John Carmarck bothered with Next gen MProcessor Consoles

[Paul]

I've said before that I don't think the average programmer will even know they are working on a multiprocessor system, it'll all be in the libraries.

Allow me to quote Brad Grenz, a engineer of thermal systems for Toshiba on the Broadband Engine project.

The whole concept of Cell hinges on it being completely transparent. Programmers will never be coding to the metal like they do now. You just tell the Cell what you need done and when you need it and it dynamically schedules, synchronizes and manages execution. You're right, it's absurd to expect programmers to juggle 512 execution units, but they'll never have to.

 

[Deadmeat]

...

Brad Grenz, a engineer of thermal systems for Toshiba on the Broadband Engine project.

I thought this "Toshiba Thermal Engineer" story was somebody's joke???

 

[Brimstone]

I don't think MS want mods in their console games, it will open the door for piracy on the platform.

New player models and such, sure, why not. Maybe the occational level, but not full-blown mods, because that requires code changes in the executables themselves.

Other manufacturers (ie, Sony, Nintendo), could just as well allow player-created content too, no particular advantage for MS here.

My point was entirely missed. People that have the talent to make games are a scarce resource. Publisher/developers making games for the X-Box 2 will be able to hire talent (programmers and artists) from the PC mod community because the PC platform and the X-Box 2 platform will be very similar (XNA tools and enviroment). These people will have real world experience with game development.

This is a very signifigant advantage.

 

[Paul]

I thought this "Toshiba Thermal Engineer" story was somebody's joke???

I've known him for a while... so no, it's not a joke. Me and Pana were even in a AIM chat with him once, he was talking about all types of shit they were thinking about to cool the BE. "overdrive" cooling systems, nitrogen, sleeping APU's(Which I'm sure was talked about in a patent).

 

[Deadmeat]

...

You know this guy's talking BS when he says

You're right, it's absurd to expect programmers to juggle 512 execution units, but they'll never have to.

512 execution units? How many APUs is that?? 128 APUs? Or 16 PEs on a chip??? This guys's so clueless it was easy to tell he was just another rumor spreader on the net..

It would sound more credible if someone claiming to be an eletrical engineering intern at Austin posted this stuff....

 

[Paul]

I have known him personally for years DM.. He's not clueless, as for the 512 execution thing.. I'm pretty sure he was exagerating it. Look at the context of the whole sentence.

 

[Dio]

Re: ...

What exactly are you defining as 'DirectX' and 'I/O' here (In terms of jobs the CPU will actually have to do)?

Those run on separate threads. Hence the "user" threads are free to do its own thing.

That's not an answer. I asked what will these threads do, not how they are implemented.

 

[MfA]

That developers might not have to manually schedule code is a distinct possibility granted, but that isnt the hard part. The hard part is structuring your code so there is enough to schedule at any given time to keep all the cores busy.

Marco

 

[ERP]

That developers might not have to manually schedule code is a distinct possibility granted, but that isnt the hard part. The hard part is structuring your code so there is enough to schedule at any given time to keep all the cores busy.

Marco

Exactly....

 

[Vince]

Re: ...

512 execution units? How many APUs is that?? 128 APUs? Or 16 PEs on a chip??? This guys's so clueless it was easy to tell he was just another rumor spreader on the net.

If I recall correctly, that exact number of execution units was posited as a What If at GDC2002.

 

[Squeak]

That developers might not have to manually schedule code is a distinct possibility granted, but that isnt the hard part. The hard part is structuring your code so there is enough to schedule at any given time to keep all the cores busy.

Marco

Exactly....

AFAIK, balancing the code, to keep an even load on system recourses, has always been an aspect of the art of programming.

 

[ERP]

That developers might not have to manually schedule code is a distinct possibility granted, but that isnt the hard part. The hard part is structuring your code so there is enough to schedule at any given time to keep all the cores busy.

Marco

Exactly....

AFAIK, balancing the code, to keep an even load on system recourses, has always been an aspect of the art of programming.

The difference with the new architectures is the scale, let's take the example of a single PE with 8 APU's, I now need to have 8 completly independant tasks to keep all the APU's busy, and realistically many times that number to have enough granularity to make dynamic rescheduling usefull.

Some game tasks have good fine grain parelellism, graphics is an example and I could dedicate 2 or even 4 APU's to geometry work, and that simplifies things somewhat. But when it gets to medium grain parallelism where the biggest potential lies, it's a far more difficult problem. Asynchronous execution of individual code fragments requires rethinking of the way game code is generally structured. I'm not saying it won't happen, just that it won't happen overnight.

Now if it turns out the the PE's power core has adequate performance it may be a complete none issue, most early games will simply run on the PowerCore and use the APU's in the same fashion that the PS2 currently uses it's VU's. The patent discusses no details of the power core, how much cache does it have, clock rate etc etc.

 

[Guden Oden]

My point was entirely missed.

I wouldn't say that, I just don't see MS having much of an advantage compared to any other developer.

People that have the talent to make games are a scarce resource. Publisher/developers making games for the X-Box 2 will be able to hire talent (programmers and artists) from the PC mod community because the PC platform and the X-Box 2 platform will be very similar (XNA tools and enviroment).

...And you think people from the mod community will be unable to accomplish anything without this XNA environment? Rofl man, talent is talent. There's no such thing as talent ONLY with one set of devtools. Besides, we have no idea how pervasive XNA will be either in the PC space or console either for that matter; there are many different sets of devtools, and today's games even come with their own. Witness UnrealED and Doom3 for example.

Sony, Nintendo, etc would be just as able to hire these people (provided they're any good working as a full-time professional developer; some hobbyists just aren't equipped to handle that), they'd need to be retrained to some degree but that's not a problem.

You can't even be certain modders used to XNA on PCs would prefer to work with Microsoft on console titles rather than Sony or Nintendo. What if those guys are big Mario fans and hate Halo?

This is a very signifigant advantage.

You're just making a lot of assumptions. Nothing's as set in stone as you're trying to make it look like.

 

[Brimstone]

My point was entirely missed.

I wouldn't say that, I just don't see MS having much of an advantage compared to any other developer.

People that have the talent to make games are a scarce resource. Publisher/developers making games for the X-Box 2 will be able to hire talent (programmers and artists) from the PC mod community because the PC platform and the X-Box 2 platform will be very similar (XNA tools and enviroment).

...And you think people from the mod community will be unable to accomplish anything without this XNA environment? Rofl man, talent is talent. There's no such thing as talent ONLY with one set of devtools. Besides, we have no idea how pervasive XNA will be either in the PC space or console either for that matter; there are many different sets of devtools, and today's games even come with their own. Witness UnrealED and Doom3 for example.

Sony, Nintendo, etc would be just as able to hire these people (provided they're any good working as a full-time professional developer; some hobbyists just aren't equipped to handle that), they'd need to be retrained to some degree but that's not a problem.

You can't even be certain modders used to XNA on PCs would prefer to work with Microsoft on console titles rather than Sony or Nintendo. What if those guys are big Mario fans and hate Halo?

This is a very signifigant advantage.

You're just making a lot of assumptions. Nothing's as set in stone as you're trying to make it look like.

Developers want people that can produce results. Hiring a person that isn't familar with a radically different enviroment is going to be close to dead weight until they get up to speed. Time is money. Some wealthy developers will have the resources to nurture and grow talent but a lot just won't be able to afford it. Sony will have to replicate their installed base with the PS3 generation or the economical advantage will shift to Microsoft because of better worker productivity. Unless Sony does pull of a great developer friendly enviroment, which they might do.

Investors don't care what platform games get made for, they just want a return on their investment.

 

[Guden Oden]

Developers want people that can produce results. Hiring a person that isn't familar with a radically different enviroment is going to be close to dead weight until they get up to speed. Time is money.

Oh that's a big load of bollocks if ever I saw it.

Where do you think game developers come from in the first place, huh? That they just pop like mushrooms out of the soil, ready-trained and ripe for picking? Naturally, there will just about always be some on-site training needed when a greenhorn arrives at his first job.

You must be out of your mind if you think other devs than MS would skip over talented people just because they don't want to invest in some extra training... It's not like they are the only guys out there with money in their pocket to spend.

You have to realize, XNA will not be fundamentally different from just about any other environment already out there. There will be differences as with every other package sure, but if there's one thing we can trust MS to do, it's to rip off other people's good ideas, so people used to other environments will likely find their way around pretty easily... No seriously, as I understand it XNA is not primarily targetted at hobbyist modders, but at already experienced developers that are used to things working in a particular way; you can't very well change everything. You wouldn't make a formula one car where the gas pedal is on the left side and the car turns in the opposite direction of the steering wheel for example because the drivers aren't used to things working like that.

I'm sure there won't be any insurmountably radical differences compared to other devtools. People used to XNA will be able to adapt, because like I said, talent is talent.

Sony will have to replicate their installed base with the PS3 generation or the economical advantage will shift to Microsoft because of better worker productivity.

I think you vastly overstate the impact of XNA even before the product's been released man.

You come back the day the world is flooding over with homegrown talent who grew up using XNA in their basements designing FPS mods and are now big-shot game devs working for MS, beavering away like ferrets running round and round in one of those little wire wheels making kickass games, leading the march of nextbox domination of the market, slowly crushing the competition like a steamroller.

Investors don't care what platform games get made for, they just want a return on their investment.

You think "investors" have much say in the way games are developed ANYWHERE? "Investors" are just suits, they don't know dick about game development, or the potential importance (or not) of XNA, if they ever even heard of it. If the thing pancakes, MS will just quietly drop the concept like a well-filled douchebag and that will be all.

 

[Simon F]

PS. message passing does not prohibit the use of shared memory as a way of communicating data, messages can contain references (there are even ways of maintaining CSP purity with reference passing, if you want that).

Sorry - I never meant to imply that it did. I can visualise all sorts of architectures that mix the 'PC' and 'Transputer' models and there are lots of other things (NUMA, DMA engines, multiplexers, etc.) that could also be involved.

FWIW, the transputer system at my last job used both message passing and some shared memory.

 

[jvd]

Personaly i think the big two are going multi chip for a reason . That reason is its the quickest way to prolong the life of the console .

You can greatly enhance the speed above what you could with one chip .

Pcs don't do that because they can keep putting out faster chips every other month if they want .

 

[rabidrabbit]

Personaly i think the big two are going multi chip for a reason . That reason is its the quickest way to prolong the life of the console .

You can greatly enhance the speed above what you could with one chip .

Pcs don't do that because they can keep putting out faster chips every other month if they want .

You mean during the console's first year they are allowed to program only for one processor, on the second year they are given libraries that unlock the second processor etc... :?

 

[jvd]

Personaly i think the big two are going multi chip for a reason . That reason is its the quickest way to prolong the life of the console .

You can greatly enhance the speed above what you could with one chip .

Pcs don't do that because they can keep putting out faster chips every other month if they want .

You mean during the console's first year they are allowed to program only for one processor, on the second year they are given libraries that unlock the second processor etc... :?

Um no

What I'm saying is that a 1 ghz chip may be top of the line when a console ships but in 5 years it will be extremly slow.

Since they can't build a 4 ghz chip they instead go multi core / chip because it will give them more of an increase in speed than a simple speed jump that they could acomplish on the same tech .

Get what i'm saying

 

[PC-Engine]

What jvd said makes a lot of sense. PCs come out with faster cpus every few months. On the console side, you can't have upgradable cpus etc. because it would fragment the market, so the only choice is to have multiple cpus in a closed box design.