CELL CHIP?? Big news for apple?

[pubwvj]

When we migrated from 68K to PPC, I remember all sorts of tales about how emulation would allow computers to pretend to be other types and work faster than the real thing. We were told in glowing terms how RISC computers would be so fast that they could pretend to be anything the software writer wanted.

The reality turned out to be even better. I have very old CISC software written back in the 1980's, some by me, some by others. It's legacy stuff we still need to use but isn't worth rewriting for the PowerPC. The beauty is it runs far faster on our current G3 and G4 machines (never mind G5 et al.) at a mear 266MHz to 500MHz than it ever ran on the original machines or even the CISC machines of the 1990's. The promise really came through.

Then there is VPC. Note something you may like using but it does do the job when you just have to touch the evil.

 

[pubwvj]

From IBM's website, it appears that the die size of the 970FX is just 65 mm square. The Cell is 221 mm square. All things being relatively equal in fab costs, the Cell will cost about 3.4 times more to produce than the G5s chips you find in the current PowerMacs.

There is a lot more to chip costs than die size. % rejection, etc make a big difference. All things aren't relatively equal.

 

[pubwvj]

If this new chip is going to be incorporated into Apple products, we can safely assume it'll happen by mid 2006. The reason being that microsoft will be shipping longhorn around that period, so alot of people will be upgrading their pcs to run it, but if Apple come along with a macmini for $400 running a cell and OS Tiger, customers will think twice before upgrading to a new intel or amd pc.

Good point about the time table. Especially if the MacCell Mini runs Windows applications thus easing the transition like Classic and emulation on the Mac did for 9 to OSX and CISC to RISC.

 

[wrldwzrd89]

The reality turned out to be even better. I have very old CISC software written back in the 1980's, some by me, some by others. It's legacy stuff we still need to use but isn't worth rewriting for the PowerPC. The beauty is it runs far faster on our current G3 and G4 machines (never mind G5 et al.) at a mear 266MHz to 500MHz than it ever ran on the original machines or even the CISC machines of the 1990's. The promise really came through.

Then there is VPC. Note something you may like using but it does do the job when you just have to touch the evil.

Apple came up with a wonderful piece of software called a 68k-to-PPC dynamic recompiling emulator and used it to run all old 68k code, including parts of the Mac OS itself before the Mac OS 8.5 era. It got better and better with each release - boy does it fly now on today's hardware.

 

[pubwvj]

I don't know about you, but I just don't see this being called the G6. The Cell processor still has a PowerPC 970 inside with multiple APU's attached-it's technically a souped-up G5, and the very few programs that cannot take advantage of Cell

Why? G4 is roughly a G3 with an vector processor inside. Same-o, same-o.

What would make a lot of sense is for Apple to standardize on a scaleable processor and stick with it for a long time. Then they can focus on stability, real-world ruggedness, battery life, etc. Imagine:

iPod - Cell with 1 APU - no screen - place music, stores data

iPal - Apple's PDA with Cell + 1 to 4 APUs with a good small screen - replaces high end iPod - carry all your data with you, backups up Home to iPal nightly (continuously?) - your data in the field.

iBook - Cell + 8 to 16 APUs - the future best seller - compute at home, work and on the road

iMac - Cell + 8 to N APUs - desktop version of the Mac (why bother but some people want them)

That is the basic lineup. Of course you can have your Power versions (PowerPod, PowerPal, PowerBook, PowerMac) of each with more features and power for the pros. Maybe additional Cells are plug-in-able like memory. "Care to super-size that iBook, sir? Can I offer you a few more Cells?"

By having a more standardized core hardware it will make software developement even easier and more consistant than it is already.

 

[GFLPraxis]

Why? G4 is roughly a G3 with an vector processor inside. Same-o, same-o.

What would make a lot of sense is for Apple to standardize on a scaleable processor and stick with it for a long time. Then they can focus on stability, real-world ruggedness, battery life, etc. Imagine:

iPod - Cell with 1 APU - no screen - place music, stores data

iPal - Apple's PDA with Cell + 1 to 4 APUs with a good small screen - replaces high end iPod - carry all your data with you, backups up Home to iPal nightly (continuously?) - your data in the field.

iBook - Cell + 8 to 16 APUs - the future best seller - compute at home, work and on the road

iMac - Cell + 8 to N APUs - desktop version of the Mac (why bother but some people want them)

That is the basic lineup. Of course you can have your Power versions (PowerPod, PowerPal, PowerBook, PowerMac) of each with more features and power for the pros. Maybe additional Cells are plug-in-able like memory. "Care to super-size that iBook, sir? Can I offer you a few more Cells?"

By having a more standardized core hardware it will make software developement even easier and more consistant than it is already.

Actually, I've changed my stance on the name. This isn't a PowerPC 970 (I was mistaken on that), it's a new PPC based on POWER5, so it WOULD be called G6.

 

[GFLPraxis]

I've been discussing this on another forum, as well. Someone smarter than me has made some interesting points.

Could someone read this and tell me how accurate it is? (I'm no processor engineer)

If he's right, Apple WOULDN'T use it.

http://bbs.stardestroyer.net/viewtopic.php?t=62208&postdays=0&postorder=asc&start=25

The post in particular:

You're not getting this are you? For desktop applications, the Cell is an oversized waste, only applications that use highly parallelized code are going to have any use for it. Apple might have use for it as an accelerator, but for 90% of the applications people used, Cell isn't going to provide much of a performance boost. This is the same reason using GPU's for general processing tasks has been mostly stillborn outside of render farms--worse in the case of cell since you are devoting millions of transitors to logic that will almost never be taken advantage of, regardless of how much you optimize the code.

And his reply to mine:

Yes, rendering would see quite a boost, but it's not something that enough Mac users do to increase the costs of Apple CPU's by that much, which is why it would make more sense to sell it as an accelerator like Durandal suggested, much like the old x87's.
...
No, games, especially on the Mac platform, aren't going to see much of a boost from Cell. Most of the parallelized code in a game is in the graphics engine, which is already handled by the GPU. It is concievable that future games will make use of a lot of parallel code for AI and background processing tasks, but game programmers are not going to add such exotic features purely for the Mac market, if they will at all.
...


Not much of an etcetera here. Most desktop software outside of rendering simply isn't going to use these vectored FP engines that the Cell offers, no matter how much you optimize it. It's not worth it to add it as a standard feature at this point in time.
...

Only in VERY select pro applications. Why do you think someone hasn't done this before? Why did the drive to use GPU's for software besides games largely fail? Because it's a lot of silicon that isn't going to be of much use outside of HPC.

The ...s are where he was quoting me.
Is he right?

 

[wrldwzrd89]

I've been discussing this on another forum, as well. Someone smarter than me has made some interesting points.

Could someone read this and tell me how accurate it is? (I'm no processor engineer)

If he's right, Apple WOULDN'T use it.

http://bbs.stardestroyer.net/viewtopic.php?t=62208&postdays=0&postorder=asc&start=25

The post in particular:


And his reply to mine:



The ...s are where he was quoting me.
Is he right?

Judging by what I know about Cell, and what was written in that post, I'd say that's 100% correct. The Cell may have the most amazing vector capabilities, but the vast majority of Mac software isn't vectorized and thus wouldn't take full advantage of what the Cell offers.

 

[GFLPraxis]

Judging by what I know about Cell, and what was written in that post, I'd say that's 100% correct. The Cell may have the most amazing vector capabilities, but the vast majority of Mac software isn't vectorized and thus wouldn't take full advantage of what the Cell offers.

So then why would Apple use it?

Would it be possible to use it to replace the graphics card like I mentioned earlier?

Or would it end up just boosting the cost of the Mac up and offering few performance boosts?

 

[pubwvj]

I don't know about you, but I just don't see this being called the G6. The Cell processor still has a PowerPC 970 inside with multiple APU's attached-it's technically a souped-up G5, and the very few programs that cannot take advantage of Cell

Why? G4 is roughly a G3 with an vector processor inside. Same-o, same-o. g3 + vectors -> G4, G5 + APUs -> G6. Makes reasonable sense.

What would make a lot of sense is for Apple to standardize on a unified scaleable processor and stick with it for a long time. Then they can focus on stability, real-world ruggedness, battery life, etc. Imagine tough machines you can take to the beach:

iPod - Cell with 1 APU - no screen - place music, stores data

iPal - Apple's PDA with Cell + 1 to 4 APUs with a good small screen - replaces high end iPod - carry all your data with you, backups up Home to iPal nightly (continuously?) - your data in the field on a slower machine but still accessible.

iBook - Cell + 8 to 16 APUs - the future best seller - compute at home, work and on the road. Expandable with extra plug-in Cells, memory, HD...

iMac - Cell + 8 to N APUs - desktop version of the Mac (why bother but some people want them)

XServer - Cells to the Nth for the ultimate in scaleable supercomputing.

That is the basic lineup. Of course you can have your Power versions (PowerPod, PowerPal, PowerBook, PowerMac) of each with more features and power for the pros. Maybe additional Cells are plug-ins like memory. "Care to super-size that iBook, sir? Can I offer you a few more Cells?"

By having a more standardized core hardware it will make software developement even easier and more consistant than it is already.

 

[wrldwzrd89]

So then why would Apple use it?

Would it be possible to use it to replace the graphics card like I mentioned earlier?

Or would it end up just boosting the cost of the Mac up and offering few performance boosts?

Apple would make it an add-on accelerator for software designed to take advantage of its capabilities if they were to use it in their Macs. Other software would simply continue running on the main PowerPC core like it always has (ever since the transition from 68k to PPC, at least).

 

[GFLPraxis]

Apple would make it an add-on accelerator for software designed to take advantage of its capabilities if they were to use it in their Macs. Other software would simply continue running on the main PowerPC core like it always has (ever since the transition from 68k to PPC, at least).

So you're saying Apple WOULDN'T sell systems with Cells in them?

 

[pubwvj]

I've been discussing this on another forum, as well. Someone smarter than me has made some interesting points.
:
Is he right?

Sort of, sort of not. One of the most fascinating aspects is that the Cell architecture could replace a bunch of other hardware. This would save space on the motherboard, cut power consumption, reduce heat loads, speed up communications, and run a whole lot faster while doing it.

Your word processor, address book and email programs would not feel a whole lot snappier.

It opens the possibility for several Cells working together in a PowerBook, or just one in an iPal (my current favorite for the handheld Mac) which would give decent processing speed so you could access all your data in the field. Or 32 in a PowerMac. Or 4096 in an XServer. It could unify the computing hardware in a way that we don't currently have. This could go all the way from the lowly embedded processor in your iPod or toaster oven (Gak!) to your closet supercomputer.

On the other hand, or the gripping claw if you prefer, applications that are written to take advantage of the Cell would scream to highest halls of Valhalla. Examples would be photo editing programs, movie programs, animation, massive databases, web servers, flight sims (e.g., X-Plane), renderers, weather forcasterers, particle physics simulators, etc.

 

[wrldwzrd89]

So you're saying Apple WOULDN'T sell systems with Cells in them?

Correct! Apple really can't justify using a Cell in a Mac as the only processor because it wouldn't be anywhere near fully utilized most of the time. Folding@Home would absolutely FLY on one of those, since the folding process is highly vectorizable...but that's beside the point.

 

[relimw]

Correct! Apple really can't justify using a Cell in a Mac as the only processor because it wouldn't be anywhere near fully utilized most of the time. Folding@Home would absolutely FLY on one of those, since the folding process is highly vectorizable...but that's beside the point.

But you're missing the point. The Cell processor is an architecture, not a specific processor. The chip being presented at ISSCC is the one Sony designed for their needs. Nothing that I have seen has said that the Sony variant is the one and only configuration.

From IBM's Cell page

STI cell processor defined
Two years ago, Sony and Toshiba and IBM (STI) announced that they had teamed up to design an architecture for what is termed a system-on-a-chip (SoC) design. Code-named Cell, chips based on the architecture will be able to use ultra high-speed broadband connectivity to interoperate with one another as one complete system, similar to the way neural cells interoperate over the brain's network.

 

[wrldwzrd89]

But you're missing the point. The Cell processor is an architecture, not a specific processor. The chip being presented at ISSCC is the one Sony designed for their needs. Nothing that I have seen has said that the Sony variant is the one and only configuration.

From IBM's Cell page

That leaves the question - how best to configure the Cell for Apple's needs? I can't think of what would work best. I guess we won't know until the patent is fully de-mystified.

 

[dontmatter]

So, IBM, Sony, and was it toshiba, I don't remember?

Why the three company alliance? Usual reasons-split the costs, get all the benefits, less risk, etc.

now... why would this new alliance let apple in on the deal, given that they weren't in initially? Apple would be getting a free ride, and would be competition. Cell is an attempt for high end pc makers to combat dell et al. Giving it over to apple would just be giving a hand to the only major company in the high end PC buisness turning a serious profit.

all I'm saying is, I read this the first time, and went, oooh, IBM, it'll come to apple. Then I kept thinking about the lack of apple, the lack of the mention of personal computers, in the otherwise EXTENSIVE list of possible uses for the chip, and the partnership...

And I frankly don't know if this chip even makes sense in a PC, exactly. But I can say, I'd be surprised if this was good news for apple.

 

[GFLPraxis]

So, IBM, Sony, and was it toshiba, I don't remember?

Why the three company alliance? Usual reasons-split the costs, get all the benefits, less risk, etc.

now... why would this new alliance let apple in on the deal, given that they weren't in initially? Apple would be getting a free ride, and would be competition. Cell is an attempt for high end pc makers to combat dell et al. Giving it over to apple would just be giving a hand to the only major company in the high end PC buisness turning a serious profit.

all I'm saying is, I read this the first time, and went, oooh, IBM, it'll come to apple. Then I kept thinking about the lack of apple, the lack of the mention of personal computers, in the otherwise EXTENSIVE list of possible uses for the chip, and the partnership...

And I frankly don't know if this chip even makes sense in a PC, exactly. But I can say, I'd be surprised if this was good news for apple.

Firstly, Apple already has an alliance with IBM and is the #1 PowerPC buyer in the world.

Secondly, Apple is working on a Sony alliance. Remember MacWorld?

Thirdly, Cell is using a PowerPC-based processor, remember? Apple is in on the PowerPC alliance (Motorola + Apple + IBM).

 

[alfismoney]

price is the main reason why apple will be using some variation of this technology sooner rather than later. if ibm can produce these in numbers high enough to meet PS3's needs while keeping costs below $300 for an entire console system, you guarantee that it will cause ripples in the processor market.

the dollar ends up being the bottom line here. if the chip costs a quarter of what the g5 does to produce (from the looks of it it will be much less even) then apple will develop the software to take advantage of this technology somehow. sure only power software will need to adapt for this but think of who writes it. avid/digidesigns will miss the wagon, as always. apple writes all their own stuff. if adobe refuses to comply they know apple is capable of producing/buying software that will cripple its market share. microsoft could refuse to port office but then apple will simply pay for development of OpenOffice. This isn't a switch that will happen tomorrow but it will be the case in the next 18 months.

 

[Evangelion]

My apologies if this has been posted already. Go to part 4 for some interesting speculation about Apple.

http://www.blachford.info/computer/Cells/Cell0.html

Like I said: take that article with a grain of salt. The whole article seems to be nothing but "oooh boy, the Cell is going to kick ass! It's just so damn great! You have no idea how awesome it is! Now, look here....". if you want really insightful articles about Cell, read Ars Technica.

 

[Evangelion]

Judging by what I know about Cell, and what was written in that post, I'd say that's 100% correct. The Cell may have the most amazing vector capabilities, but the vast majority of Mac software isn't vectorized and thus wouldn't take full advantage of what the Cell offers.

Well, GCC 4.0 will bring autovectorizarion to the table. And IBM is working with "open-source compiler" to add support for Cell in it. I wonder what compiler that could be.....

 

[trebblekicked]

price is the main reason why apple will be using some variation of this technology sooner rather than later. if ibm can produce these in numbers high enough to meet PS3's needs while keeping costs below $300 for an entire console system, you guarantee that it will cause ripples in the processor market.

not quite. iirc, sony, microsoft and nintendo all take a bath on hardware. they make up the loss in software sales and licensing.

 

[wrldwzrd89]

Well, GCC 4.0 will bring autovectorizarion to the table. And IBM is working with "open-source compiler" to add support for Cell in it. I wonder what compiler that could be.....

I meant to bring up GCC 4.0...

Good point though! GCC 4.0 will help everyone with a vector unit, including Cell.

 

[orangedv]

Why not? Unless they delay the XBox to let the PS3 come out first (yeah, right. They got killed on the first one, being last to the market), the PS3 will probably have better graphics. Cell + a bit more time for new things to come on the market = probably a more powerful system.

If the PS3 ends up with better graphics, and the Nintendo with better gameplay, the XBox 2 will go down the gutter.

Cant agree with you there fellah, if you follow the graphic card scrap between Sony, Microsoft, ATI and Nvidia, the current state of play is Microsoft have the ATI good stuff and Sony have the Nvidia leftovers.

Nintendo have always had better quality regarding gameplay than the others, but the competition use that as an advantage rather than a disadvantage. Stringent quality control is a time burner, meaning slow release of titles. Sony took the view with PS1 'lets let anyone release titles on this, flood the market with choice". It works too; the magazines and the net sort the wheat from the chaff of PS games and you walk into a gamestore and see PS games EVERYWHERE. Also, don't beat the xbox up too badly, version 2 is closer to your Mac than any other console will be, twin dual cpu G5's are the dev tool for this console. Also, over here in the UK the Xbox pushed the Nintendo system so quickly into third place in market share that the Uk's biggest supplier of consumer electronics to the public phoned Nintendo up and politely told them they were not going to sell their stuff anymore. That sent shockwaves around the game industry I can tell you.

 

[AidenShaw]

Will the Cell processor be the new Itanium?

http://news.com.com/Cell+chip+Hit+or+hype/2010-1006_3-5568046.html

This sort of excitement and speculation about chips is driven by what I call the "Battlestar Galactica" principle. It goes as follows: If the domination of the universe isn't contested on a weekly basis, ratings will go down. Analysts, reporters, consumers and even executives need a gladiatorial contest to keep the job interesting.

The high-public profile of Sun Microsystems can partly be attributed to its role as the William Shatner of computing--donning a new uniform every three seasons to battle a new nemesis.

Put in that perspective, the Cell story starts to look different.