Quakecon 2008 keynote notes
- Thread starter silhouette
- Start date
Has Carmack ever professed 'godlike' skills? If not, isn't it a bit much to demand such things of him? You can't deny the progress the guy's made in getting a fabulous engine driven by new tech of his invention working across multiple platforms...
I think he would like it "exclusively" on the Cell, i think it would be an interesting experiment as well. But he usually don´t like to spend time on something he doesn´t find interesting. And in this case it seems it annoys him more than it intrigues him.
Panajev2001a
Veteran
I think that if Carmack got nice funding to make an exclusive PS3 title he would push it as far as it could go, I think that it is the practical side of him talking that, considering a certain standard of graphical fidelity, a certain deadline, and a budget that has to fit for a multi-platform release, is influencing things. He knows that there is a way to make PS3 sing, CELL especially, but unless you are given a very big budget it might be hard to follow that path as well as tuning very well for other paths too. Still, they went ahead and created id Tech 5 and are reaping the fruits with RAGE now, so it is hard to tell them they are doing it the wrong way. Still, he has said many times that he prefers a SMP approach much more than a heterogeneous one.
I'd like to see a CELL vs regular SMP panel with Carmack and John Olick (ex ND guy, both at id), Mark Cerny, and Mike Acton... with a polite and technical discussion following (hey, a nice GDC idea
?).It's not fair, but then we do treat his every pronouncement like the word of god. Like I've said earlier, we're not in 1999. Back then, if he said that down was the new up you'd know that most upcoming games would have inverted gravity, and graphics card companies would change their drivers to better implement upside down graphics (anyone remember the quaek3[sic] mip-mapping optimizations?).
Meanwhile, not only has id struggled to release good games (though I suppose Orcs and Elves counts) for quite a while (if Doom 3 counts, then it's been nearly 4 years, otherwise 7-9), their engines haven't had the importance they used to. It's not that there are pretenders to the throne, the throne was lost entirely: Epic bet on consoles and it paid off, they have a much greater influence in the gaming world than Carmack has right now.
Maybe that'll change with idTech 5... but how's this: we let him release a multi-platform game on it, first, before taking every single word of his as gospel? I mean, we haven't even seen gameplay footage of Rage and even people on B3D are going 'that video shows how the game's gonna look!'.
I'd like to see a CELL vs regular SMP panel with Carmack and Jon Olick (ex ND guy, both at id), Mark Cerny, and Mike Acton... with a polite and technical discussion following (hey, a nice GDC idea
This! (and i took the liberty to subtly fix your post
)Something that would also be great, although I don't know what the chances are, would be a B3D interview of Carmack and especially Olick, now that he's working in a multiplatform house.
Technical debate among those who have the highest profile in game technology 'pronoucements' is sorely, sorely lacking.
Has Carmack ever engaged, recently or otherwise, in a public debate on the technology issues?
I think it'd be healthy, everyone might learn new things. I don't think it's healthy to accept anything anyone says too easily, even among those smarter or apparently smarter than their audience. Best to put them all together (them being the smartest people in the field) and let them shake out an even higher and better level of understanding and insight. It's easy to make convincing arguments in the absence of challenge.
It could also be uncomfortable for some observers though who like to think in black and white, when they realise the frequency of opinion and perspective in matters, even technical matters, rather than absolute truth.
Has Carmack ever engaged, recently or otherwise, in a public debate on the technology issues?
I think it'd be healthy, everyone might learn new things. I don't think it's healthy to accept anything anyone says too easily, even among those smarter or apparently smarter than their audience. Best to put them all together (them being the smartest people in the field) and let them shake out an even higher and better level of understanding and insight. It's easy to make convincing arguments in the absence of challenge.
It could also be uncomfortable for some observers though who like to think in black and white, when they realise the frequency of opinion and perspective in matters, even technical matters, rather than absolute truth.
Carmack leaved assembly behind many, many years ago. There's something like 'we don't code hardware overlays in assembly anymore' or so in his last .plan updates back in 2000-2001 or so. He doesn't like to waste time on unimportant stuff like hunting bits in low-level code if it takes away developer time from large scale issues, he's always been like that. Back in the Quake days everything was a lot more simple, and even then most of the P6 optimizations were done by Abrash AFAIK.
Yeah, exactly what I've meant.
Also note that he's talking about issues that have led to MegaTexture's full implementation in Rage today. His mind has been on it for all this time...
And I think that solving this problem is is far, far more important than messing around in ASM.
The id site lists 60 employees, that's an all time record. Even though some of them are working on the mobile games, it's still a decent sized team. And they're hiring for at least a half dozen positions.
Probably? Considering there are PC's GPU's which can already achieve this very easily I would think its more of a definate
In a year from now when we might see Doom 4 we should be expecting PC GPU's to be pushing the game at 1080p, 16xAF, 8xMSAA, 60fps and all with much higher visual fidelity than the console versions..
My desire to see him going head to head with the Cell (which is a dream since it wont happen for all the mentioned reasons) isn´t about him putting the assembler to work, again. The interesting thing would be how he would end up coding a PS3 only graphics engine, from the ground up.
It´s fair to think that he can´t invent the wheel again and that it wouldn´t be anything special compared to the current engines, but it would be interesting to see what he would come up with
There has and still is alot of talk of how games (and all software i guess) has to take care of more than one core, and in the PS3 case it´s even more complicated. John Carmack was one of the first to take baby steps in that direction with the famous r_smp option.
IMHO, Carmack's main skills haven't really been invention, but creative problem solving through gathering vast amounts of knowledge, and highly efficient implementation.
We've found existing research even for MegaTexture's first (Quake Wars) and second implementations, not to mention stencil shadows or BSP trees or raycasting. It's mostly academic stuff, results of a lot of people's hard work and ideas, sometimes up to a decade old.
But it does take Carmack to think, what if you only do raycasting for every column - and you get Wolfenstein; what if you precalculate visibility and store it in a BSP - you get Quake, what if you do the stencil stuff in reverse - you get Doom; and I guess you see where I'm going with this. And then he polishes the technology enough so that it can run fairly well on the current generation hardware, and sets the way for the rest of the industry for years.
True, stencil shadows did not take off that much for lack of efficiency and advancements in shadow buffers; but look at normal mapping, which was once again academic research applied to game development. Everyone was floored on the first Doom3 related test stuff, and it still lives with us in practically every game.
And he is already looking into the future, and it's more about high level stuff, virtualizing geometry and such. He couldn't care less for the underlying hardware as long as it's fast enough, and he leaves optimization to the programming team and system specialists.
So expecting anything extra on the PS3 from him is kinda pointless - what he's into is platform independent, and it's the right thing to do
We've found existing research even for MegaTexture's first (Quake Wars) and second implementations, not to mention stencil shadows or BSP trees or raycasting. It's mostly academic stuff, results of a lot of people's hard work and ideas, sometimes up to a decade old.
But it does take Carmack to think, what if you only do raycasting for every column - and you get Wolfenstein; what if you precalculate visibility and store it in a BSP - you get Quake, what if you do the stencil stuff in reverse - you get Doom; and I guess you see where I'm going with this. And then he polishes the technology enough so that it can run fairly well on the current generation hardware, and sets the way for the rest of the industry for years.
True, stencil shadows did not take off that much for lack of efficiency and advancements in shadow buffers; but look at normal mapping, which was once again academic research applied to game development. Everyone was floored on the first Doom3 related test stuff, and it still lives with us in practically every game.
And he is already looking into the future, and it's more about high level stuff, virtualizing geometry and such. He couldn't care less for the underlying hardware as long as it's fast enough, and he leaves optimization to the programming team and system specialists.
So expecting anything extra on the PS3 from him is kinda pointless - what he's into is platform independent, and it's the right thing to do
It was the first thing he commented on - how they have always been a small company, but recently they have increased in size considerably.
silhouette
Regular
I believe he has never worked on PS3. I guess he is handling PC and 360, and some other ppl in the company have been porting the engine to PS3.
Not this year, but in last year's keynote, he told that he still enjoys programming by writing directly to graphic hardware registers as in the old days, but only when it is managable to do that (like in DS). With current GPUs complexities, it is not practical anymore.
Btw, he mentioned id Tech 5 engine has 10M+ lines of C++ code. Good luck to anyone who wants to do similar -incredbile- stuff in assembly.
Not this year, but in last year's keynote, he told that he still enjoys programming by writing directly to graphic hardware registers as in the old days, but only when it is managable to do that (like in DS). With current GPUs complexities, it is not practical anymore.
Btw, he mentioned id Tech 5 engine has 10M+ lines of C++ code. Good luck to anyone who wants to do similar -incredbile- stuff in assembly.
I don't believe at any point he's said he wasn't going to develop on the PS3 or that he wouldn't be pushing the Cell as hard as possible to achieve the results. This makes me curious as to why so many people class his comments as whining.
Are we so used to developers being so diplomatic when they speak that when someone like Carmack speaks so bluntly, everyone thinks he's being so negative?
Are we so used to developers being so diplomatic when they speak that when someone like Carmack speaks so bluntly, everyone thinks he's being so negative?
Interesting. You're saying that unlike the uber-geek impression the internet has often presented Carmack with, he's actually a proper software engineer, working on the software theory and implementation side regardless of hardware. This would definitely make keyn's disappointments misplaced, as he'd be expecting something other than what the man is really about.
He's certainly skilled in low-level work, but I believe that he now considers it a waste of time compared to high level research. At least that's my impression based on Abrash's articles on the development of Quake.
Almost every chapter ends up describing some complicated problem (now obsolete thanks to 3D hardware acceleration), and then Carmack trying wildly different solutions until he arrives at one that works.
Also remember that between Quake3 and Doom3 he's written like a dozen test engines. Ain't exactly bit hunting either, and I'd say that a lot of his current work has its root in that short research period.
Almost every chapter ends up describing some complicated problem (now obsolete thanks to 3D hardware acceleration), and then Carmack trying wildly different solutions until he arrives at one that works.
Also remember that between Quake3 and Doom3 he's written like a dozen test engines. Ain't exactly bit hunting either, and I'd say that a lot of his current work has its root in that short research period.
Similar threads
