair that Carmack is getting credit for this "mega texture" technology, and I think that the process of developing and adapting existing research for use in games is something that should get more media attention.
Let me say again, that I am not trying to diminish the accomplishment of John Carmack, or Ben Garney, or anyone who takes the time to do research and applies it. I'm also not trying to put any kind of fault on Carmack for answering the questions he was asked...I just really feel like those who do research should get proper credit, when it is due.
06/14/2006 (10:01 am)
This ticks me off, personally. I don't think the fault is Carmack's, though, entirely.Most of the coverage I've seen of this "new technology" doesn't mention at all previous research. Just about every technique used in real-time graphics now, had it's start as a SigGraph paper by someone you'll never hear of like 10 years ago. Personally, I like to attribute credit to research, and make it known when I am using that research because I think it supports the people who do the research.
I don't think Ben has made any secret that Atlas is based on research done by someone else. This doesn't at all diminish the task of taking a paper and/or source code for a research project and turning it into a working implementation for a game engine. It is a massive undertaking to do research, understand research, and apply research. I just don't think it's very fair that Carmack is getting credit for this "mega texture" technology, and I think that the process of developing and adapting existing research for use in games is something that should get more media attention.
Let me say again, that I am not trying to diminish the accomplishment of John Carmack, or Ben Garney, or anyone who takes the time to do research and applies it. I'm also not trying to put any kind of fault on Carmack for answering the questions he was asked...I just really feel like those who do research should get proper credit, when it is due.
#12
Whilst I realise now Carmack hasn't himself claimed that it's all his work and it was bad wording on behalf of the article, he's also not done anything to deny it either or credit previous developers who have done pretty much the same thing either. Him not doing that is a pretty bad move it does nothing to help his case on things such as the Creative patent suit with regards to Doom 3's lighting, Carmack previously claimed that he invented it himself seperately to Creative, Creative claimed otherwise.
06/16/2006 (2:16 am)
Maybe that's it Pat, perhaps the feeling is that Carmack with all his experience and knowledge should be capable of doing the research himself, not just implementing it from a paper?Whilst I realise now Carmack hasn't himself claimed that it's all his work and it was bad wording on behalf of the article, he's also not done anything to deny it either or credit previous developers who have done pretty much the same thing either. Him not doing that is a pretty bad move it does nothing to help his case on things such as the Creative patent suit with regards to Doom 3's lighting, Carmack previously claimed that he invented it himself seperately to Creative, Creative claimed otherwise.
#13
06/16/2006 (8:38 am)
MegaTexture sounds a lot more sexy than Atlas, maybe Atlas should be changed to MegaAtlas ;)
#14
06/17/2006 (10:30 am)
I totally disagree, Ian.Quote:
If I have seen further it is by standing on ye shoulders of Giants.
-Isaac Newton, Letter to Robert Hooke 15 February 1676
#15
All that said however, that quote also highlights where I take issue in the second part of my last post - that great people like Newton are willing to acknowledge that their work is built on that of others, something Carmack generally never seems to do. Then again, the irony is that that quote is a saying that dates back long before Newton so the fact that it's often mistakenly attributed to him is in itself perhaps a possible example of part of the problem that Carmack faces ;)
06/23/2006 (4:44 am)
Always been one of my favourite quotes so I guess in the first case you're probably right, maybe it's not such a bad thing - in fact looking back I'm not totally sure why I said that but then it was 10am on a workday so I'll still have been half asleep ;)All that said however, that quote also highlights where I take issue in the second part of my last post - that great people like Newton are willing to acknowledge that their work is built on that of others, something Carmack generally never seems to do. Then again, the irony is that that quote is a saying that dates back long before Newton so the fact that it's often mistakenly attributed to him is in itself perhaps a possible example of part of the problem that Carmack faces ;)
#16
I wrote him a thank-you e-mail when I shipped the first version of Atlas, too. :)
On the subject of credit, Atlas also owes a big debt to Tim Gift and Mark Frohnmayer, the authors of the Tribes 1 and Tribes 2 terrain systems, respectively. It's a lot easier to do a good terrain system when you have experienced guys with decades of experience writing terrain systems to help you out with the tricky bits. ;)
About the only thing Atlas didn't benefit from was all the academics out there implementing complex CLOD schemes (ie, ROAM). Sorry guys - you should find something more useful to research. :)
06/30/2006 (10:59 am)
Incidentally, geometry-wise, Atlas is based (increasingly loosely) on work by the illustrious Thatcher Ulrich - his chunklod technique was what sparked the basic geometry rendering architecture. The previous version had his name mentioned in every Atlas header comment block; the new version is much less directly based on his work so I removed the credit from the source files (although it is still in the docs, and I definitely owe him a beer ;).I wrote him a thank-you e-mail when I shipped the first version of Atlas, too. :)
On the subject of credit, Atlas also owes a big debt to Tim Gift and Mark Frohnmayer, the authors of the Tribes 1 and Tribes 2 terrain systems, respectively. It's a lot easier to do a good terrain system when you have experienced guys with decades of experience writing terrain systems to help you out with the tricky bits. ;)
About the only thing Atlas didn't benefit from was all the academics out there implementing complex CLOD schemes (ie, ROAM). Sorry guys - you should find something more useful to research. :)
#17
Carmack and id have always been class acts from where I'm standing, and I just don't get the same impression that he is claiming that he engineered this thing entirely on his own, just because he doesn't mention it in a magazine interview.
06/30/2006 (12:30 pm)
How many of you have seen the header files of this new engine, and can say that Carmack is not giving similar credit where credit is due?Carmack and id have always been class acts from where I'm standing, and I just don't get the same impression that he is claiming that he engineered this thing entirely on his own, just because he doesn't mention it in a magazine interview.
#18
As with this "megatexture", its basically a form of clipmapping as far as anyone can tell. I just wish he'd call a spade a spade. Making up a term like megatexture really does (I think) detract from the previous works.
I doubt he does it deliberately.
I think its a cool technique, but as has been pointed out in another forum. Look at the screenshots of EWQW which uses this and show me a shot that doesnt look pretty much the same as any and all other games shots that use splatting/blending/whateverelse.
What I think *IS* interesting, is his assertion that the toriodal memory management and fundamental freedoms of unlimited texture size does as a basic practice within a game. In one interview I read, it sounded like he was thinking of generalising the whole idea so that it could be used on all geometry.
06/30/2006 (12:57 pm)
Well, fundamentally Carmack is a great implementor, not a great inventor. All of his work has always been based on past papers. BSP tree's werent big news until Carmack made an implementation work and work fantastically well.As with this "megatexture", its basically a form of clipmapping as far as anyone can tell. I just wish he'd call a spade a spade. Making up a term like megatexture really does (I think) detract from the previous works.
I doubt he does it deliberately.
I think its a cool technique, but as has been pointed out in another forum. Look at the screenshots of EWQW which uses this and show me a shot that doesnt look pretty much the same as any and all other games shots that use splatting/blending/whateverelse.
What I think *IS* interesting, is his assertion that the toriodal memory management and fundamental freedoms of unlimited texture size does as a basic practice within a game. In one interview I read, it sounded like he was thinking of generalising the whole idea so that it could be used on all geometry.
#19
If it were JUST clipmapping, maybe he would. But the Megatexture technology encapsulates more than just one thing. Physical properties are also embeddable in the terrain, as part of the technology.
I'm not aware of many people or companies that don't attach codenames to their own implementations of things, when their implementation is different. If your implementation is an exact duplicate of something else, then maybe you don't have any business giving it your own name, but if it is just derived from something else, and uses a hybrid of techniques that are put together in a unique implementation, then there is nothing wrong with giving it a unique name.
And just because something looks like something else in a screenshot, doesn't mean it is the same thing.
06/30/2006 (1:12 pm)
Quote:
I just wish he'd call a spade a spade. Making up a term like megatexture really does (I think) detract from the previous works.
If it were JUST clipmapping, maybe he would. But the Megatexture technology encapsulates more than just one thing. Physical properties are also embeddable in the terrain, as part of the technology.
I'm not aware of many people or companies that don't attach codenames to their own implementations of things, when their implementation is different. If your implementation is an exact duplicate of something else, then maybe you don't have any business giving it your own name, but if it is just derived from something else, and uses a hybrid of techniques that are put together in a unique implementation, then there is nothing wrong with giving it a unique name.
And just because something looks like something else in a screenshot, doesn't mean it is the same thing.
#20
I have nothing but respect for John Carmack; he's a brilliant programmer. We'd be in a different place entirely right now without him. (Just in the Torque engine alone, there is visible evidence of the Quake engine's impact...)
I can't recall him ever saying "This is MY invention" with regard to anything he's done, with the exception of the ATI lighting/shadowing fiasco. And in that case, I believe he in all likelyhood did come up with that particular solution, but felt it would be a waste of resources to debate it with ATI. But even then, it was all clearly based on research that had been done by others. He is well known for taking concepts/techniques detailed in SigGraph papers (or from other sources) and refining/extending/optimizing them to apply to realtime rendering. And his solutions have generally proven to be better than competing solutions.
So with regard to the "new" Megatexture technology, I'm sure it's going to rock. It's definitely a step in the right direction; I would love to see something beyond the tiled/blended texturing that we see in engines right now.
06/30/2006 (1:27 pm)
MY $0.02 cents on this:I have nothing but respect for John Carmack; he's a brilliant programmer. We'd be in a different place entirely right now without him. (Just in the Torque engine alone, there is visible evidence of the Quake engine's impact...)
I can't recall him ever saying "This is MY invention" with regard to anything he's done, with the exception of the ATI lighting/shadowing fiasco. And in that case, I believe he in all likelyhood did come up with that particular solution, but felt it would be a waste of resources to debate it with ATI. But even then, it was all clearly based on research that had been done by others. He is well known for taking concepts/techniques detailed in SigGraph papers (or from other sources) and refining/extending/optimizing them to apply to realtime rendering. And his solutions have generally proven to be better than competing solutions.
So with regard to the "new" Megatexture technology, I'm sure it's going to rock. It's definitely a step in the right direction; I would love to see something beyond the tiled/blended texturing that we see in engines right now.
Torque Owner David Miller
Default Studio Name
Dave