A year behind schedule, can the PlayStation 3’s Cell shrug off vapourware claims?
> News Comments > SPOnG Comments Index
Topic started: Tue, 8 Feb 2005 13:06
Click here to view the news article this topic refers to.
Click here to view the news article this topic refers to.
Wed, 9 Feb 2005 02:52
Will you people listen to yourselves? Some of you are talking about the cell as if it were the holy grail of computing! All Sony have done is succeeded in making a more complicated device for developers to program, while boasting about something that has an ever decreasing relevance in games today: the processor speed. Since NVidia are developing the video processors, the only point of any relevance is the bus speed between video processors and the cell processors. If I had to bet which next gen machine developers will like the most, I would bet the farm on the XBox, as MS do one thing extremely well... Development tools. The developer API's (mostly DirectX) will be largely unchange for the next gen XBox and will be a no-brainer for developers to upskill, but I suspect the PS3 will be an entirely different beast, requiring developers to effectively start from scratch.
Being that a system's success is very much reliant on good software and games, I think we'll see a much greater percentage of XBox 2 games taking full advantage of the hardware available at launch, as opposed to a much smaller percentage of PS3 games, where developers will need time to come to grips with the architecture. Many people seem to use the fact that XBox architecture is similar to a PC, as a negative aspect of the console. It is in fact it's greatest asset, holding the door open for thousands of PC game development companies to develop for XBox. With the next gen, they retain this advantage, and bound ahead of the obstacles facing PS3 game developers.
Being that a system's success is very much reliant on good software and games, I think we'll see a much greater percentage of XBox 2 games taking full advantage of the hardware available at launch, as opposed to a much smaller percentage of PS3 games, where developers will need time to come to grips with the architecture. Many people seem to use the fact that XBox architecture is similar to a PC, as a negative aspect of the console. It is in fact it's greatest asset, holding the door open for thousands of PC game development companies to develop for XBox. With the next gen, they retain this advantage, and bound ahead of the obstacles facing PS3 game developers.
Wed, 9 Feb 2005 09:56
Elysium wrote:
If I had to bet which next gen machine developers will like the most, I would bet the farm on the XBox, as MS do one thing extremely well... Development tools.
And lets not forget XNA. If M$ really get behind this, dev studios will flock to them, because they need to design games with as little hassle/cost as possible.
Sony did this with élan when they implemented their 3rd party development support for PS, and perhaps M$ have been smart enough to get the dev-support ball rolling with XB 360?
If the Cell architecture is a pig to develop for, and if Sony doesn’t spend the time implementing excellent dev tools, then surely they'll loose (US) dev co.s?
The PS3 may have the potential to blow the XB 360 out of the water, but, like Elysium says, developers may struggle with the 1st wave of games, and 1st impressions count. If this happens, we could have a shift in power.
Wed, 9 Feb 2005 10:14
You've got to remember in all this that Cell is really just a more advanced PowerPC processor. The "Cell" is just a piece of software.
There is no doubt in my mind that this makes a lot of sense. It means that it is based on an already well-supported platform and it will be possible for cross-platform ports between PowerPC Xbox and Revolution. However ports would not take advantage of the Cell.
Kid_77 makes the most valid point about development tools. Just look at the rise of Visual Studio and the decline of Borland. Microsoft's biggest problem with XDA is that Sony owns the market and is hardly going to support their technology, and because of this most developers won't support it either.
Prehaps they could come to a compromise. At least for the next generation I can't see Microsoft XDA/Xbox affecting Sony at all. It has the power in the market at the moment and can set the standards.
There is no doubt in my mind that this makes a lot of sense. It means that it is based on an already well-supported platform and it will be possible for cross-platform ports between PowerPC Xbox and Revolution. However ports would not take advantage of the Cell.
Kid_77 makes the most valid point about development tools. Just look at the rise of Visual Studio and the decline of Borland. Microsoft's biggest problem with XDA is that Sony owns the market and is hardly going to support their technology, and because of this most developers won't support it either.
Prehaps they could come to a compromise. At least for the next generation I can't see Microsoft XDA/Xbox affecting Sony at all. It has the power in the market at the moment and can set the standards.
Wed, 9 Feb 2005 12:26
Autobot wrote:
Sony has a very weird quirk, they like to force there new technologies and when the find one that sticks they begin to try and take over the medium and eventually the market.
On what basis do you make these claims? Show me one market that Sony has used its proprietary techology to take control. I cannot think of a single market that the company has done this. There are hundreds of consumer electronics firms out there, prolly a dozen of them in just as strong a position as Sony - though their brand may not have the caché that Sony seems to have. At PS1 launch there was SEGA and Nintendo, both of which lost ground due to their own inept methods.
The only Sony maneouvering I'm troubled by is it relentless aqcuisition frenzy in the music and movie production and publication arena, but that has nothing directly to do with technology.
Sure Sony and Toshiba are making TV's with the cell chip but they are only 2 of many tv companies.
It doesn't mean they won't licence out the tech to other companies. This goes on all the time in TV manufacture. I'll bet my hat that Sony doesn't manufacture all of the components inside its products, and that some of the components are produced by apparent competitors. Daewoo & Philips, for example, produce a lot of the CRTs in TVs and monitors. It's no big leap to see Toshiba or Sony producing the driver or signal processing components for TVs built and badged by other firms.
just think, how long did it take for the mass to accept computers in there life? and how long did it take for the internet to become mainstream? the cell chip just seems to want to take things soo far soo fast that its going to be too much to swallow to fast. If any one of there promises fails the whole project can be doomed.
I think you're expecting the Cell's built into TVs to do too much. There's all this posturing about Cells being able to chatter and share load, but in fact I highly doubt we'll see the Cells in TVs being able to communicate with PS3's from day one. It'll just be a processor to handle the computing currently done by a bunch of discreet components.
Wed, 9 Feb 2005 13:42
OK, some very valid points in this thread. I'm not going to reply to them all directly, but I'd like to talk in general terms about where I see the Cell fitting into the digital landscape.
First and foremost, Cell is not a Sony-only technology. Sony can't use it to dominate any market except maybe the games console market. All the other markets that Sony competes in are covered by Toshiba and IBM - Sony's partners in the Cell consortium.
OK, that said, I think that Cell has some design "features" that will make it a hard sell into the general computer manufacturing world.
Much of my assumptions are based on the following article about Cell by someone hardware-geeky enough to have been at ISSCC and understood what they were on about - don't worry there are also pictures!
Introducing the IBM/Sony/Toshiba Cell Processor
There's also a part two to the article that goes into more depth than just the SPEs.
Basically, Cell is only capable of single precision floating point calculations, at 32bit. This is not a good thing for heavy scientific analysis. It is an extremely good thing for games and embedded devices since it allows all instructions to have fixed execution times.
The Cell is not a drop-in replacement for G4 or G5, so Apple are very unlikely to use Cell in a Mac. One of the instruction sets that MacOS X relies heavily on is called "VMX/Altivec" - a system to allow a single CPU instruction to work on large sets of data. Cell has a worse than G4 implementation of this instruction set. This also makes it very unlikely that Microsoft will port Windows to Cell, even more unlikely than them re-porting XP to G5 or another PPC chip.
Also managing the on-SPE memory is entirely under software control, in the same way the OS manages the main memory of a PC. This is unlike the normal cache memory on a "traditional" CPU. This will make writing a general purpose compiler harder. Not that it will make it impossible, mind you, but getting the optimisation out of generated code that you would be able to with hand rolled is going to take a lot of work by compiler writers. Luckily for a lot of people, IBM have some very, very good compiler writers who will, no doubt, create a back-end for gcc (the GNU C/C++ compiler) that will jump start Linux and other open source projects on Cell.
Note: The above points are not "a bad thing" they are part of the reason Cell can execute so quickly and run at such high clock speeds.
In other news, the Cell is very similar to the PS2 in overall architecture, the G4 on the Cell takes the role of the Emotion Engine and the SPEs take the roles of the PS2's vector units. What this means is that people who write software close to the hardware on PS2 will feel at home writing software for Cell.
OK, all of the above points have led me to the following conclusions;
1) Apart from a few specialised workstations, Cell will not be used in destop computers.
I base this on the lack of the full set of VMX/Altivec instructions. This will make it difficult to get Windows or MacOS X onto Cell. Linux will appear, but I think it will be a specialised distribution, not a desktop one like Fedora Core. This will make it hard to buy computers with "Cell Inside" stickers in the short term, since Linux is not 100% ready for general desktop use. Please note that I use Linux as my home desktop and HTPC OS, but then I'm not "normal", in so many ways. :-)
2) For games, Cell will not be much harder to program than PS2.
The similar architecture to PS2 will enable games developers to move quite smoothly. This will be most apparent with middleware developers such as Criterion and Havok. This will make XNA largely irrelevant in the developer's mind. More and more companies are using middleware to develop their games, I can only see this trend accelerating. Development tools will mostly be created by the platform owners to help middleware companies develop their products.
3) Cell will be used in a lot of entertainment devices.
The Cell consortium will need to license the use of Cell to other manufacturers to recoup the cost of development. No way will Sony, IBM and Toshiba have enough product lines to make their money back themselves. If the Cell consortium produces a good embedded OS for TVs, DVD players and the like, other manufacturers will jump at the chance to just buy such a set-up. This OS will probably be an embedded version of Linux, keeping development and licensing costs down, ensuring a quick development and uptake. Note that I still believe that HD-DVD and Blu-Ray players will not be on Cell processors quickly due to MS licensing of WMP codecs.
4) Cell networking will not take off quickly.
Not until we have a high bandwidth wireless standard that can be built onto the motherboard of Cell-based units. Using a wired connection will require a lot of extra hassle that "normal" people will not put up with. Network switches, signal boosters and the cabling will put off so many people it's not true. However, a wireless networking technology that "just works" like Bluetooth with enable a distributed processing house as long as the bandwidth is high enough. By high bandwidth, I mean able to keep up with the Cell's 100GB/s IO speed. This means that PS3 will not make use of computing power from your TV.
OK, there are my four big predictions and some analysis of facts to back them up.
Discuss.
First and foremost, Cell is not a Sony-only technology. Sony can't use it to dominate any market except maybe the games console market. All the other markets that Sony competes in are covered by Toshiba and IBM - Sony's partners in the Cell consortium.
OK, that said, I think that Cell has some design "features" that will make it a hard sell into the general computer manufacturing world.
Much of my assumptions are based on the following article about Cell by someone hardware-geeky enough to have been at ISSCC and understood what they were on about - don't worry there are also pictures!
Introducing the IBM/Sony/Toshiba Cell Processor
There's also a part two to the article that goes into more depth than just the SPEs.
Basically, Cell is only capable of single precision floating point calculations, at 32bit. This is not a good thing for heavy scientific analysis. It is an extremely good thing for games and embedded devices since it allows all instructions to have fixed execution times.
The Cell is not a drop-in replacement for G4 or G5, so Apple are very unlikely to use Cell in a Mac. One of the instruction sets that MacOS X relies heavily on is called "VMX/Altivec" - a system to allow a single CPU instruction to work on large sets of data. Cell has a worse than G4 implementation of this instruction set. This also makes it very unlikely that Microsoft will port Windows to Cell, even more unlikely than them re-porting XP to G5 or another PPC chip.
Also managing the on-SPE memory is entirely under software control, in the same way the OS manages the main memory of a PC. This is unlike the normal cache memory on a "traditional" CPU. This will make writing a general purpose compiler harder. Not that it will make it impossible, mind you, but getting the optimisation out of generated code that you would be able to with hand rolled is going to take a lot of work by compiler writers. Luckily for a lot of people, IBM have some very, very good compiler writers who will, no doubt, create a back-end for gcc (the GNU C/C++ compiler) that will jump start Linux and other open source projects on Cell.
Note: The above points are not "a bad thing" they are part of the reason Cell can execute so quickly and run at such high clock speeds.
In other news, the Cell is very similar to the PS2 in overall architecture, the G4 on the Cell takes the role of the Emotion Engine and the SPEs take the roles of the PS2's vector units. What this means is that people who write software close to the hardware on PS2 will feel at home writing software for Cell.
OK, all of the above points have led me to the following conclusions;
1) Apart from a few specialised workstations, Cell will not be used in destop computers.
I base this on the lack of the full set of VMX/Altivec instructions. This will make it difficult to get Windows or MacOS X onto Cell. Linux will appear, but I think it will be a specialised distribution, not a desktop one like Fedora Core. This will make it hard to buy computers with "Cell Inside" stickers in the short term, since Linux is not 100% ready for general desktop use. Please note that I use Linux as my home desktop and HTPC OS, but then I'm not "normal", in so many ways. :-)
2) For games, Cell will not be much harder to program than PS2.
The similar architecture to PS2 will enable games developers to move quite smoothly. This will be most apparent with middleware developers such as Criterion and Havok. This will make XNA largely irrelevant in the developer's mind. More and more companies are using middleware to develop their games, I can only see this trend accelerating. Development tools will mostly be created by the platform owners to help middleware companies develop their products.
3) Cell will be used in a lot of entertainment devices.
The Cell consortium will need to license the use of Cell to other manufacturers to recoup the cost of development. No way will Sony, IBM and Toshiba have enough product lines to make their money back themselves. If the Cell consortium produces a good embedded OS for TVs, DVD players and the like, other manufacturers will jump at the chance to just buy such a set-up. This OS will probably be an embedded version of Linux, keeping development and licensing costs down, ensuring a quick development and uptake. Note that I still believe that HD-DVD and Blu-Ray players will not be on Cell processors quickly due to MS licensing of WMP codecs.
4) Cell networking will not take off quickly.
Not until we have a high bandwidth wireless standard that can be built onto the motherboard of Cell-based units. Using a wired connection will require a lot of extra hassle that "normal" people will not put up with. Network switches, signal boosters and the cabling will put off so many people it's not true. However, a wireless networking technology that "just works" like Bluetooth with enable a distributed processing house as long as the bandwidth is high enough. By high bandwidth, I mean able to keep up with the Cell's 100GB/s IO speed. This means that PS3 will not make use of computing power from your TV.
OK, there are my four big predictions and some analysis of facts to back them up.
Discuss.
Wed, 9 Feb 2005 13:51
config wrote:
It'll just be a processor to handle the computing currently done by a bunch of discreet components.
yea theres that and... sony sucks!!! WOOOOO!!!!!
sorry i had to throw it in, its my calling card.
anyway, ive said it before, the whole cell thing looks really intreiguing, if it does half what they say it can do it shoudl easily be doign twice what its closest competetor is doing. I do agree that the ability of multipl cell device to work in tandem will most likely not really be used in any worthwhiel way for some time. More than likely i see it as a way to increase functionality in preexisting items. Example: I saw an add for a new tv the other day, made by HP... it has a memory card slot, cdr drive and a tiny photo printer all built into it. The interface and processign is probably done by something much more expensive or much weaker than a cell will suposedly be. Now take that same thing, throw a cell in it, and suddenly resiszing pictures, runing simple filter effects, and other more complex tasks seem resonable on a tv set (and probably cheaper).
Now they coudl also be sued in DVR's and DVD+-R's and in your console. Over the next few years i imagien the poor guys at best buy having a real hard time tryign to figure out how to display certain items as they all start to implement some of the functionality of each other.
I really believe it is a very genius angle to take on the whole thing, if the timeing is right and they can get people interested(sony is damn good ad advertising, not realy IBM or Toshis strong suit) in the tech then they could take a huge role in all branches of tech development for the forseable future. Hell even robotics would have an intrest in such a cheap powerfull solution... and then that leads to miltary contracting... thats some freaky skynet s**t right there! j/k(sort off)
:)
________
Wed, 9 Feb 2005 14:14
Oh my god! Yeah I agree! No way!
Okay so I'm posting to test out if I can post here okay again, so that's why I cleverly disguised this message with the valuable comments again.
Gotta love that Playstation! Right on! Go Nintendo!
Okay so I'm posting to test out if I can post here okay again, so that's why I cleverly disguised this message with the valuable comments again.
Gotta love that Playstation! Right on! Go Nintendo!
Log-in or register to permanently change your layout setting.
466 comments