No Cell chip in PS3

Games
1

I didn’t see this posted in here, so I figured I would.

Electronics Design Chain has an interesting [url=http://www.designchain.com/coverstory.asp?issue=spring03]article online about the join design of the Cell chip, which was long planned to be the heart of the Playstation3.

Far down in the article is the following choice quote:

“IBM has not announced when it will release the final chip, but the Cell project was originally envisioned as taking five years, suggesting that the final product may not be ready until as late as 2007. Although Sony refuses to talk about its future plans for the chip, it does admit that the Cell chip will not be the CPU in the Playstation3, reportedly slated for release in 2005.”

There’s actually plenty of interesting stuff not related to the PS3 in that article, too. Worth a read on a lazy Sunday afternoon.

Personally, I think the whole Cell thing is going to end up being a bit of a disappointment. Sure, the Sony/Toshiba/IBM alliance is powerful, but Intel and AMD regularly spend just as much money and have just as many engineers working on their next-gen chips. And as good as IBM’s process technology is, it’s still second to Intel’s.

The whole “1,000 times as powerful as a Pentium 4” thing gets tossed around a lot, but it’s all hot air. Measuring things in gigaflops and teraflops is tricky business – you can do a billion floating point operations a second. Great. Which operations? Adds? Copies? Can you retrieve enough DATA to realistically feed your processor for a billion operations?

Comparisons to a P4 seem bunk in a processor coming out in 2007. Or even 2005. Do we think a Pentium 4 will still be Intel’s best chip in two years, let alone four?

One of the big advantages of Cell is that they’re starting with a blank slate, with no need for backward compatibility. Yeah, that worked great for Itanium - which took about three years longer than it was supposed to, from the most experienced and well-funded CPU designers in the world. Oh, and good luck programming for a radically new CPU with no ties to any prior chip. Optimization, compilers, it’ll all be a bitch for the first couple years.

I admire the concept, but I think it’s way WAY too early to get excited about Cell. Or PS3 for that matter.

2

Sony did the same thing before they announced PS2 - remember all those stories about how Sony wouldn’t be allowed to export the “emotion engine” because it was more powerful than supercomputers?

It’s just their PR machine ramping up to make the PS3 seem more powerful. I personally never understood what they meant by the whole cell thing anyway - every couple months some new person from Sony says something about how the PS3 won’t even have removable media, and that it’ll be powered by distributed computing via the Net - whatever.

3

Personally, I think the whole Cell thing is going to end up being a bit of a disappointment. Sure, the Sony/Toshiba/IBM alliance is powerful, but Intel and AMD regularly spend just as much money and have just as many engineers working on their next-gen chips. And as good as IBM’s process technology is, it’s still second to Intel’s.

Are you saying that Dean “Buoyed by so much processing power, consumers will be able to interact with these worlds without worrying about hackers, viruses, or lost connections.” Takahashi is wrong, silly, and a grossly uninformed motherfucker? Is that what you’re saying, you know, in a nutshell?

4

"And as good as IBM’s process technology is, it’s still second to Intel’s. "

You mean their fabbing process or the whole ability to design new and faster chips? When it comes to making chips, IBM is as good as anyone in the world.

5

With process technology I think you’re usually referring to the fabbing process only, and in this area I believe Intel is #1.

6

Reading the patent application, one does get the idea thaat the design was too ambitious to be mass produced practically in a 2005 time frame. I still think it’s a really exciting design, maybe they’ll get it going for the PS4. Now it’s just a question of what they can use in the PS3. Will they beef up the Emotion Engine family, use a scaled back version of the Cell or use something off the shelf? Maybe even go with a slighly custom IBM CPU? And what about the GPU? ATI and nVidia are all but spoken for. I wouldn’t trust Via (S3) or NEC with something so important. I suppose ATI and nVidia would be eager to slip away from their respective partners in order to be inside the next Playstation if given a chance. And Sony might keep it in house, beef up the Graphics Synthesizer or build an all new GPU. I also wonder how long they’ve known they couldn’t use the Cell, and what they might have been up to in that time…

7

“With process technology I think you’re usually referring to the fabbing process only, and in this area I believe Intel is #1.”

Well everything I’ve ever read says IBM is as good at fabbing as anybody in the world. I think people just don’t think about it because they don’t see IBM that way. There are a number of companies that have deals with them in the fab area because of their reputation.

8

Didn’t nVidia just sign up with IBM to fasb their GPUs?

9

Kutaragi is far too proud to use someone else’s chips if CELL isn’t ready. I think you’ll see an evolution of the Emotion Engine in PS3 or maybe two of them running in parallel or something equally outrageous. What better way to get your developers ready for massively parallel computing than to force them into it with a chip they already know.

…and as a bonus, PS3 is immediately backwards compatible with PS2 and PSOne.

–Dave

10

Correct - I’m referring chip manufaturing process technology (ie - 90 nanometer interconnects, strained silicon, silicon on insulator, wafer sizes, yields per wafer, etc etc). IBM is a close second behind Intel in this regard, but Intel still has the capability to produce more advanced chips from a fabrications standpoint, in greater quantity, earlier than anyone else.

Intel will ship millions of 90-nanometer very complex CPUs this year (Prescott), when other fabs will just be getting similar technology out on the market in a tenth the volume.

Thing is, Intel doesn’t do what IBM/TSMC/UMC/etc does. They simply aren’t available for outsourced chip fabrication. Or believe me, people would use them.

It’s just their PR machine ramping up to make the PS3 seem more powerful.

Yup. The PS2 was “1,000 times more powerful than the PS1” also, as I recall. Which of course, it’s not. That’s a great statement to make, though - there’s no standard definition of “power.”

Will they beef up the Emotion Engine family, use a scaled back version of the Cell or use something off the shelf? Maybe even go with a slighly custom IBM CPU?

Like Dave said, Sony’s too proud to go with anything off the shelf. In fact, Sony is proud to the point of hubris, and it bit them with the PS2. They thought they could spend billions to build their own chip fabrication plant and build the Graphics Synthesizer chips themselves - at .13 micron, one of the most complex chips of the time with tons of on-die cache. And it turns out they didn’t know as much as they thought they did about building chips, there were giant supply problems, so they sold the plant off to another company to run (at a big loss) and yeilds went way up. (how come nobody ever factors in costs like R&D and stuff like Sony’s manufacturing problems when they discuss how much each company loses on each console sold?)

Sony does this a lot - develop something that’s not really better than the competing stuff, but that they are convinced is better in every way, and then think they can make it the standard. Memory Stick comes to mind (is there a non-Sony product out there that uses it?). Remember when they said the PS2 was going to be the way we all got online, and was going to replace our PCs and would be the online and digital media gateway for the entire home?

I don’t know if backward-compatibility will be a priority for PS3. It works well enough with PS2, but I don’t know that it’s a real selling point.

The talk is that Sony will design and produce, probably again in combination with Toshiba, it’s own CPU for the PS3. Which will utilize some of the Cell concepts.

BTW - I spoke with some of the engine guys at GDC (Touchdown Entertainment, NDL, Renderware) and jokingly asked if the PS3 is going to be easier to program for than the PS2, which developers weren’t too fond of. They all just sighed and shook their heads - everyone expects it to be much HARDER to code well for. Which if you’re a middleware provider, is both good and bad.

11

I don’t know if backward-compatibility will be a priority for PS3. It works well enough with PS2, but I don’t know that it’s a real selling point.

For a guy that follows PCs, you should know better than to say that. :)

Backward-compatibility is such a strong selling point for any established format. When the next device you buy does not invalidate the entire collection you’ve built up (including peripherals, often expensive ones), it’s a strong reason for purchase of the next device. I’d argue it’s an even stronger draw than things like DVD-movie playback and other non-game uses. With no backward-compatibility, you’re immediately on even ground with your competitors. Sony’s strongest card right now is their enormous installed base of PSOne/PS2 machines. They need to use that to stay on top.

Just look at Nintendo with GBA. You can play games that were made ten years ago on that damn thing. It’s a huge draw for people that already own a library of games for the machine. Obviously, it’s slightly different due to the lack of competition in the space in recent years, but Nintendo has made no noises about ever losing said compatibility. What’s really wacky is that you’ll soon be able to play those ten year old games on a Gamecube via the Game Boy Player… once again almost ten years after you could play them on your TV through your SNES and the Super Game Boy.

Microsoft’s dominance of home computing is predicated in large part to compatibility. It’s the big sticking point with 64-bit computing if I’m not mistaken because Intel wasn’t focusing on backwards compatibility but AMD is making sure things run perfectly well on their 64-bit CPU.

–Dave

12

This just in…Looks like the Cell is in to me.

13

I don’t have the link handy, but Kutaragi pretty much confirmed it won’t be the brains for PS3. Looks like this CNET article is supposing some things based on the fact they’re building a plant…which they were doing anyway for future Cell production.

–Dave

14

If you’ve built up a large collection of games, then chances are good that you already own a device that you can play them on. I’m not saying that it isn’t convenient to have one machine that runs them all, but it’s not like a non-backwards compatible PS3 would make your PS2 games useless if you also own a PS2.

Sony’s strongest card right now is their enormous installed base of PSOne/PS2 machines. They need to use that to stay on top.

I agree, but not for the reasons you state. Sony has a lot of consumer mindshare, much like Nintendo did in the late 80s. When people think “console games,” they think “PlayStation.”

But I doubt that backwards compatibility is a huge selling point. It’s nice to have, but people buy shiny new consoles primarily to play shiny new games. Sure, you can play 10 year old Gameboy games on your GBA… but I wonder how many 10 year old GB games retailers are actually selling to GBA owners? Probably very few.

Microsoft’s dominance of home computing is predicated in large part to compatibility. It’s the big sticking point with 64-bit computing if I’m not mistaken because Intel wasn’t focusing on backwards compatibility but AMD is making sure things run perfectly well on their 64-bit CPU.

That’s a little different, though. Non-game computer software (especially business software) is not as transient as games. Gamers–especially gamers in the market for a new console–want to buy all-new games. Businesses would rather never spend money on new software if they could avoid it. A new PC architecture that is not backwards-compatible with any existing software is a big deal, but backwards compatibility in consoles probably only appeals to a small minority of users.

15

I think you’re right on the point about how it’s more important to PC users.

There’s proof that it’s okay not be backwards compatible for consoles, and deadly not to be compatible with the current mainstream PC OS.

Having said that I think that Sony got a huge boost with compatibility for the PS2. They basically gave the early adopters a big present, and made the idea of switching a more comfortable choice for the majority of people who would rather not have two consoles under their TV.

Also, though it may be hard to remember now, Sony was facing the possibility that other people would be running Playstation software on the PC and the Dreamcast and it would look better. So the PS2 would run your PS1 games, but they would look better!

In the end the question of backwards compatibility will be answered by just how cheaply Sony can stick a PS2 chipset into your PS3.

Your Power Pill

16

"Intel will ship millions of 90-nanometer very complex CPUs this year (Prescott), when other fabs will just be getting similar technology out on the market in a tenth the volume. "

The CPU isn’t the only complex piece in a computer anymore. I’d argue in a top of the line machine the chip on your graphics card is as complicated in design as your CPU is if not not more. You can’t say that producing a chip like Prescott is more complicated than making a R300/NV30 level GPU.

17

And lookie here:

Separately, SCEI said it plans to move production of the main chips for its PlayStation 2 console to a more advanced 90-nanometer production line later this year. That will enable the chips to be manufactured with 4Mbytes of memory embedded on the same silicon as the main and graphics microprocessors, which should result in lower power consumption and improved performance.

The success of that project is going to have a direct effect on whether we see backwards compability in the PS3.

and from the same article:

The companies need such advanced production technology to produce the “Cell” microprocessor, which is being developed jointly by Toshiba, IBM and Sony Computer Entertainment (SCEI) to form the basis of the PlayStation 3 video gaming console and other future consumer electronics products.

The chips are being developed as part of a five-year project that began in 2001 and has as its goal a 1 teraflop-class consumer microprocessor. If successful, the result would be a processor that could be fitted into household electronics yet is more powerful than IBM’s Deep Blue supercomputer.

So Sony is going to go for the gusto… maybe.

I wonder what the back up plan is?

Your Power Pill

18

Backward compatability (at least for me) means I’m more likely to buy the device sooner, because I don’t have to shell out big money to buy many games early. Yeah, you buy the system movers at launch, but if I know I’m going to be able to play 20 games out of the box instead of 1 (or however many I shell out for), it’s easier to stomach the #300 purchase.

It didn’t kill the GameCube (or the SNES, or the N64…), though, so its not nearly as important for consoles to be backward compatible as it is for PCs

19

This seems like to sort of logic that I use on myself when I’m trying to justify something that I’ve already decided to do. ;)

I mean, it seems like you are trying to justify spending $300 on a machine that you will (at least for a while) use to play one or two games. The rationale: you are also using it to play games that you already have.

But you can already play the games that you already have, so you don’t need to buy the new system for that. So you are still spending $300 to play (at least for now) one or two games.

Just sayin’…

20

I think you folks underestimate the power of backward-compatibility. Sony has sold 50 million PS2 units already…that’s some 20 million ahead of where they were with PSOne by this time. They’re at about 95 million PSOne sold overall. At the rate they’re going, you could see 100 million PS2’s within two years at the most.

They’re also putting the Emotion Engine and the Graphics Synthesizer together on one chip with RAM embedded. That means they could turn it into a single chip in PS3 effectively creating a new system with the hardware to run PS2 games easily. Since PSOne’s brains are the I/O chip in PS2, it seems likely they’ll hold onto that one too or possibly just run PSOne games through software emulation which by then would probably be a piece of cake. They can also most likely add cool things like anti-aliasing and the like to PSOne games by then with no hit.

Sony also has one big advantage over Microsoft. With all this proprietary chip technology, they’ve been able to shrink it, make it cheaper and create a situation where they can sell systems for less and still make money long term. Microsoft is stuck with off the shelf parts essentially. The Xbox won’t ever get that cheap to produce because of the stuff in it and their inability to integrate it. They’ll either learn from that and create a far more integrated design next time around, or they’ll stick with the PC legacy to retain backwards-compatibility and go through the same situation of never being able to make money on the console itself or worse, continue losing money hand over fist with each sale. Their biggest mistake was following the Sega model of system sales at a loss. It only works if you get to be number one. Microsoft, with less than 10 million consoles worldwide compared to Sony’s 50 million, is far from number one.

Oh, here’s another one…guess what happens when you buy the PS3 and you’ve already got a PS2?. The PS2 ends up hooked to a different TV in the house. Sony sells TVs, so they know the average house has more than one. With backward compatibility, it’s even more likely that you’ll keep that other system in another room and be able to take games from one to the other without invalidating that system’s games on TV number one.

–Dave