Multiproccessor and multitasking in next gen consoles

[chaphack]

Since nextgen consoles are going to be used for more than just games, i wonder how cool will it be for multitasking.

You guys played emus on your PC? It is nice to be able to switch between gaming and going online to find a FAQs almost instantly, or how about switching to watching a divx once you get bored of the game. You can play switcharoo with the programs while leaving everything else in pause. Too bad there is a slight lag when you do that even with emus. I hope PS3/Xb2 can do away with that lag.

When it comes to actual PC games, the "multitasking" is horrible. Quiting a game takes time before your PC can function again. Switching programs totally sucks.

I wonder if CELL will allows for such instantenonous realtime switcharoo?

Ya getcha whatcha i meana?

 

[notAFanB]

'lag' switching between would seem to be enevitable given that most games are rendered top priority on the OS.

besides consider that your reclaiming lost memory when switching between apps an it ain't so suprising.


more memory == faster switching my guess, but that leads to non optimal mem usage.

Ya getcha whatcha i meana?

frankly .. no.

 

[chaphack]

'lag' switching between would seem to be enevitable given that most games are rendered top priority on the OS.

besides consider that your reclaiming lost memory when switching between apps an it ain't so suprising.


more memory == faster switching my guess, but that leads to non optimal mem usage.

yay, i know about the reclaiming memory stuff. So I just wonder if future consoles, using multiple cpus, can change programs(games/dvd recording/mp3/online etc) in an instant?

I mean, if you want a multipurpose system, it be cool to use all those functions in realtime. And not restricted to games only or movies only or internet only etc, at any period of usage.

 

[notAFanB]

again probably not, the problem is not with the current concurrency model. rather the hardware which it operates on is continually stretched to it's limits by users.

give me 1G memory and 16 CPU's in parrellel and I'll run so many apps to 'create' the lag.

I supposse we could throw More CPU's at the problem but then for games the extra 'grunt' has to remain unused for the solution to your problem.

 

[DeadmeatGA]

...

1. Multiprocessing.

I am quite sure MS will stick with a uniprocessor model for Xbox2, while I do not know for sure what to make of PSX3, since CELL is rumored to be scrapped for the platform. Multiprocessing is best suited for server applications that serve a number of client applications concurrently, and games have traditionally been single threaded.

2. Multitasking.
Game consoles serve only one purpose only, that is to play game in the living room. Multitasking will not be a common place with next gen consoles.

 

[Panajev2001a]

while I do not know for sure what to make of PSX3, since CELL is rumored to be scrapped for the platform.

The Design Chair mis-interpretation and your repeating it in the dark corner of your room does not make such a thing a credible rumor...

 

[Grall]

I wouldn't expect any console to come with multiple CPUs. I fully expect the XB to come with a hype-threaded P4 (pun definitely intended) however, though I dunno how much good it would do in a console where careful programming and zero irrelevant threads running in the background would make sure pipeline stalls are kept to a minimum.

Besides, hype-threading can cause weird cache behavior, at least in current implementations, so it's not all fun and dance either.

*G*

 

[Paul]

Try reading up on some recent news Deadmeat... I guess SCEI just wasted those billions they just recently invested in Cell production in mid April right?

Please..

 

[DeadmeatGA]

...

Try reading up on some recent news Deadmeat... I guess SCEI just wasted those billions they just recently invested in Cell production in mid April right?

A hardware is useless without the software to animate it.

If I were you, I would keep an eye on CELL OS and compiler development progress to guess the eventual release date of PSX3 , not the chipset itself.

 

[cybamerc]

DeadmeatGA:

> consoles serve only one purpose only, that is to play in the living room.
> Multitasking will not be a common place with next gen consoles.

Multitasking might not be commonplace but the next generation Sony and M$ set-top boxes will do much more than play games. The PC roots of the Xbox will be particularly evident.

 

[Paul]

A hardware is useless without the software to animate it.

If I were you, I would keep an eye on CELL OS and compiler development progress to guess the eventual release date of PSX3 , not the chipset itself.

Your telling me IBM is really as retarded as you claim and are not even working on a Cell OS and the tools to go with it? So I guess they will have a 1TFLOPS consumer chip in 2005 but with no software to make it work for another year and a half correct? That's what your implying and it's pretty dumb, let's use a little logic here.

 

[cybamerc]

Paul:

> Your telling me IBM is really as retarded as you claim and are not even
> working on a Cell OS and the tools to go with it?

That's not what he's saying. IBM began working on OS design quite some time ago but development is still far behind that of the chip.

 

[Paul]

No, it's exactly what he's saying.

Care to provide recent links stating that IBM is behind on the tools for a PS3 version of Cell? Anyone who thinks that billions can be spent on a chip and than having the companies magicly forget about the software to get it work isn't only dumb, it's almost bias.

 

[DeadmeatGA]

Multitasking might not be commonplace

Of course multitasking is already commonplace at the OS level, it is just that most games are not multithreaded and games do not render themselves easily to multithreading.

but the next generation Sony and M$ set-top boxes will do much more than play games. The PC roots of the Xbox will be particularly evident.

Have game consoles been anything other than game consoles?? Who uses their PSX2 to browse web and play DVDs??? How many people have bought the DVD player remote controller for Xbox?? A game console is a game console. Consumers already have rejected any attempts at hydrid devices for over 10 years, and I doubt the consumer preference is about to change anytime soon.

Your telling me IBM is really as retarded as you claim and are not even working on a Cell OS and the tools to go with it?

I haven't heard any word on the progress of CELL OS and compilers. Software tool development is more complex than the hardware itself and always lags behind hardware schedule. It is a known fact.

So I guess they will have a 1TFLOPS consumer chip in 2005

Still buying into that 1 TFLOPS story????

but with no software to make it work for another year and a half correct?

More like 2~3 years at least given the project complexity.

Anyone who thinks that billions can be spent on a chip and than having the companies magicly forget about the software to get it work isn't only dumb, it's almost bias.

Well, Intel threw in an insane amount of money on IA64 Linux and NT, but those came out years later than the chip itself. Remember that IA64 Linux and NT were mere port jobs, nothing new. What Sony is proposing is far more ambitious than Itanium will ever be, and the OS and tool development is to be delayed several times.

 

[DeadmeatGA]

...

The only way you can make some thing like CELL work is if you perfect the concept of "autothreading"; in other word, the compiler itself breaks down the source code into several micro processes based on data dependency analysis and stitch them together to produce an application(game). Any manual thread synchronization management via OS calls is bound to fail.

 

[cybamerc]

DeadmeatGA:

> Who uses their PSX2 to browse web and play DVDs???

It's my impression that there are plenty of people who use the DVD function in the PS2. Even if there aren't it's really not relevant as the functionality is there. A PC does not become a console if you only use it for games.

> A game console is a game console.

And a set-top box is a set-top box.

> I doubt the consumer preference is about to change anytime soon

Tell that to Xbox management.

 

[Paul]

I haven't heard any word on the progress of CELL OS and compilers.

Your argument ends with that quote.

Still buying into that 1 TFLOPS story????

Oh please..

More like 2~3 years at least given the project complexity.

So Cell isn't going to be ready to put into a device until 2008? Despite Kutagari saying the chip will launch in 2005, Launch meaning the chip will actually come inside some device. But how can this be if you say that the
chip will infact be useless until 2008...

Well, Intel threw in an insane amount of money on IA64 Linux and NT, but those came out years later than the chip itself.

A ps3 version of Cell is quite different. What deadlines did Intel have with IA64? Cell needs to be ready for PS3, Sony, IBM and Toshiba know this.

Also, why would the ram contract for PS3 be settled now? If as you say Cell will not be ready for PS3 until 2007-8 wouldn't Sony just wait until 2006 to decide this? You don't make any sense.[/quote]

 

[Vince]

Have game consoles been anything other than game consoles?? Who uses their PSX2 to browse web and play DVDs??? How many people have bought the DVD player remote controller for Xbox?? A game console is a game console. Consumers already have rejected any attempts at hydrid devices for over 10 years, and I doubt the consumer preference is about to change anytime soon.

Deadmeat, anyone thinking about it in that paradigm is of course goign to fail in the maketpalce. Nobody wants a PC replacement in their livingroom.. eww.

What you need to do is make the design very usable and user-friendly. No open OS, no installing software. You keep it tightly focused and say FU to the current MS-based PC-paradigm. Make it more like Apple, all Plug & Play, No/minimal set-up for the consumer... you need to be able to have a trained monkey operate these.

And what you'll see is that is it's adoption is rather like that of the PS2 and it's DVD Player - which many people use. OT, but I cna't help but see the irony of you stating that nobody uses the PS2 to play DVDs, but I thought that was like it's only selling pt for the first year - being cheaper than stand-alone DVD players.

I haven't heard any word on the progress of CELL OS and compilers. Software tool development is more complex than the hardware itself and always lags behind hardware schedule. It is a known fact.

I also haven't heard a confirmed thing about Osama, Saddam, XBox2, NV40, R400 - doesn't mean they don't exist... (I'm sorry, but this deserved the roll)

The idea being that just because something isn't known with definity to us, doesn't mean it just doesn't exist.

Still buying into that 1 TFLOPS story????

Uh huh.

The only way you can make some thing like CELL work is if you perfect the concept of "autothreading"; in other word, the compiler itself breaks down the source code into several micro processes based on data dependency analysis and stitch them together to produce an application(game). Any manual thread synchronization management via OS calls is bound to fail.

Well, thankfuly they have around 30 years of research to use and expand upon. Always so negative... the glass isn't allways half empty.

Hey, I got one for you (and Zurich whose into the thought experiments): If a full glass is in a closed box within IBM's Austin labs when the project starts (signifying it's all good)... can you say it's still full, or empty before opening the box? Or is it in a state of both? Maybe we can draw us the pre-collapse wave function and end your debate.. wait, no.

OK.. I've had enough

 

[qwerty2000]

of course the ps3 will have over 1 tflop maybe 1.256-1.512tflop

 

[DeadmeatGA]

...

So Cell isn't going to be ready to put into a device until 2008?

Chip might be done earlier, but it has no soul(The OS)...

A ps3 version of Cell is quite different. What deadlines did Intel have with IA64?

1999. Missed by 2 years because of software problems... Sony will face a bigger hurdle than porting Linux and NT to IA64...

Cell needs to be ready for PS3, Sony, IBM and Toshiba know this.

Hardware engineering is rather simple; this is why X86 has clones while Windows has none.

Also, why would the ram contract for PS3 be settled now?

Sony is going with a proprietary memory technology, so they wouldn't just be able to pick a standard memory device later.

If as you say Cell will not be ready for PS3 until 2007-8 wouldn't Sony just wait until 2006 to decide this? You don't make any sense.

A computing platform always consists of two parts, the body(hardware) and the soul(software). The hardware will be done earlier but the OS engineering will face a delay no doubt.

What you need to do is make the design very usable and user-friendly. No open OS, no installing software.

I was talking about developer's perspective of CELL OS.

Well, thankfuly they have around 30 years of research to use and expand upon.

And I have seen the fruit of 30-year research in the form of MPI2.0; nothing automatic, you do it mentality....

What's with "You do it" mentality of software engineering anyway??? By now shouldn't developers enjoy auto I/O and auto threading????