Short Tech Interview with John Carmack and Tim Sweeney

Brent_Justice

Moderator
Joined
Apr 17, 2000
Messages
17,755
posted on the front page

http://www.beyond3d.com/#news27836

some pointed tech questions here

Seems Tim agrees with the ALU increase over TMUs, though what ratio is best is debatable, seems he also will be using Vertex Texture fetch sometime in the future, doesn't mention R2VB.

John seems to be more on the fence, not really stating which side he is for.

I also like what Tim talks about at the end, letting IHV's play with a 'benchmark' months before game release so they can optimize and have good gaming perf in drivers when the game ships. I like this idea, I hope more game devs pick up on it.
 
Sooo, that was short and kinda vague.

Other than the new benchmark.

They basically agree with ATI's directioning?

Wish Carmack would talk more about ID's next generation game. They usually keep things very quiet up until near the unveiling. I doubt we'll see an early preview of it like we did with Doom3 at Macworld w/ GF3 how many years and graphics card generations before release?
 
Is it me or did that not clear up anything as far as VTF and ALU/TMU ratio?
 
Thats it! I'm calling shens

Since when does Carmack respond without paragraph after paragraph of information full of his idea or opnions on some type of rendering method or whatever. Confuses me with the amount of information he's talked about rather than that lack of...much of anything at all.
 
PRIME1 said:
Is it me or did that not clear up anything as far as VTF and ALU/TMU ratio?
Neither of them can clear up VTF (WHQL waiver or not). MS has been mum.

The trade-off ratio is going to vary game to game.
 
Menelmarar said:
Since when does Carmack respond without paragraph after paragraph of information full of his idea or opnions on some type of rendering method or whatever.
Read the next issue of PCGamer UK. :p
 
RoffleCopter said:
Sorry for the noob question, but what's R2VB?

Render to Vertex Buffer, a feature ATI has implemented in their X1000 series that is the equivalent of Vertex Texture Fetch.
 
You will never get a straight answer on ATi features from id, at least not in public.

Bad blood I tell you...bad blood.
 
You will never get a straight answer on ATi features from id, at least not in public.

Nope. For straight talk about ATi products, you need to talk to Valve :rolleyes:

I can't be too confident, but I think Sweeney is very close to surpassing Carmack in overall prowess, although both are kings in the realm of realtime rendering. Very surprised and happy about the UE3 early benchmark for IHVs.
 
phide said:
Nope. For straight talk about ATi products, you need to talk to Valve :rolleyes:

I can't be too confident, but I think Sweeney is very close to surpassing Carmack in overall prowess, although both are kings in the realm of realtime rendering. Very surprised and happy about the UE3 early benchmark for IHVs.
Carmack has jsut been focusing his attention in other areas of technology. Space Flight? Java cellphone games? I mean wtf.....
 
Back in the day Carmacks engines where the only game in town. Almost every developer clammored over them. But other developers improved greatly on their engines and now that Carmack is on pretty much even ground with the likes of Valves Source engine and the upcoming UE3 engine, Carmack does have the clout in the industry that he used to.
 
Space Flight? Java cellphone games? I mean wtf.....

Hey, Armadillo Airspace is/was an important venture. And programming cellphone games is child's play for someone like Carmack. How long do you think it took him to write the renderer/engine for Doom RPG? Not too long, I'm willing to bet.

We'll be impressed with what he's been working on once it's revealed. We always are.

On the subject of Unreal Engine in general, I do think it's a more viable platform. Very solid, intuitive developer tools and good support. You won't really find the same when you license an engine from Mesquite.
 
Menelmarar said:
Carmack has jsut been focusing his attention in other areas of technology. Space Flight? Java cellphone games? I mean wtf.....
:D lol
 
phide said:
Hey, Armadillo Airspace is/was an important venture. And programming cellphone games is child's play for someone like Carmack. How long do you think it took him to write the renderer/engine for Doom RPG? Not too long, I'm willing to bet.

We'll be impressed with what he's been working on once it's revealed. We always are.

On the subject of Unreal Engine in general, I do think it's a more viable platform. Very solid, intuitive developer tools and good support. You won't really find the same when you license an engine from Mesquite.

I dunno man I'm less and less impressed with Carmack's recent work. Doom 3 was a big letdown and the engine isn't all that revolutionary either. I think Sweeney is definitely ahead of him now in terms of making great game engines for the masses.
 
5150Joker said:
I dunno man I'm less and less impressed with Carmack's recent work. Doom 3 was a big letdown and the engine isn't all that revolutionary either.

Doom 3 is just such a solid engine, though. Memory management is fantastic. Stability is absolutely, positively rock solid. I've tried to crash Doom 3 using traditional means - I haven't managed to make it happen yet. The renderer is also pretty quick, though I still think there's a little room for improvement. Not sure about that one, really. I don't think we need to talk about netcode - it's just not there, but that wasn't ever the point of Doom 3.

FarCry did the graphical tricks Doom 3 did, sure. The renderer's solid, but the rest leaves much to be desired. Memory management is atrocious. It's as if FarCry doesn't do memory dumps at all. Stability is, to this day, a little iffy. Patches fix bugs and create a host of new ones. It seems like Crytek really isn't too keen on what's going on with their engine deep down.

If you want to talk about Doom 3's gameplay, this is really just the wrong topic.
 
phide said:
Doom 3 is just such a solid engine, though. Memory management is fantastic. Stability is absolutely, positively rock solid. I've tried to crash Doom 3 using traditional means - I haven't managed to make it happen yet. The renderer is also pretty quick, though I still think there's a little room for improvement. Not sure about that one, really. I don't think we need to talk about netcode - it's just not there, but that wasn't ever the point of Doom 3.

FarCry did the graphical tricks Doom 3 did, sure. The renderer's solid, but the rest leaves much to be desired. Memory management is atrocious. It's as if FarCry doesn't do memory dumps at all. Stability is, to this day, a little iffy. Patches fix bugs and create a host of new ones. It seems like Crytek really isn't too keen on what's going on with their engine deep down.

If you want to talk about Doom 3's gameplay, this is really just the wrong topic.


Well phide, I don't pay attention to memory dumps when running a game. Far Cry was rock solid for me and never once crashed - it's graphics were fantastic as well. When I played Doom 3 I thought, "this is what we've been waiting for?" when I saw the graphics. It was pure shadows and granted it does that well but that's about all that gave me a positive feeling about the engine. Maybe I'll change my mind when I see Quake Wars.
 
Not to take anything away from Tim (he's a master himself) but John is in a world all by himself. You people seem to forget that Carmack pushed the envelope on technology long before others caught up to him. It's much more standardized today. But years back, it wasn't so. I mean look at Quake 3....the game still had a polished look some four years after release. That's how far John pushed technology. John pushed the Doom 3 engine before anyone else with the realtime shadows and lighting and it's becoming standard now, as an example. The guys been coding for so long....if he want's to venture into other realms, whose to say he can't? Just remember each game engine he's put out and how it pushed the envelope ahead of any other games at the time. That's his legacy.
 
Correct me if i'm wrong here but I'd imagine when unified shaders release VTF will be a given.

I'd imagine the unified cards are going to be similar to the current cards with the vertex processing portions chopped out. Seeing as how a pixel pilelines is a superset of a vertex pipeline it would make sense to cut the vertex and drop in a bunch more "pixel" pipelines.

Seeing that VTF is nothing more than texture lookups in a vertex pipeline why waste time adding them in to be replaced less than a year down the road? R2VB seems like a better method to hold things over until that time. It'll also likely be similar to how a unified shader would approach the situation by using the pixel pipelines to do the processing. And in case it wasn't obvious pixel pipelines seem to be rather good at doing texture lookups.

Also seeing how ATI's branching/threading only affects their pixel pipeline it seems like this is the direction they've been moving.

In response to the article. Talking about VTF and which they prefer seems to be a minor point since it's implementation is sort of a given with unified pipes. The ALU/Tex ratio seems to be the only major point that would be of interest. I'm not aware of any other major features that either IHV is pushing to implement so it's all there really is to talk about.
 
BBA said:
You will never get a straight answer on ATi features from id, at least not in public.

Bad blood I tell you...bad blood.
Then why won't Sweeny say anthing?

I just don't think it's in anyone's best interest to say Card-x is better than Card-y.

I bet if you got them drunk and asked them, they would bitch about both cards.
 
5150Joker said:
. Doom 3 was a big letdown and the engine isn't all that revolutionary either.

You know...it's easy to confuse Doom3 being a bad game if you feel that way and the power the engine has. Quake 4 is just AWESOME looking (They turned on the lights ;) )
Whether you hate quake 4 or not too *grin* Prey is also just looking to be incredible.
 
Back
Top