Carmack's Hands On Impressions of Xbox 360 & PS3

[MonkeyLicker]

Rage, for example, looks remarkably good for being 60 fps and multiplatform title but I'd like to remind to everyone that years ago ID was stunning us with things we have never seen before in realtime, now we have ppl that thinks Rage does not look even that good compared to what we have already seen in other multiplatform games (COD4?).

Some people also think Killzone 2 came "very close" to the E305 CGI.
Point being, some people really have no clue.

 

[Subtlesnake]

My acknowledgment as to whether Carmack could "in fact program effectively for the Cell" is irrelevant because it was never a question to begin with.

It's relevant to establish what we're not arguing about! That was my point - if Carmack's skills aren't in question then the only thing left is his commitment/motivation. And assessing the latter isn't exactly easy.

And to lower you batting average even further I'm also going to disagree with you that I was questioning his commitment for developing for the Cell. After all, he's so committed that he even hired one of the more talented PS3 developers available - I consider that a commitment relative to the best interests of id.

Committed to developing for the Cell himself. I thought that would be clear - we are talking about whether he himself is prepared to adopt the appropriate programming techniques to effectively take advantage of the Cell, if necessary.

Yes, I'm sure he really contemplated re-writing his entire engine to make it more data centric just so that it would run well on the PS3.

That's not the context of the quote! He made those comments at Quakecon 2005 when he was only just starting the new codebase, and wasn't completely sure which direction to go in WRT to multithreading. The fact that he was considering the 'worker thread' style model indicates that he wasn't closed off to that methodology, contrary to your statements that:

"[Carmack won't] even contemplate switching to a data centric view of processing where the flow and packaging of data is just as important as the code processing it"

"a methodology that places so much importance on data will never permeate into mind of Carmark"

So he was prepared to implement such a system, and according to you has the skills to do this successfully, so what else is there to argue about? Unless, as I said, you refuse to take Carmack's words at face value.

But I'm doing all the hard work here. All I'm asking for is the evidence for your assessment that Carmack isn't prepared to consider this 'data centric' programming model.

Let's be realistic here - there was no way he was going to ever radically rewrite his engine code. If there was any contemplation then it was for reasons for not doing it as opposed to doing it. Once again, I'm not superimposing my beliefs, but rather just looking at the herculean task from a realistic perspective.

Calling it a herculean task doesn't make it so. And again, the comments were made at the start of development, where Carmack was looking at developing for both platforms. So the question is not 'would Carmack rewrite everything for the PS3' but rather 'would Carmack periodically bring the PS3 codebase up to date'? And when you look at it like that it seems a lot less herculean.

And the context for this is that Carmack committed to 60 FPS platform parity, so you'd think that he would do whatever is necessary to bring the PS3 version up to par. Even hiring another specialist programmer....

Well, anytime you need to re-architect an entire graphics engine it's going to take a lot more effort - this is pretty obvious. But, it's something that's done once, if done correctly, and optimized thereafter.

Isn't that the perfect argument for Carmack taking full advantage of the PS3?

As for speculation on how the engine was modified, well, once again since the PS3 developer was an EDGE contributer I would hazard to guess that the same techniques or a variation thereof were applied to the PS3 id Tech5 engine.

Doesn't EDGE involve "a data centric processing perspective that uses schedulers to parse out work"?. At least for geometry processing, animation, compression etc.

You seem to be saying this thread is pointless since we don't have enough context to weigh his feedback either way to begin with ? e.g., He could have done X but didn't do Y, or he may not have started on X even, or he may have accomplished everything he wanted but couldn't achieve the results he desired, etc. etc.

I'm not saying the thread is pointless, because there's plenty to talk about! But yeah, we don't know the nature of the problems Carmack's faced or how (if!) he's solved them, so we can't really judge the nature of his success with or commitment to Cell development.

 

[GreekDNA]

Some people also think Killzone 2 came "very close" to the E305 CGI.
Point being, some people really have no clue.

Came close enough to fool 99% of people out there that it was infact CGI

me including .

 

[nAo]

Some people also think Killzone 2 came "very close" to the E305 CGI.
Point being, some people really have no clue.

I think I have a clue (hey..I might be wrong!) and I don't find Rage visually stunning.
Doom amazed me, Quake amazed me, Doom 3 was still amazing me.. KZ2 made me jump on my chair (yep, it looks that good ), but Rage does not, if they wouldn't have told me about this new texturing tech I wouldn't have probably noticed the difference with other games/engines (but a 30 seconds clip is not really enough to give a clear judgment..)

Marco

 

[Xenon]

I'm not saying the thread is pointless, because there's plenty to talk about! But yeah, we don't know the nature of the problems Carmack's faced or how (if!) he's solved them, so we can't really judge the nature of his success with or commitment to Cell development.

OK, I will - this argument is pointless. I believe Carmack could care less about the PS3 architecture and has relegated all PS3 related tasks to his hand picked PS3 developer. You, on the other hand, think Carmack and the PS3 developer are tag teaming the PS3. Our opinions differ. Case closed.

 

[patsu]

I'm not saying the thread is pointless, because there's plenty to talk about! But yeah, we don't know the nature of the problems Carmack's faced or how (if!) he's solved them, so we can't really judge the nature of his success with or commitment to Cell development.

I don't know... in fact, I am somewhat confused. People might talk about plenty of stuff but what is there to surmise ? ...without acknowledging/knowing "the nature of the problems Carmack's faced or how (if!) he's solved them" and "we can't really judge the nature of his success with or commitment to Cell development".

All the above conjectures and reasonings would seem futile since no one can prove/disprove them without hard data. Even JC's technical comments and commitment would raise questions since everyone knew about his preferences and no one knows what _exactly_ went into RAGE.

I would have to agree more with Vysez/Farrid in this case. In my view, it would seem that this thread is a celebrity talk/gossip.

Or we could try to find faults in each others' arguments, which would be for the sake of arguing like what I do below:

That's not the context of the quote! He made those comments at Quakecon 2005 when he was only just starting the new codebase, and wasn't completely sure which direction to go in WRT to multithreading. The fact that he was considering the 'worker thread' style model indicates that he wasn't closed off to that methodology, contrary to your statements that:

"[Carmack won't] even contemplate switching to a data centric view of processing where the flow and packaging of data is just as important as the code processing it"

"a methodology that places so much importance on data will never permeate into mind of Carmark"

Just a minor clarification... 'worker thread' style model is different from 'data centric view of processing'. I am not sure how one can lump them under the same methodology.

Calling it a herculean task doesn't make it so. And again, the comments were made at the start of development, where Carmack was looking at developing for both platforms. So the question is not 'would Carmack rewrite everything for the PS3' but rather 'would Carmack periodically bring the PS3 codebase up to date'? And when you look at it like that it seems a lot less herculean.

Why less herculean ? Why can't it be a massive design and implementation headache initially _plus_ "periodically bring PS3 codebase up to date" thereafter, especially in the face of cross-platform development ?

I mean there is no indication either way regarding what they did and not do. Commiting to 60fps is great and I'm rooting for the developers, but we only have 1 good example so far.

 

[Darkon]

I stand corrected on the second SPE, however, I find it hard to believe they could run the system code purely on a SPE without any PPU involvement, so I still firmly believe that they reserve some percentage of the PPU.

Why do you find it hard to believe ? and it better not be something stupid as SPEs can’t run scalar code.

 

[Geo]

Why do you find it hard to believe ? and it better not be something stupid as SPEs can’t run scalar code.

Or what?

If the hostility in this thread doesn't dial back about 3 notches it's going to look like the Democratic Convention of 1968 in here, with me in the role of the Chicago Police and y'all in the role of the poor bloody protestors who never did anything wrong.

 

[MonkeyLicker]

I think I have a clue (hey..I might be wrong!) and I don't find Rage visually stunning.
Doom amazed me, Quake amazed me, Doom 3 was still amazing me.. KZ2 made me jump on my chair (yep, it looks that good ), but Rage does not, if they wouldn't have told me about this new texturing tech I wouldn't have probably noticed the difference with other games/engines (but a 30 seconds clip is not really enough to give a clear judgment..)

Marco

I was referring to it on a technical level. It's painfully obvious you've probably forgotten more about technology than I've ever known, so I'm not dumb enough to try to debate you on that.
We've really not seen enough of Rage to say that Carmack hasn't delivered. He always has in the past, so i think it's only fair he gets the benefit of the doubt at this point.

Came close enough to fool 99% of people out there that it was infact CGI

me including .

I guess it really depends on the circles you frequent. Most of the people I saw seemed to know better.
Rest assured it was nowhere near 99%.

 

[Fafalada]

I was referring to it on a technical level.

I'll have to agree with nAo here.
Tech just for tech sake is ultimately pointless (both from marketting and endproduct perspective). Gaming history is littered with examples of this, some of us learned it the hard way too.
These presentations are done to impress with what is shown, we all know things will change as products come closer to completion, but we're discussing what's seen here and now.

Mind you I am actually quite fond of the tech-concepts they are using (more so then I ever was of how Doom3 was designed), the tradeoffs in this case actually make a lot of sense to me for certain types of games.

And on the other end we still have static, flat 2D displays..

Personally I've been more impressed with the stuff I've seen in regards to 3d display tech over any controller innovations in past 15 years. But I am sure the former is also a lot harder to translate to consumer products.

 

[MonkeyLicker]

Faf, nAo

What would you say are the most technically and visually impressive games shown thus far? I know nAo prefers the high poly look of games like Resistance to games like Gears, but I'm curious to hear more about what you guys think.

 

[GreekDNA]

Faf, nAo

What would you say are the most technically and visually impressive games shown thus far? I know nAo prefers the high poly look of games like Resistance to games like Gears, but I'm curious to hear more about what you guys think.

Resistance fall of man

- 4x AA

- High polygon count with subtle normal mapping where its needed "cough gears cough"

- Constant 30fps with complex physics simulations , partical effects - special effects ,ateast 20 enemies on screen and individual AI patterns without a hickup

-Real time shadows

- Impressive geometry

Resistance does have its flaws don't get me wrong but its still a steller looking game , very underrated in this regard .

Heavenly Sword

- amazing animations

- sub surface scattering giving the character a natural look

- self shadowing ( as far as i can tell )

- amazing lighting

- Clean IQ

- high poly count

- impressive textures and again subtle use of normal mapping

Killzone 2

- Best IQ in a next gen game hands down

- amazing lighting

- self shadowing

- complex animations

- constant 30- 35 fps with tons of particals and effects

- excellent use of post processing effects giving the a game a pre rendered look

- Guns have little to no visible polygon edges/ with no visable aliasing - smart use of motion blue and post processing effects to achieve a cgi look to them

- complex physics system (death animations)


Crysis

- Amazing draw distance

- incredibly physics engine

- interactive world (cutting down trees and destorying houses etc etc in real time is an impressive feat )

- incredible water shaders

Cod4

- Great animations ( makes a next gen game imho)

- natural lighting

- subtle use of normal mapping but effective

- looked like it ran at a constant 30fps with alot of particals , physics , etc etc

- self shadowing

GT5

- incredible real time lighting and reflections

- real time shadows

- 60fps

- 360fps physics simulations

- impressive textures

Ratchet and Clank: Tools of destruction

- constant 30 fps with dozen of particals and special effects

- clever use of normal and bump mapping

- Impressive geometry

- high polygon count and fur shaders that "almost" rivals pixar quality

- incredible animations

Uncharted

- 4x AA

- Amazing textures

- incredible foliage

- complex water shaders

- 3000 layerd animations

my 2 cents...

 

[Gubbi]

OOE circuits take up more than half of a typical modern CPU. Consoles traded this space for more execution units and left the programmers responsible for writing efficient code. Adding OOE to the PS3 or X360 would cut their potential in half, in exchange for easier development...

Off topic, but....

No it doesn't. The old PPRO architecture used about 10-15% of it's core Si real estate for the ROB and the RAT. K7/8s uses less. The PPRO/P2/P3/Core architecture is the one that is extended into current Core 2s, where the number of ROB entries is more than doubled from 40 to 96, but at the same time other functional units have been beefed up considerably, so I'd take a wild guess and predict that the Si real estate used by the OOO apparatus is still around 10-15%.

PA Semi has a 3-way superscalar OOO PowerPC core, complete with Altivec execution units and all that only takes up 10mm^2 in a 65nm process. That's smaller than XCPU (doubling it's size for a 90nm comparison).

I think the reason XCPU and PPE looks the way they do is because of a general trend at IBM where they aim for 14 FO4 logic levels per pipestage and then go all out on clock frequency (just look at their brand new Power 6). In that sense I don't think that either the XCPU or CELL succeeded (ie. I think IBM expected both cores to clock higher given the current power envelope, or clock as it is now in a narrower power envelope).

Cheers

 

[[maven]]

This is really what has bothered me about Carmack's (and Newell's) comments over the last couple of years. parallelism is hard, it's annoying, it's fraught with peril. What use is there in complaining about it though?

I may well accept the inevitable, but that doesn't mean I have to like it (especially if it means more effort). Why should I be happy about additional problems (even if they're interesting ones) if I'm time-constrained enough as it already is?

That'd be my guess. But all you armchair coders obviously know better...

 

[pjbliverpool]

We've really not seen enough of Rage to say that Carmack hasn't delivered. He always has in the past, so i think it's only fair he gets the benefit of the doubt at this point.

Bare in mind that in the past Carmack has been targetting high end PC's. Rage is now targetting the lowest common denominator in consoles at 60fps. Thats a lot less power to work with so we can't expect him to impress as much as in the past.

That said, I personally think Rage looks as good or better than anything else out there (baring perhaps Crysis).

 

[Nite_Hawk]

I may well accept the inevitable, but that doesn't mean I have to like it (especially if it means more effort). Why should I be happy about additional problems (even if they're interesting ones) if I'm time-constrained enough as it already is?

That'd be my guess. But all you armchair coders obviously know better...

I'm a software development lead at a supercomputer institute. Having said that, carmack is a better coder than I am, but I find his attitude disappointing. I think this discussion is a lot more about attitude than it is about skill.

Nite_Hawk

 

[nAo]

I may well accept the inevitable, but that doesn't mean I have to like it (especially if it means more effort). Why should I be happy about additional problems (even if they're interesting ones) if I'm time-constrained enough as it already is?

I'd wait for the final game to see how multicores computational power is use, maybe the game is so simple that does not need any particularly powerful CPU.
That would be a good way to eliminate part of the problems with these CPUs.. if you are not interested in working on games that do something more with a CPU than submitting some thousands rendering commands per framne, handling controls and some cheap ai/physics than I guess you'd really get to hate these in order/painfully slow next gen consoles CPUs.

 

[Xenon]

In that sense I don't think that either the XCPU or CELL succeeded (ie. I think IBM expected both cores to clock higher given the current power envelope, or clock as it is now in a narrower power envelope).

So you don't consider a Cell BE clocked at 6GHz high enough?

Link

 

[AlphaWolf]

So you don't consider a Cell BE clocked at 6GHz high enough?

Link

I'd say you are a victim of selective reading. Clocking at a given power envelope was what he said.

 

[function]

I think the reason XCPU and PPE looks the way they do is because of a general trend at IBM where they aim for 14 FO4 logic levels per pipestage and then go all out on clock frequency (just look at their brand new Power 6). In that sense I don't think that either the XCPU or CELL succeeded (ie. I think IBM expected both cores to clock higher given the current power envelope, or clock as it is now in a narrower power envelope).

Given that the first set of leaked 360 specs (was it 2003 or 2004?) showed the CPU cores running at "3.5+ gHz", there could well be some truth to that.