Apple Switched Chips Too Soon?

Ctrl+Alt+De1337 writes "C|Net is reporting that IBM has announced a method of altering silicon that will allow its next generation of Power chips to run at speeds between 4 and 5 gigahertz, and consume less power as well. From the article: 'Instead of just making transistors smaller, IBM came up with a process to alter how silicon behaves by placing a layer of insulator underneath a layer of silicon less than 500 atoms thick ... The higher speed of the Power6 will be achieved with existing chip manufacturing technology that etches transistors only 65 nanometers wide, several hundred times smaller than a human blood cell.' These won't be out until 2007, but it still raises the question: did Apple jump the gun by switching to Intel?"

Apple too soon or IBM too late?

Apple switched because Intel offers a better deal right now. When IBM offers a better deal, Apple can pretty well switch to (note: switch to, not switch back) this new chip.

Apple would be silly sticking to an inferior product for 2 more years.

Re:Apple too soon or IBM too late?

I'm still waiting for the eight pound, 10-inch screen iPod that Steve Jobs announced during his podcast demo at MacWorld last month. :P

Re:Apple too soon or IBM too late?

That would be really good thing.

I've often wondered about that. We have a lot of asymmetrical multi proccessor machines where there's general purpose processor which offloads tasks to GPUs, signal processors etc, but are there any computers which use multiple diffent general purpose CPUs?

Amiga's Sidecar had an Intel CPU and ran DOS from within the Amiga OS using Janus software and there were bridgeboards that could run Windows for later models. How hard would it be to write an OS which could address both CPUs and pass instructions to the most suitable processor?

Re:Apple too soon or IBM too late?

I agree. The tech will be available in for the expensive servers in 07? Will be ready for mass production at the level Apple would need in ??? Will be cheap enough for the PC market in ???? Will be energy efficient in ?????

Remember, IBM likes to make high-end chips where the cost of the chip is secondary to raw power for its servers. That's where the new tech generally goes first. IBMs first goal isn't cost efficient processors for PCs -which is why the IBM/Apple "breakup" can be viewed as a good thing for both companies.

Re:Apple too soon or IBM too late?

I seem to recall it was also a matter of supply problems

That, and the impossibility of getting a G5 into a laptop.

Apple probably lost a billion dollars or more every quarter since the G5 came out, because of supply restrictions. It's a fine CPU, but we just couldn't get enough of them.

-jcr

Re:Good news

You misspeled Armageddon.

IBM and AMD makes this good for apple too

One of the reasons AMD caught up with intel was IIRC they liscenced IBMs Silicon-on-insulator technology to get lower heat dissipation. If IBM once again liscences this to AMD then you will have this technology running on 0x86 instruction sets. Or conversely, if it's a world beating technology IBM may be able to persuade Intel to liscence it's 0x86 instruction sets.

No matter how fast the chip is, unless it runs 0x86 it's never going to show up in home or bussiness computers. Windows is the glue that holds that enterpise together and unless windows runs on it, people wont buy it and dell wont sell it unless there's a market.

So Apples will probably by able to access this in the new 0X86 mode. but it's not going to be just a simple processor replacement since you also will need RAM and busses that can handle the suction this processors is going to have. So motherboards are going to have to be entirely redeisnged to cope.

So this is going to be good news for apple since they are an agile hardware manufacturer that is not locked into the PC motherboard paradigm and are free to create their own firmware and software to run on radical hardware variants.

Re:IBM and AMD makes this good for apple too

Windows may hold the Enterprise Desktop environment, but they are a far cry from holding the server market in business. IBM is the leader there. They build the top end, most reliable, and highly available systems short of a custom engineered product for complete failure detection. In short, IBM is doing more for large scale business than anyone else these days. If you were running a large company, say a Fortune 50 retailer, would you install a bunch of Windows servers and setup nightly reboot jobs to keep things running smoothly, or would you setup AIX servers with HACMP (IBM's high availability offering) to run even if the hardware dies, sites go down, net connections drop, etc... you get the drift. It's all about what people can afford and the best is IBM.

Unlikely they'll switch again

I rather doubt Apple would switch again. Intel will eventually develop a similar capability or will license the production of it from IBM. I rather doubt Windows laptops will be switching to PowerPC anytime soon so there's a long road ahead for laptop chips at Apple.

Apple didn't move because of the performance of Intel versus IBM, it was that IBM was very unresponsive when it came to making a laptop variant of the G5. Now that Apple's on the Intel ship, they'll benefit from working with a company that has a vested interest in developing laptop chips. Name me one manufacturer other than Apple that made PowerPC based laptops and you'll see what I mean.

Re:Unlikely they'll switch again

Do you mean Microsoft? Sony? Nintendo? Ford? BMW?

The PowerPC market is huge. Worldwide, half of all the cars manufactured this year will use at least one PPC.

Re:The beauty of Universal Binaries...

Universal Binaries are a transitional stopgap; Apple is, after all, referring to this as a transition to Intel.

Let's get a few things out of the way:

1.) This silicon technology isn't new, it's just the first news of it being rolled out in the desktop in major waves.

2.) This is IBM, who is famous for promising in press releases but never delivering. I still remember when the IBM guy said at WWDC '03 that the G5s would hit 3Ghz "by next summer."

3.) Apple isn't going to "switch back." For Pete's sake, how could anyone actually think they'd do it all again next year? Apple switched to have faster, cooler chips so they could update their Powerbook line. Portables outsell desktop machines in today's computer industry. They liked Intel's future low-power roadmap (particularly Merom). Steve Jobs originally considered x86 in 2000, and again in 2003 (but was dissuaded with the G5). Remember that Rhapsody ran on Windows NT for a while.

4.) Intel chips aren't magically going to sit still until 2007. Intel has already announced a dual-core 3.4Ghz Xeon with a unified 16MB cache, available this fall (AMD's fall server chips won't have unified cache until next year...they'll have two 512kb caches). And of course, Merom and Conroe are due out.

Re:Apple too soon or IBM too late?

Apple switched because they're sick of IBM marketing writing checks that IBM manufacturing can't deliver.

This vapor-chip is no different than the 1GHz multicore G4 (in large numbers) IBM promised when the first G4's were delivered.

Apple is probably not worried that IBM might actually deliver on this promise.

Look, I'd much rather see Apple stick to a multi-chip strategy, or just stick with PPC, myself. Because I think (and we're already seeing it) that even with VERY slick emulators and fat binary technology, switching from PPC to Intel is going to end up being a usability nightmare for non-technical users: (hey, your web browser still works, but not your plugins), and a compatability nightmare for technical users: (hey, your Photoshop/Video Editing/Audio Editing software still works, but not your favorite 5 year old set of plugins). But at the end of the day, there's only so many broken promises and marketing bullshit you can put up with from IBM. True - with the G5, it seemed, IBM was FINALLY delivering on the promise that was made when the PowerPC platform first was dreamed up in the early 1990's. Except that they hobbled the chip by getting rid of the litte/big endian translation, which made x86 emulation SLOWER than on the previous generation. Then they promised low heat and power consumption - making the high-end G5 Power Macs "whisper quiet" - until Apple learned that these machines were running dangerously hot, and had to patch the firmware to crank up the fans (yeah, I remember when I first got my dual G5, it *was* whisper-quiet. But not after the second OS update. . . ) - face it. Apple trusted Motorola, and got screwed. They trusted IBM, and got screwed. They know they can trust Intel, because if Intel screws them, then Dell, Gateway, and a zillion other manufacturers will go to AMD, and Apple can to. That's really the bottom line.

Re:Apple too soon or IBM too late?

But right now there are applications that can run on BOTH platforms (Intel and PPC). So, why not just keep the option open and have applications be Universal and then it doesn't matter what Chip is inside. It will just simply run.

I think the dumbest thing is to get locked into a chip again. I think it would be really genius to be able to have an OS that runs on anything.

Re:Apple too soon or IBM too late?

I'm sorry, but your post is incredibly uninformed.

NeXTSTEP was Mach + BSD with Display PostScript and objc libraries.

Mac OS X is Mach + BSD with Display PDF and objc libraries (the latter rewritten with CoreFoundation), plus Carbon.

Mac OS X is basically NeXTSTEP.

On a side note, I'm really tired of Apple fanboys who are unwilling to admit that these technologies are nearly 20 years old. Apple marketing wants people to think that Mac OS X is shiny and new. It isn't. They also want free software/open source enthusiasts to think that it has more to do with FreeBSD than NeXT. It's a bit frustrating to hear Mac OS X compared to Linux or even a typical BSD because they are really apples and oranges (no puns intended)

Re:Apple too soon or IBM too late?

Actually, it does not double testing costs.

You use the same installer, the same compiler, the same debugger, the same test procedures (even the same test-harness software - if you're that lucky). It doubles your hardware costs, and doubles the hours billed to actual test performance, but that's actually a small fraction of what's involved in a proper testing process. Unfortunately, most software vendors don't use a proper testing process, and only use the minimal components, which is why their cost is doubled, and utterly ineffective on either platform alone.

Re:Apple too soon or IBM too late?

But right now there are applications that can run on BOTH platforms (Intel and PPC). So, why not just keep the option open and have applications be Universal and then it doesn't matter what Chip is inside. It will just simply run.

The thing is, it doesn't just simply run. If you're application developer you now have to run a complete QA cycle on two totally disparate architectures. Or choose to develop on x86 or PPC and hope it works on the other. QA costs an absolute fortunate so most companies know and Apple certainly does that PPC and Universal binaries are a stepping stone. I doubt that if all but the most mainstream apps even pay lip service to it in a couple of years from now. PPC users will basically be left hanging out to dry and I seriously doubt Apple is going to produce a Rosetta for PPC anytime soon.

Re:Apple too soon or IBM too late?

I think it would be really genius to be able to have an OS that runs on anything.

You might want to download a copy of NetBSD then.

Re:They can switch again!

Who says Apple won't switch chips again?

I think it's pretty well established, even with this development, that Appple won't be switching chips again anytime soon. The move was more about laptop chips than anything else. Laptop sales kept growing and IBM kept making promises it couldn't keep. Intel had a solution available and Apple liked the product roadmap of future chips so it jumped ship. I doubt Apple would suddenly switch back because IBM might have a much faster desktop chip in 2007. Desktop sales will probably be even further marginalized by then and IBM has a well established history of making promises about it's processors that it can't keep.

Both supported

Doesn't mean Apple can't rethink and switch back. From the looks of it, right now they are supporting both platforms. Since they have the ability to go back and forth (with performance loss of course) I can see strategic changes as they see fit.

Re:Both supported

Doesn't mean Apple can't rethink and switch back

Don't you mean "think different"?

Re:Both supported

Grammer nazi says, "think differently".

Re:Both supported

Spelling Nazi says "Grammar".

-jcr

Re:Both supported

Bannkontrolleur National Sozialist sagt Grammatik

Re:Both supported

Das echte Bannkontrollier sagt: "Bannkontrolleur" sieht Französisch aus.

Re:Both supported

Actually. Considering the installed base of PPC Macs, it's likely that universal binaries are going to be the standard release from here on out - *universal* binaries, not "intel-only" binaries. So this will really give Apple a lot of freedom to "switch back" from a technical perspective.

Of course, considering what must have happened in the background when Apple did the switch - I've heard rumours IBM found out about the switch the same as all of us from the announcement at WWDC - whether IBM will be interested in Apple's business again in the future is a separate matter. It's entirely possible the only G5s Apple has are ones being delivered under the pre-WWDC contracts and that the iMac G5 is being retired so soon not only because they WANT to switch over to Intel ASAP, but because they don't have the G5s to put in 'em. IBM's probably been busy retooling the fabs to make Cells for Sony et al and won't be switching them back anytime soon...

or is it just an IBM PR stunt

with all the bad PR they've been getting related to Apple's switch, this seems more likely something a desperate PR department would do.
I mean one and a half years is a lot when it comes to CPU research & development, why tell everyone what you're doing?

Re:Begs the question?

Yes, just like "hacking" and all the other words people have taken over to mean something different.

For all intensive purposes, I could care less.

Apple Switched Chips Too Soon?

No, it was the best thing to do, instead of having one company as a supplier they now got at least 2 , AMD and Intel. I think we get better and cheaper Apple boxes out of the x86 move.

Its not all about performance either, its the ability to ship large quantities of chips also, if you want to grab a larger market share.

Re:Apple Switched Chips Too Soon?

Technically, didn't they have two before? IBM and Freescale?

Universal Binaries

I thought the idea of Universal Binaries was that the packages were compiled for multiple architectures, selectable at runtime? The same binaries are now running on Macintels and G5s, so Apple should be able to continue running apps on either architecture...

Not so fast

IBM was facing a pretty big loss of business, and would have let Apple know before the official announcement. Apple knew, and decided to switch anyway.

Less than 2% of one fabs capacity

Actually, chips for Apple accounted for less than 2% of the capacity [nytimes.com] of just one IBM fab. IBM's tech division (which does chip fabbing) accounted for less than 3% of IBM's total revenue. That's a really small piece of IBM's global business. It's kind of like an oil company losing one gas station...not really gonna hurt them that much.

Re:Less than 2% of one fabs capacity

And I think that's really the point: IBM didn't care much about the success of their desktop/laptop processors. They care about servers. They care about the embedded market. Desktops? That's not a big concern right now.

So it doesn't matter if IBM has some new tech in 2 years. Their tech, if it materializes as promised, will be focussed on server and embedded markets. Intel's chief business, however, is making processors, motherboards, and associated devices for PCs. In two years, they'll also have better chips than those currently available, but for desktops/laptops/portable devices. That's who Apple wanted to cozy up next to.

Nothing new

SOI is nothing new. It's been around for decades for radiation hardened ICs used in space and military electornics. The only news is that it is now being considered for large scale commercial production. IBM has been hinting at a transition to SOI for years and rest assured that Apple planners were well informed of this when they made the decision to switch.

This is strained silicon, though.

SOI just makes the chip run cooler.
They are talking about strained silicon, which makes the electron mobility larger in one direction. Intel, in fact, is working on that too, as are others.

Chickens and Eggs

Don't count your chickens before they hatch. IBM's chip is theoretical and not in production, Intel's is here now. The better question is, why didn't Apple switch to Intel chips earlier?

Too Soon?

I thought the major impetus for switching to Intel was the supply and timely delivery of the PowerPC (or lack thereof). IBM was not willing to meet Apple's requirements. There is no guarantee they would meet them with this chip, either.

So no, Apple did not move too soon.

No way.

No way did Apple jump the gun. Both Apple and their users wanted more speed, especially in their mobile products, and Intel delivered on that today - not in mid-2007. I see in no way how a chip process that wont be available until 2007 is compared against a decision made in 2005. I think anyone in the right mind can see why Apple made the switch in the first place.

Apple hasn't switched

This is one subtle, yet important distinction. Apple has added the intel processor to their lineup, but they haven't abandoned the PPC architecture. Although Steve et al. have implied a complete switch through the various pr statements made on the subject, Apple could just as easily stick with both chips indefinitely. Or they could retreat back to PPC if intel suddenly died and IBM came out with a blockbuster. That is of course if Intel doesn't lift this technology from IBM for their own chips in the future. Which they will.

I don't see a difference

The G5 was an amazing chip when it came out. But the speeds stagnated, the heat production was incredible, and they could never design a laptop chip.

How is this chip different?

And what would it cost?

With Intel, Apple gets a low-cost chip that they can use NOW, in their laptops and desktops. They get low-power consumption today, and low-heat today. Not in 2009, when the POWER6 chip has been tamed... Or hell, maybe never, AGAIN.

So yes, this seems like a good chip. But it doesn't really affect the reasons that Apple changed. It doesn't say it's a good chip for laptops, and they would still need to change the architecture of their systems. AND they'd have to stick with a company that was creating lower yields.

Plus, this writeup makes it sound like IBM didn't tell Apple that they were about to make POWER6 chips. I'm sure they knew, and I'm sure they realized the advantages and disadvantages.

Not really.

The Power6 chip will compete against offerings from IBM rivals such as Intel, Advanced Micro Devices and Sun Microsystems...But the process also tends to make chips run hotter

So these are server chips. The area of Apple's lineup that was suffering the worst was their laptop line. These breakthroughs from IBM don't address that at all.

Reasons for switch...

Of course it was reported all over that the reason for Apple switching to Intel processors was because of speed and power consumption -- this is what makes consumer happier ("Hey, an extra GHz!".

But another reason was that Apple was VERY unhappy for a while with the rate IBM produced PPC processors and their rather poor chip yields. Introducing more exotic SOI process would not help keeping these yields up, for sure!

We will see if IBM will be able to fulfill demand for PS3 Cell processors -- I wish them best, but...

Paul B.

No.

No, Apple did not switch too soon.

Remember, we (the loyal Apple customers) have been waiting for a significant increase in computing power within the portable market.

IBM made promises to Apple but were unable to deliver on those promises. Remember the statements about 3 GHz within a year? [pcworld.com] Apple couldn't sit by while IBM broke promise after promise on upcoming product lines.

If Apple had waited any longer, they would have lost momentum in the portables market, and in turn the desktop computer market, eventually pulling down the servers and everything else with it.

On the other hand, Apple could always keep their servers on the IBM product line. I doubt they would, but it's always a possibility. Apple might just not be done with the PPC for good.

Answer: Depends on what they want

If they want speed, then the answer is "maybe" - but then again, Apple could have considered AMD (please, fanboys on either side, before there's an Intel v AMD argument - just shut it).

However, if Apple is going for more than speed, and wants Intel's DRM technology, their vivo (I think that's the acronym) certification for projects that would make Hollywood happy, and other things to allow the company to cozy up with the entertainment market - then Intel was the right choice.

Personally, I'm pleased with the Intel switch. Speed is looking up, once Wine or an Intel virtual PC is up and running that lets me play Half-Life 2 at nearly full speed I'll be set with my games, and besides, IBM had how long to get a G5 into a laptop and couldn't deliver?

So while IBM's technology looks pretty damn cool, I'm not worried about Apple making the "right" or "wrong" choice. As long as my apps and terminal work on my Powerbook (oops - sorry, "Macbook Pro"), then I think I'll be all right.

Of course, that's just my opinion. I could be wrong.

the switch was about money not technology

I work in a world where a variation of the PowerPC drives a business. From iSeries (AS/400 new name) to xSeries and eventually the pSeries. The processor and the technology behind it are simply amazing. We went from 48bit to 64bit computing in the late 90s without recompiling or any such nonsense because iSeries engineers separated the processor from the OS. The tech has always been there. We have PowerPC powered thin clients as well - fanless to boot!

Switching to the Intel platform allowed Apple to get those sitting on the fence waiting for the next greatest thing to have a reason to buy a new Apple computer. It will even garner more buyers from the previously Intel-Only world in the form of linux and windows geeks. Continuing the PowerPC line would not generate the boost in revenue Steve needed. There are only so many variations of the iPod they can crank out before someone either starts to truly compete (overseas the iPod saturation level is only near 40%) or the market moves to further integration perhaps out of Apple's area of expertise.

I know its working, almost everyone of my friends who have Macs are going to buy into the new machines. The laptops are where its going to be the biggest until the mini comes out intel flavored. After that IntelMini comes out I expect another surge once someone shows Linux and Windows running on it easily.

Two things:

One) It doesn't "beg the question". Begging the question is a logical fallacy in which you assume, implicitly or explicitly, the very thing you are trying to prove.

Two) Apple primarily switched because the laptop-suitable G4 line speeds had been stagnant forever. Freescale's 7448 is over a year late and counting. PA Semi's everything and the kitchen sink promises are still vapour-ware. And IBM couldn't provide a G5 that ran cool enough to put in a laptop.

This technology won't be out in the Power6 until 2007 if everything goes as planned, a never-safe assumption when it comes to IBM's fabs. Add more time to that for them to retool the Power6 into a desktop-suitable G6. So in return for not switching, Apple would have to leave their desktop speeds stagnant for another year, and still have no guarantee of any new chips to offer in their laptop line.

Selling 1.42 Ghz, 133 Mhz front side bus iBooks is tough enough now. They'd have had to be absolutely suicidal to stick to IBM's roadmap and the near certainty that they be trying to sell the exact some mobile processors in late 2007.

Probably not too soon...

Sure, IBM just came out with better chips. But if Apple didn't switch, it would probably be only the 4th largest customer for IBM in the next few years.

Apple would have had to wait in line behind Microsoft, Sony and Nintendo. Xbox 360, PS3 and Revolution are all going to use IBM chips. These 3 systems will use the same chips for years. So once things are running, it'll be an easy job of IBM to supply them. Apple, in the meantime, will be constantly asking for faster and better chips from IBM.

If you were IBM, would you like to deal with 3 easy customers or one tough one?

Keeping One's Options Open

You'll know Apple is keeping their options open if they extend the life of their current G5 PPC machines by bringing out new models. Sure it will be explained as, "Some of our customers can't/won't switch to Intel yet so we're continuing to support their demands." (If Dell actually supported their own customer's demands, they'd be selling AMD64 processors long since.)

So that's what to watch for. Any extension of the G5 line. Anything so much as a bump in processor speeds will give Intel some well-deserved heartburn.

And remember, the only Apple Intel machines currently available are 32-bit models. And it looks to stay this way until at least mid-year. For the life of me I cannot understand why Apple wants to support both 32-bit and 64-bit Intel machines in addition to 32-bit and 64-bit PPC machines. That's a huge drain on resources -- especially when you are not only not nearly the biggest player in the field, but won't be anytime soon. And all your software partners are also going to be required to support 32/64 as well. I'm surprised SJ hasn't been assassinated by his own operating system engineers by now.

it's not like Apple didn't know this was coming...

Does CNET really think that in private meetings with IBM, this technology wasn't discussed months if not over a year ago, with Apple? I love how the press thinks that when THEY find out about it, the rest of the world is first hearing about it too...

"Stupid Apple", they chant. Except:

• IBM has said they can do it, but not for over a year. Intel is here, now, shipping.
• IBM has historically had problems meeting supply. Intel doesn't.
• IBM has made it clear they don't care about Apple- they were running around telling everyone how Apple represented a single-digit percentage of their output of PPC's. Not a good sign when your supplier is dismissing how 'trivial' you are.

Maybe these Power chips will end up in Xserves or something...seems fairly unlikely though.

Non-Disclosure

Most companies brief their customers (under non-disclosure) about upcoming products before they are announced to the public. Especially large customers like Apple.

I'd wager that Apple knew about this long before they decided to switch.

Strained Silicon On Insulator

TFA is imprecise enough that it's hard to be sure, but seems to be talking about a combination of Silicon On Insulator (SOI) and strained silicon. In the x86 world, neither of these is terribly new.

Intel announced their use of strained silicon [intel.com] back in 2002, and I'm pretty sure all new Pentiums for at least the last couple of years have used this technology. It's essentially certain that every Intel-based Macintosh already uses strained silicon in its CPU.

As an aside, TFA only talks about "squeezing" silicon, but it's actually possible to either tighten or loosen the lattice. CMOS uses complementary pairs of NMOS and PMOS transistors, and for best results you (normally) want to strain the silicon in opposite directions for each -- though NMOS generally has slightly better characteristics to start with, so IBM may have decided to apply the strain only to the PMOS transistors (or the article may simply be incomplete, and they're really doing both, just like Intel and others do).

OTOH, AMD has been using SOI [amd.com] (also since they went to 90 nm). I'm reasonably certain that all their current x86 processors use this technology. Their dual core processors certainly do, though some of their low-end processors may not use it (I'm afraid I've lost track of which cores use what technology anymore).

What IBM has announced is (apparently) successfully using both of these technologies in the same chip. AFAIK, that hasn't been done in an x86 CPU before, but it's not entirely new either. One thing that should be kept in mind is that x86 CPUs are (mostly) built for the mass-market -- that means using fabrication technology that you can dependably produce in large quantities with decent yields. The IBM POWER series chips have a drastically smaller market and substantially higher price tags. A yield level that's perfectly reasonable for that market would virtually put an x86 supplier out of business. As such, both Intel and AMD are somewhat conservative in what they use in production chips, as opposed to what they can manage to do under lab conditions and such (though their volume also lets them put lots of money into R&D to really push the technology as well).

Re:Apple had its own reasons...

Apple's claim that Intel won on watts has been thoroughly discredited in the press and in the blogosphere.

It has been discredited everywhere except in reality. IBM had no good competitor to Yonah and Conroe. The G5 was a long-pipeline, high-frequency design, and it just plain ran too hot for a laptop. Yonah is offering integer performance competitive with the top 970MP, with a power budget 1/3 the size and a CPU die about half the size. POWER6 is just another step in the wrong direction as far as Apple is concerned. It's got a higher frequency, longer pipeline, lower IPC, and an even worse INT/FP performance balance than the G5 had. It's the Pentium 4 all over again. Perhaps POWER6 will be the Pentium 4 done right, but no matter what, its not going to be a good chip for Apple's machines. Especially when you consider what will happen when you take a long-pipeline (inherently bandwidth hungry) design like POWER6, which is optimized for 32GB/sec of memory bandwidth and tens of megabytes of cache, and stuff it into a PC system with 8GB/sec of memory bandwidth and a power envelope of 60W.

Re:Apple had its own reasons...

The game console cores suck. They are 2-issue in-order designs with crappy branch prediction. Initial reports suggest that they are barely fast enough on integer code to keep the FPU fed, and that's with low-level gaming code. God help you if you're trying to run generic, unoptimized C code on it.

It's 2006 --- no programmer of desktop/workstation/server programs is going to spend time optimizing their code to make up for a flawed processor design. It's 2006, and a few things have happened that apparently no-one told the "Cell on the desktop" folks about:

1) Programs are becoming platform-agnostic. Especially at the workstation/server level, many important applications run on multiple platforms. This often means they are not highly optimized on any platform. This was always one of the things that held the G5 back --- it's high theoretical performance was often nullified by its reliance on tight, well-scheduled code tuned to its idiosyncracies. Super-optimized apps is a luxury few users have. Hell, as an engineer, much of the code I write runs in Matlab's JIT. You think that does G5 optimizations? A processor that does not run all these minimally-optimized apps well is not going to fly on the desktop/workstation.

2) The world is moving towards higher-level languages and higher-level programming constructs. If your CPU can't run machine code with whatever optimizations the JIT can spit out in 100 milliseconds, it sucks. As someone who does a fair bit of programming, I love the Opteron for one reason: it doesn't care how much my code sucks (from a performance standpoint). It lets me write clear, clean code, and runs it with decent performance. I don't have to drop into SHARK to figure out why my 5-issue processor is behaving like a 2-issue one because of instruction scheduling issues, I don't have to sacrifice virgin blood on the alter of code alignment, and I don't have to bust out Altivec to get good FPU performance. Programmers in the desktop/workstation/server markets have gotten used to processors that serve the software, not force the software to serve the hardware. A 2-issue in-order core is not going to fly with them.

3) Vector performance has largely become irrelevent except in a few markets. Yonah has shitty vector performance, and nobody in x86 land really cares. Most desktop CPUs these days spend their time running integer logic code, or double-precision floating-point, letting the heavy vector lifting be handled by the GPU. As API's like CoreImage/CoreVideo take off, things like VMX and AltiVec will become still more irrelevent, except perhaps to those people running FFTs all day long.

Did Anybody Read the Article?

The article does nothing to speculate on whether or not this is an indication that Apple made a bad decision switching away from IBM. Apple is not even mentioned once.

The article mentions that the process makes the chips run hotter, and that engineers are trying to figure out how to counter this so that the chips don't fry themselves.

Decent article, bad post. Still sounds like Apple made the right decision.

Please don't confuse the POWER with the PowerPC

Call it unfortunate naming, but these two processor families don't really have much in common (other than possibly some marketing material). A POWER processor is the stuff dreams are made of. See http://www-03.ibm.com/systems/power/ [ibm.com]. A PowerPC processor is the stuff printers are made of. And until recently; Macs.

Laptops, laptops, laptops.

Power6 is a desktop/server chip. Laptops started outselling desktops [yahoo.com] last year. Intel is offering relatively fast, low power chips.

Ergo, the answer is no. Apple did not switch too soon.

That's nice and all...

...but nothing in the article said IBM could sell those miraculous CPUs at a price affordable enough for consumer computers.

IBM reserved their best performance for themself

Yes, from a technical perspective, I am inclined to say that Apple's switch from the PowerPC was not necessarily a brilliant move. However, the real reason for the switch was in my opinion this:

Apple could no longer live with a processor manufacturer that reserved its best performing processors for their own use

IBM has a huge business of their own to protect, making servers and workstations using the same technology that Apple does. IBM's issue is that these systems are priced at 2 to 4 times higher than the same performance from Apple. This became very evident when Apple shipped the G5 Xserve and completely undercut IBM in large cluster configurations (which is clearly IBM core markets.) Why has the Xserve not yet shipped with the dual-core IBM 970MP? Why has Apple never shipped anyhthing but dual processor machines even if it was possible all they way back to the PPC 604 days to build 8 way systems. IBM had them. No coincidence if you ask me.

Intel does not have any such hangups og dependencies. Intel is all about delivering its best performing processors to those who can build systems from them.

Intel will even throw marketing efforts into the equation -- something IBM never, ever did to help Apple promote the PowerPC plattform. I think IBM's - and IBM Software's complete lack of support for Mac OS X is a telltale sign [andwest.com] why Apple had no choice but to switch even if the PowerPC/POWER processors at the technical level perhaps would be better.

RTFA

If you RTFA you will see that while speeds of 4 to 5Ghz are possible the chips also run hotter therefore the switch to Intel, especially for small form factor machines like the laptops, no to mention the Mac Mini, was still a step up from the G4 without the heat issues of the G5. The Power6 architecture might be great for a PowerMac however.

And the Cell processor is almost as pie in the sky, until there's some real information about the Cell everything is just conjecture and hope.