John Carmarck bothered with Next gen MProcessor Consoles

[Farid]

http://www.computerandvideogames.com/news/news_story.php(que)id=102803

Carmack also used the conference call to launch a stinging attack on next-generation console development, blasting: "It's not a business model I like. I'm bothered by multiple processors; I would have thought people would have learned their less by now - [multiple processors] rarely get used as much and hinder development."

Speaking on the belief that next-gen architecture will be structured around more than one processor, he continued: "I'm not thrilled with these directions; we've had several generations showing [multiple processors] don't work."

http://www.homelanfed.com/index.php?id=21933

In a Q&A after the keynote, Carmack stated that he wondered why future game consoles like the PS2 and the Xbox 2 will likely use multi-processors, saying such technology has not worked well for consoles in the past and wondered why this way of working was back in vogue again from console makers. ( I think many comsole programmers agree on this point. )

Looks like J.Carmack is not enthusiastic about the BE and the XCPU 2

 

[Bjorn]

Looks like J.Carmack is not enthusiastic about the BE and the XCPU 2

And then you have to take into consideration that he is one of the few that has actually supported multiple CPU's in his engines.

 

[London Geezer]

And let's not forget that current PC architectures are nothing more than multiple processors in one die... VS, PS (which will be the same thing soon enough)... This guy's on a roll, first the "no more polygons please" thing, now this... Meh...
All i can say is, he knows a lot more about games development than i do, so... :|

 

[McFly]

But then what is he asking for? A single CPU with 100GHz and 10'000W? I don't see how anyone can have similar power with a single processor then what we could get with multi processors. (In the same timeframe of course)

Fredi

 

[London Geezer]

And it's worth remembering that apart from GPUs (which are already vastly reliant on parallelism and "multi processor"-like architectures), CPUs in the PC world will become multi-core soon enough (some already are), whether he likes it or not. Both Intel and AMD already made it clear that future architectures will be multi-core... So... (again) And that's ignoring Macs which already have G5 today, a double core CPU, which is often used in multi-processor setups. And how about all the dual Opteron and dual Xeon setups. Granted, they are all for workstations, but i feel such setups will ultimately make it into the high-end gaming crowd too one day.

 

[Simon F]

Just playing devil's advocate here.... how many of you guys who are being critical of Carmack, have written complex and big selling computer games?

 

[London Geezer]

Just playing devil's advocate here.... how many of you guys who are being critical of Carmack, have written complex and big selling computer games?

That's not the point, as i said, who am i to talk next to this guy, however i'm merely clarifying that, whether he likes it or not, everyone is going the multi-processor, multi-core route...

 

[sir doris]

from this i gather he is more concerned with having lower quality models and much higher quality effects (more rendering passes). I guess he feels the next gen consoles should be putting more effort into the graphics and rendering capabilities and matching this with a single core CPU. Have the CPU to run the game but use the GPU part to do the graphics and maybe physics? Just an idea.

 

[bleon]

Carmacks turf is in PC land, I dunno why he's so concerned about the way consoles are headed, I mean if he doesn't like it he can stick PC development. *scratches head*

 

[London Geezer]

Carmacks turf is in PC land, I dunno why he's so concerned about the way consoles are headed, I mean if he doesn't like it he can stick PC development. *scratches head*

The point is, he WILL get multi-cores and/or multi-processors in PCland too... Which is even more confusing when you think about it.

Of course this could be another ambiguous quote, like the "no more polygons" one posted around here. In which case we should know the whole story before commenting.

 

[Simon F]

Just playing devil's advocate here.... how many of you guys who are being critical of Carmack, have written complex and big selling computer games?

That's not the point, as i said, who am i to talk next to this guy, however i'm merely clarifying that, whether he likes it or not, everyone is going the multi-processor, multi-core route...

And have any of you tried programming multi-cpu systems and, by this, I mean >4 CPUs? (I have BTW - IIRC 48 CPUs). It is not easy. Not all applications map nicely to such a system.

OTOH, it is relatively easy for HW designers to bung down multiple cores and say that they have a high performance part. I would say that JC is lamenting this direction.

Of course, massively parallel systems have their fortes (and of course the pixel processing in a graphics chip fits this model), but just throwing CPUs at a problem might not be the best solution.

 

[Guden Oden]

Just playing devil's advocate here.... how many of you guys who are being critical of Carmack, have written complex and big selling computer games?

No doubt Carmack's a smart guy, but he's not a console developer EITHER, just like Sweeney isn't. The last console he ever did was for the fecking Atari Jaguar for chrissakes.

When he says multiprocessing for consoles don't work I do a big , because there's a multiprocessing console out there right now that happens to be the MARKET LEADER. So, what's he talking about, really?

I guess he's referring to the Saturn as the 'multiple generations that didn't work', or whatever the quote was, but the Saturn was just badly designed. Of course six processors are going to be a horrible mess to deal with if virtually all those are of different architectures with different purposes, capabilities, performance...

Next-gen consoles will be symmetrical in design, several of THE SAME processors, with equal architecture, performance etc. It won't be the same kind of messy designs seen in the past.

Now, there's nothing wrong with the guy's intelligence, he's become immortal through programming techniques such as "Carmack's Reverse" and such, and is a veritable rocket scientist to boot, but the guy is LAZY. He quickly embraced high-level languages in games programming (DOOM from 93 had just two assembler routines spliced into it, and after that I guess there's nothing at all), because it's quicker and easier. Now he sees multiprocessing as something bothersome instead rather than the asset it is. A monolithic PC-like machine can never match a similarly configured multiprocessor device in raw performance. The latter is harder to program for - well duh - but hey, that's his job, to overcome challenges, not complain!

The monolithic, cast-iron steam-powered, clay-heeled PC colossus is all Carmack really knows, so it seems he judges everything else with that as the reference. He's brilliant, but a conservative. Bet he voted for Bush.

 

[rabidrabbit]

CPU's as in PC architecture CPU's.
Or processor elements (or whatever they'll be called) like in 'Cell' architecture.
Isn't there a difference.

 

[London Geezer]

Re: ...

It is good that influential people like Carmack and Sweeny are speaking against CELL......

.... And XCPU, new multi cores CPUs from Intel and AMD, dual G5's, multi cores solutions from IBM...the list goes one and on...

Pretty much against everything really... Go figure...

 

[Entropy]

Just playing devil's advocate here.... how many of you guys who are being critical of Carmack, have written complex and big selling computer games?

He quickly embraced high-level languages in games programming (DOOM from 93 had just two assembler routines spliced into it, and after that I guess there's nothing at all), because it's quicker and easier. Now he sees multiprocessing as something bothersome instead rather than the asset it is. A monolithic PC-like machine can never match a similarly configured multiprocessor device in raw performance. The latter is harder to program for - well duh - but hey, that's his job, to overcome challenges, not complain!

The monolithic, cast-iron steam-powered, clay-heeled PC colossus is all Carmack really knows, so it seems he judges everything else with that as the reference. He's brilliant, but a conservative. Bet he voted for Bush.

Well, I've not written "complex and big-selling computer games", but I've waded hip-deep in complex software running on Very Expensive Supercomputers. And I have to say that Guden Oden has a point. Sure it's easier to write code for a single classical CPU, but multiprocessors simply offer greater capabilities IF your problem can be expressed in a form which is sympathetic to the particular architecture. Amdahls law never to be forgotten. To a certain extent it's a problem of education and habits - if you have been taught and practised programming using typical classical algorithic design targeting PC type architectures, multiprocessors will seem exotic.

But for example, a decent programmer today has to have a firm grip on the peculiarities of memory subsystems, and arrange data access and organization accordingly. Adding awareness of data access of multiple processors and inter-thread communications and the other little wrinkles of multiprocessors are just another layer. The more difficult part is IMHO breaking algorithmic habits and starting to "think parallell", rather than translating your habitual thinking to a different platform.

By and large, the Cell concept seems fairly straightforward, and given the small number of processors, the penalty for not making optimal use of them is relatively low. (As the number of processors grow however, this will change, but that's not an immediate problem.)

I have to say I'm a bit disappointed in the attitude of some programmers. I've seen science graduate students just buckle up and attack the problem of wringing optimal performance out of quite gnarly architectures. No whining that it doesn't look like a PC, they just do the job and actually enjoy working on something more interesting. If you're a programmer, I'd say if you don't enjoy working on hardware that is taking a new architectural approach, an approach that generally seems likely will grow in importance of over time, then maybe you should consider switching to another line of work.
Like selling burgers.
Burgers are nice.
You know where you have them.
They stay put.

Playing a bit of Devils Advocate myself, how would you feel Simon if confronted with programmers who felt that "working with TBRs are a hassle. IMRs are proven, and we should just focus on improving them"?

PS. I have the feeling that Carmack might be talking about the next XBox rather than Sonys new console. I have no idea how the new XBox might be organized.

 

[Simon F]

I have to say I'm a bit disappointed in the attitude of some programmers. I've seen science graduate students just buckle up and attack the problem of wringing optimal performance out of quite gnarly architectures.

But he was also talking about the problems of the development cycles increasing and of the spiralling costs. Making the HW harder to use doesn't help matters. I guess it all depends on how good the development tools are in the first years before and after launch - usually these start out as "less than ideal".

Playing a bit of Devils Advocate myself, how would you feel Simon if confronted with programmers who felt that "working with TBRs are a hassle. IMRs are proven, and we should just focus on improving them"?

Touche.
FWIW, we've seen it on a couple of occasions and, yes, you can demonstrate that they are wrong (i.e. the interface is transparent, you can do T&L, and the bang-for-the-buck is there) but it's the old adage "You can lead a programmer to water but you really have to hold his head down for a long time."

Then again, you sometimes get developers saying... "I really wish you had that fantastic translucency sorting that was in Dreamcast".... Very very difficult for an IMR.

 

[jvd]

Well the only multi cpu systems I can remember are the jaguar and saturn. Perhaps 3do was one too ?

Anyway perhaps he is saying for graphics. Making graphics engines with the effects mostly in software must be a bitch to do. On the pc side of thigns the os can allways assign which cpu or core gets which thread for a.i and physics but on the console side things get tricker .

But the industry is moving to this and I guess while he would rather have it another way he will make due with it .


Personaly i'm waiting for dual cpu (not core) pcs to become mainstream and cheap . Once that starts taking off then games will get another big boost on the pc side.

 

[London Geezer]

Personaly i'm waiting for dual cpu (not core) pcs to become mainstream and cheap . Once that starts taking off then games will get another big boost on the pc side.

Same here, so much so i was almost going to buy a dual Opteron system, but then my best friend slapped me in the face and made me understand i don't have that kind of money to spend in other than alcohol and junk food... So now i'm stuck with an Athlon64 3200 and loads of cash for drinks and burgers! YAY!!

 

[function]

John Carmack owns a game development studio, and I think to take these quotes as JC saying "I'm a lazy / limited programmer" would be unfair.

A development studio has to make software that comes in on schedule, on budget, and be of a high standard from beginning to end. It has to do that title after title. Risk is something you want minimised, whether it's the game concept or the target platform.

While some may feel programmers should enjoy tackling new hardware, there's a lot more to the business of making games than this. The less you can spend on an area of development, and the less risk that area of development involves, the better. I think this was where JC was coming from, but I should really look for the full transcript of the Q+A session.

From what I understand, many programmers liked "banging on the hardware" of the Saturn, but ultimately that didn't seem to help the platform.

 

[jvd]

Personaly i'm waiting for dual cpu (not core) pcs to become mainstream and cheap . Once that starts taking off then games will get another big boost on the pc side.

Same here, so much so i was almost going to buy a dual Opteron system, but then my best friend slapped me in the face and made me understand i don't have that kind of money to spend in other than alcohol and junk food... So now i'm stuck with an Athlon64 3200 and loads of cash for drinks and burgers! YAY!!

yea i need lots of beer money too haha .

I'm sure that before the end of the year amd will put out the a64 non fx dual system boards .

I wouldn't mind building an a64 3000+ dual . Would be sweet