Visit Your Local PBS Station PBS Home PBS Home Programs A-Z TV Schedules Watch Video Donate Shop PBS Search PBS
I, Cringely - The Survival of the Nerdiest with Robert X. Cringely
Search I,Cringely:

The Pulpit
Pulpit Comments
April 25, 2008 -- Apple to the Core
Status: [CLOSED]

Bob, I think you're underestimating the value of the virtualization software (Fusion, Parallels) that the X86 architecture brings to the Mac platform.

Given that Windows is inextricably tied to X86 (for the time being), you couldn't efficiently run Windows under those virtualization tools unless they're running on X86 hardware.

I can't see this changing any time soon. Maybe when/if Windows becomes irrelevant, but that may take 5-10 years.

Chris Ryland | Apr 25, 2008 | 2:44PM

Are you sure that this is not an attempt to get away from white box Mac OS clones. If Apple can build a chip optimized for OS X they won't have to worry about the hackers building better, cheaper Mac boxes, since they won't run as well.

Tim Oppenheim | Apr 25, 2008 | 2:49PM

Nice take

t-maxx cowboy | Apr 25, 2008 | 2:51PM

The preceding comment about the advantage of an x86 chip for virtualizing Windows is well-taken but Apple only needs to provide this security blanket for a while -- most folks switching from Windows to Mac OS X end up giving up Windows fairly quickly, so when they buy their NEXT Mac, the x86 thing may not matter.

I think part of the equation is that Apple would like to have truly proprietary hardware (again) and not have to deal with the possibility of a clone market.

Tonio Loewald | Apr 25, 2008 | 2:57PM

So if I'm understanding you correctly, Apple bought the CPU maker not because of some freaky business strategy or conspiracy theory, but because they actually want to make their own CPUs for running Apple machines (likely the non-desktop/laptop variety like Apple TV and XServ).

This sounds reasonable, but can Apple build a CPU better than the likes of Intel, AMD, and IBM with a small $278 million company? Sounds like a big stretch to me.

boxlight | Apr 25, 2008 | 3:06PM

Bob. Hammer. Nail. Head.

William Donelson | Apr 25, 2008 | 3:11PM

The begin Clone Wars has.

Steve Dean | Apr 25, 2008 | 3:11PM

Confused by Tonio's comment - the old PPC chips were not proprietary to Apple. It's always been Apple's lawyers that keep people from making clones, never their chips.

Dave | Apr 25, 2008 | 3:12PM

Interesting...my first thought was that this had to do with either:

a) Lowering manufacturing costs on iPods, which are no longer the rabid profit centre they once were (future profit increases will come from reducing manufacturing more than anything else.)

b) h.264 decoding hardware encrypted into devices.

Darcy McGee | Apr 25, 2008 | 3:25PM

Blurg. Didn't mean "encrypted." meant "embedded"

Darcy Mcgee | Apr 25, 2008 | 3:28PM

One thing on x86 virtualization, though. Windows (or any OS) will always run faster on native silicon, virtualization or not, period.

So the idea that Windows will be just as fast on RISC-based chips as CISC-based chips because of virtualization is not true. Yes, you can run virtualized Windows on PowerPC, Itanium, Alphas, or a Wii, but there will always be a significant translation overhead compared to running (virtualized or not) the silicon on which it has been compiled to run.

If Apple switches to a custom PowerPC based chip, you can kiss your Parallels or VMWare'd Windows performance goodbye. You could throw more cores at it, sure, but it would definitely be an uphill battle. And one Apple probably doesn't want to face any time soon - significant gains market share are coming in large part from Windows switchers and/or Bootcampers. That would certainly put a kink in this nice little perfect storm they've created.

Will | Apr 25, 2008 | 3:37PM

Seeing how well virtualization works with Parallels and VMWare on my 4 core MacPro, and hearing how Intel and others are starting to put virtualization instructions on chip, I've been wondering how long it'll be before OS's become like windowing systems in *nix. Base system loads from chip and then you decide which OS to run on top of that. And of course, you can run apps written for any OS within it, just running other OS's in the background.

Gilmoure | Apr 25, 2008 | 3:38PM

Not telling you what to do or anything but do you have any interest in the energy industry? I have a feeling things are going to get busy there very soon. And as always I can't wait for your next arty.

Fred X | Apr 25, 2008 | 3:50PM

If Apple makes its own CPUs it can lock down the hardware again; right now its not difficult to make a Hackintosh which typically uses hacked versions of OSX from torrent sites; Apple makes nothing in that case unless the user eventually switches to real Apple hardware.

geekraver | Apr 25, 2008 | 3:51PM

Will,
You are forgetting several things:
- full cpu emulation is simply a performance issue
- if you have enough margin, efficiency does not matter - look at the move to byte code in .net and Java
- if you control the cpu architecture, you can design for virtualization and specific emulations adding to the instruction set and register architecture etc.
- the importance of Windows will continue to decline.
- Windows is only an issue on the desktop, not portable or set-top devices. OS X is...

I think, Cringe is on the right track here.


HaraldS | Apr 25, 2008 | 3:53PM

I have to disagree with Cringley on this one.

Apple would have become a competitor to Intel for microprocessors, in the same way as Sun is today. This makes no sense for the desktop, notebook, or server markets.

The portable market, iPhone, Touch, and maybe future products with larger screens, will need fully integrated solutions. This market wants a single chip, with one or more CPUs, graphics, memory controllers, networking, and everything else except for DRAM and Flash memories, in a single chip.

I think Apple will create their own ARM based solutions, since Apple uses Arm and the chip company has a lot of experience designing ARM and other PowerPC cores with low power. These will be fully integrated systems, minus memory.

Dogzilla | Apr 25, 2008 | 3:57PM

My understanding of the x86 ISA advantage is that it costs the same amount to design and more importantly fabricate a device for a given performance level, but you can sell a lot more units of x86 compatible CPUs than any other ISA thus granting Intel economies of scale its competitors can't match.

Having a CPU design house doesn't help you with the most expensive part of chip production and if you don't sell enough units no fabrication house is going to manufacture your chip on cutting edge processes because it's not worth it. So threatening to jump off the Intel ship seems to me an empty bluff since there is no way you can match their performance at the same price.

The only way out of this trap is to stop going for total performance assuming your performance is already good enough at an older process level. While this may be true for personal computers (what can an 8 core machine do for the average user that 2 couldn't?) the performance battle for cell phone and similar sized portable devices is only starting. If Intel can cross sell a mobile device chip into the desktop replacement market (and Centrino is already used in low end desktops) they can still beat any other fabricator's cost at a given performance level.

In other words, if Apple's goal is to reduce costs I can't imagine how PA can make a credible threat to Intel, unless Apple sells chips to Dell or HP too which prevents the OS X lock down idea. Of course I don't have any good idea why Apple bought PA Semi, but then again I'm not Steve "the apparently infallible" Jobs.

Q-Tip | Apr 25, 2008 | 4:13PM

Apple's been funding LLVM, a very nice virtual machine (etc), and actually using it in products. Now perhaps they want to make their own processors. What if the new processor implements LLVM in hardware? Less power, fewer transistors, and Apple software already written...

Rod Price | Apr 25, 2008 | 4:19PM

I've been reading comments here about how Windows will always run faster on native silicon. While this is probably true, it doesn't matter. With the speed and number of cores these days, you only need it to run fast enough. You don't need the fastest possible speed--only a marginal speed which is quick and powerful enough to get your work done.

Scott Johnson | Apr 25, 2008 | 4:24PM

Not likely Apple will make the mistake of going to another botique cpu maker, even if the botique is Apple. The last decade has been the story of every cpu arch that isn't x86/x86_64 dying a painful death with the sole exception of arm in the handset/embedded world and even those days are likely drawing to a close as x86 System On Chip solutions are beginning to encroach.



Apple will not leave Intel either. Forget technical arguments in the Intel/AMD competition because they don't matter here. Apple doesn't sell software. Apple doesn't sell hardware. Apple sells a 'brand experience' to upscale customers with lots of disposable income. Intel has the brand ID with the non-technical sort who make up the bulk of Apple's base and AMD is seen as the upstart cloner by that demographic. Again let me try to avert the flamewars by noting that the truth of the matter isn't under discussion, only perception by the NON INFORMED matters.



As for moving to a more controllable platform to stop the hackintoshes, good thinking but nope. A totally closed platform is just as easy to create with an x86 CPU on the board as one with something else. All Apple need do is actually start using the TCPA chip.



But for now the hackers are seen as a net positive much like Microsoft tolerated widespread piracy until they saturated every other market on the logical notion that it would be easier to eventually extract revenue from bootleg Windows users than from Linux users. Anybody who builds a kachintosh is going to fall into one a few groups:




  1. Just curious about OS X. Either they will play around and move on to something else or they will join the Cult. A Mac sale is probably in their future in that event.
  2. They want/need a configuration Apple doesn't want to sell. Allowing those who REALLY need it to remain in the Cult is still a win because they probably have bought Apple before and probably will again. And if only a few build hackintoshes it still makes better economic sense to not make configurable/budget machines available at the Apple Store and threaten the totally sweet margins they currently enjoy.

John Morris | Apr 25, 2008 | 4:43PM

As a small market share player, there is a huge benefit for Apple to use Industry Standard chips...lowest costs and parity with the competition. Apple can always sell boxes at a premium based on its sexy case designs and slick OS. However, there is also a huge disadvantage: cloners. Apple's inability to stop OSX86 is the reason why Steve is wooing a proprietary chip maker. Steve doesn't like clones and this is a hedge strategy in case cloning takes off the way jailbreaking iPhones has. This card won't be played unless it has to, but this is Steve's ace in the sleeve against erosion of hardware sales due to cheap beige boxes capable of running OSX. A couple of hundred million is chump change, and of course, it keeps Intel from getting uppity.

Larry MacPhee | Apr 25, 2008 | 4:44PM

Interesting analysis, but when you say...

CPUs optimized for OS X would be smaller, cheaper, and use less power than any Intel or AMD alternative simply because they could be simpler overall.

...I think the there's a bit too much magic hand waving going on. Maybe someone with more hardware knowledge than I can explain why Apple would switch from PowerPC, a simpler instruction set, to x86, a more complex instruction set burdened by decades of compatibility hacks. If the processor design was so "bad" comparatively, then why can Intel keep it competitive?

Kevin Watters | Apr 25, 2008 | 5:03PM

Has anyone thought of the fact that Apple might sell Macs with PPC chips, that would run OSX, say, 2X faster, and allow other large manufacturers (Dell, HP) to sell OSX on Intel chips, if there was a significant speed difference?

Apple is sitting very well right now. Who else can run their OS on multiple chips and still control the OS and the hardware it rides on? No one.

Apple is set to explode. I wonder if certain businesses (IBM, and now salesforce.com) aren't pressuring Apple into the scenario above: let Dell sell cheap Intel OSX boxes while Apple gets to sell faster Macs running PPC chips. Who knows what could happen? I manage 55 Macs. Only two of us need Windows running in Fusion. So why should we have to buy an Intel Mac?

Blad_Rnr | Apr 25, 2008 | 5:09PM

Daniel Dilger at roughlydrafted.com has a similar analysis but with different conclusions. As Bob states, x86 chips are commodities and the competition between AMD and Intel should keep Apple's costs low. Certainly Apple is not going to have any impact in this market with a small fabless chip maker. They can always switch to or include AMD chipsets in their product line if they're looking for leverage on Intel. That's how commodities work. Apple left PowerPC for x86 to benefit from the economies of scale and product roadmaps of the larger PC chip market. A new PowerPC chip for Macs only makes sense well into the future if Mac sales volume continues to increase as it has in the past couple of years (Bob seems to have hinted as such).

I think Daniel is more correct to guess that Apple intends to use PA Semi's highly qualified engineering team to build Mac-only hardware acceleration chipsets and specialized embedded chipsets for mobile devices that make it much more difficult to clone iPod and iPhone designs. They can use their unique position as an OS and hardware maker to make OS X perform best on Apple hardware, leaving Mac-clone performance in the dust. Building in a true advantage for Mac hardware is a better approach than locking down the hardware or software. Locks can be picked. Non-commodity proprietary hardware that actually provides value is not so easy to replicate. Unless you have a team of talented Palo Alto engineers.

davidLBC | Apr 25, 2008 | 5:13PM

HaraldS,

I agree with you; however, keep in mind the added cost of developing additional chip logic just to support Windows would be significant and necessary to be equivalent to the speed with which one can currently run virtualized Windows on a Mac.

.NET and Java are good examples of effeciently executed byte-code, but virtualized hardware is MUCH less effecient. There's even a noticable difference in performance of .NET apps vs. Java apps in Windows - the .NET apps are just plain faster because they are WinX86 specific.

Sure, the desktop will be much less significant in the future and Windows along with it, but unless some amazing strides are made in hardware and software virtualization very quickly, it's too much of a stretch to expect x86 virtualization to run smoothly on a different chip architecture. The performance margin is WAY higher than the margin when executing bytecode.

Will | Apr 25, 2008 | 5:18PM

There is a huge disconnect between high level programmers and lower level hardware architecture because it is costly. Although Jobs says OSX was running on Intel years before and Cringley's opinion that OSX would run faster on dedicated hardware that could take less then the original 2 years to amass -I find this theory has holes.

Even 'native' Intel applications hardly use the vast instruction set in X86 hardware. MMX anyone? It took a very long time for users to see any benefits of Intel's enhance instruction set for multimedia applications dubbed 'MMX'. The sticker on the front of the machine was nice, but the software didn't take hold for years.

Case in point: When getting real work done in accounting, or architectural engineering or say video production, one uses a software application that runs on an OS that runs on hardware, right? The problem is that if that low-level engineer fails to take advantage of the hardware all applications above it suffer.

Adobe AfterAffects and Premier (used in video editing) for instance has a 2gigabyte barrier this is inherited from the OS on Wintel boxes. Why? Because XP can only address 2-4gigabytes max! Now that puts in kink in your next 4.7 gigabyte DVD -and don't get me started on Blueray or HD content.

Now Vista with true 64bit has hit the market breaking that memory barrier but the applications have yet to take advantage of the OS. The same is true for the compiler that made the OS in the first place. We experienced the same effects back when 8 bit code went 16 bit and that dreaded 64k barrier stopped new application development in its tracks (Commodor128 anyone? Naw... its all about the Commodore64!). There were more applications for C64 that got the job done that held back 16bit applications from taking hold.

This delay is costly because labor is not cheap and the lower level programmers that create the compilers are always one step behind the hardware, and the high level programmers are one step behind the low-level guys, and so on, and so on.

So how did Jobs get OSX ported so fast? Because the OpenSource community where labor is quite a bit cheaper, already did it for him in ‘Darwin’ (the code name for the Open Source guts of OSX). Darwin was around for years prior and was all about the X86 hardware platform from the start. OSX was always on X86 to begin with! Now the switch to processor Y, WILL take more then two years to amass –and that is something that’s gonna create a very big bottleneck for high-level programmers and end product developers. Getting users to buy applications that run in emulation or slower then normal when it worked just fine is a very big hurdle. I doubt and Apple low level processor switch would be as succesfull this time around. In face, it could be a disaster.

Steveorevo | Apr 25, 2008 | 5:23PM

The one big gotcha in this scenario is support for accelerated graphics in Windows apps. There are enough applications that are run in Bootcamp because they need DirectX 9.0c that I don't think Apple can easily move away from being able to run Windows natively right now. If virtualization systems add support for 9.0c (or if Apple produces virtualization that can run Windows with 9.0c support), then the game changes and virtualization and non-X86-compatible chips become quite feasible.

Brian Gilstrap | Apr 25, 2008 | 5:30PM

Hmmm... I'm actually thinking that PA Semi designs would be used as "accelerator" processors for AV codecs: offload encoding and decoding to dedicated hardware, so the computer can do other things, like show advertising based on the movie/scene/actor etc.

Paste M&Ms over the Reeses Pieces in "ET" fer instance.

Tony D | Apr 25, 2008 | 5:53PM

According to Mark Cuban, video over/on the internet will be a 3 to 7 TRILLION DOLLAR PER YEAR business (every year).

Jobs aims to capture most (all?) of that 7 TRILLION dollar a year market. Makes iPod and iPhone look like chicken feed.

Jobs loves H.264 HD - which requires a heavy duty computer and/or graphic card and /or hardware based decoder. Imagine ..just imagine rolling it all into one chip at 1/10 th the cost. BIG PROFIT.

At least Jobs isn't stealing Xerox's property this time. ;)

Harry | Apr 25, 2008 | 5:58PM

My 2 cents. The reason Apple went to Intel x86 and away from the PowerPC was all about laptops. IBM was focused on making CPUs for servers, game machines, and supercomputers. Number crunchers. Intel was going for lower power.


More importantly, it was not about just the CPU chips, it was all about the support chips - the onboard northbridge/southbridge "glue" chips that are needed to make a complete computer. Intel can provide Apple with the complete set...CPU, support chips, and graphic chips. Before Apple had a whole team focused on building custom designs to make the PowerPCs workable as desktops and laptops. These chips would be custom fabricated (in lower volumes = more expensive) They even had a Cray just to work on these designs. I would bet that cost Apple as much to build these support chips as the actual PowerPC CPUs cost for each computer.


The reason Apple bought PA Semiconductor was all about the iPhone, iPod, and iNewton. Remember while PA Semi did mess around with PowerPC designs they have worked on the StrongARM processors in the past. In the iPhone it takes four different chips using ARM cores to run the box. Different chip vendors wrap stuff around an ARM core to create the chips that do the graphics, touch controller, broadband/wireless, sound, etc. That is 4 chips consuming power and 4 different vendors that Apple must rely on. Now they can craft one or two chips that run ARM cores that do what they want. They can get lower power and custom features other handheld device makers don't have. (Remember that the Newton was one of the original devices to use a low power ARM).

One of the benefits of Apple going from PowerPC to the X86 is that it forced all the developers to move away from the old Carbon/Quickdraw world of coding that had lots of hooks into the hardware. In moving to x86 they were able to force developers (slowly for some like Adobe) to get rid of that code and move to Cocoa and C or Objective C code, which is much easier to port to other platforms. This is how they were able to create OSX for the ARM in iPhone, and is also why Apple is being very tight on what can be developed on the iPhone. They don't want people hacking around the interfaces provide in the Cocoa/OSX world and talking directlty to hardware. Now that Apple has killed Carbon they can much more easily switch to any CPU. PowerPC, X86 or ARM. Or the Apple PC Semi ARM core super low power babies.

Matt Simpson | Apr 25, 2008 | 6:33PM

NeXTStep was running above : Motorola, Intel, Sparc, HP processors

Nothing new under the sun :)

Differrentiating from competitor is the 'raison de vivre' for SJ

gerard | Apr 25, 2008 | 6:36PM

I would say that seeing the x86 Instruction Set Architecture as a liability is a tremendous mistake. I'd thought that the recent transition to Intel had taught us a very important lesson here: the ISA is irrelevant, what matters is the vendor's engineering and manufacturing prowess. The G4s and G5s were utterly crushed by Intel's Core Duos and such despite their arguably superior ISA. Does anyone here expect Apple and the PA-Semi guys to beat Intel at their very game? Won't you remember how agonizing things got to be in the G4 era, only for histoy to repeat itself in the G5 one?

A few things:

-Saying that OS X, because of "it being Unix" (which is somewhat debatable, no matter the certs), was not designed for x86 so it won't run optimally on it feels rather strange (and just what was Unix designed for, in the days of the OS dinosaurs? Wasn't it all about platform independence, anyway?). You optimize for the target hardware platform via the best compilers you can get your hands on (and Intel's are really good), profiling and tuning, getting close to the metal where it matters, whatever it takes. You bet Apple x86-optimizes OS X and apps to death if it wants to stay competitive, just like any other Unix-on-x86 vendor. You bet any x86 ISA trick in the book is taken advantage of, SSEs included. And you bet PPC is the one no longer receiving so much tender loving care anymore.

-Choosing a manufacturer partner isn't such an easy choice: as prestigious as IBM is in that respect, it's been no fun for AMD or Sony sometimes.

As a long time Apple customer (since the Apple II and Mac 68xxx eras), If I was to suffer another idiotic transition I'd be EXTREMELY pissed.

Plus, if I was Intel, I would call Apple's bluff if that was the reason behind this adquisition.

(Actually, After Effects having a 2-3 GB ceiling has to do with it being a 32-bit application and Adobe's 64-bit development cost/benefit assessments). I don't think Windows is the culprit here: if 64-bitness was so fundamental to Adobe years ago, they could have offered a high-end AE64 version for some Unixes. Plus there was a fully featured 64-bit Windows available well before an equivalent OS X-64 appeared (and Apple cancelled the Carbon64 elements Adobe needed to do the port to 64-bitness anyway, adding at least two years or so to the port).

Snafu | Apr 25, 2008 | 6:43PM

Please don't start up again about how modern multi-core cpus are so powerful now that emulation will work fine and be "fast enough", "you can always throw more cores at it, etc., etc." The point is that Windows on native hardware will always be much faster, and that is what sets the expectation of users. It seems very unlikely that Apple would be able to sustain a hardware advantage of an order of magnitude or more necessary to make emulation anywhere near as speedy as native running.

Windows ran "great" under emulation on PowerPC --- as fast as on previous generation native hardware. But users saw it as slow relative to their current native Windows machines. And that was (and would be) unacceptable.

Jeff B | Apr 25, 2008 | 6:49PM

This still doesn't make sense to me. Intel spends billions on R&D to create the next generation of speed advances. AMD supposedly can't keep pace and has been falling behind. I think Intel's spend is mostly productive (as opposed to, say, MSFT's massive R&D spend). How is this small division in Apple supposed to keep pace? Why would Apple want Macs to leave Intel (and its cutting edge technologies) to create a competitor that will mean slower computers than PCs? Taking away that stigma by moving to Intel has been a big win for Apple. Bob glosses over way too much by just saying chips have been "commoditized". If so, why is AMD getting killed? Their chips haven't been able to keep up with Intel's gains in performance and performance/power.


Far more likely that this is for mobile integrated chipsets, as many others here suggest. Competing with Intel in chip development seems completely pointless for Apple.

JT | Apr 25, 2008 | 6:56PM

Please don't start up again about how modern multi-core cpus are so powerful now that emulation will work fine and be "fast enough", "you can always throw more cores at it, etc., etc." The point is that Windows on native hardware will always be much faster, and that is what sets the expectation of users. It seems very unlikely that Apple would be able to sustain a hardware advantage of an order of magnitude or more necessary to make emulation anywhere near as speedy as native running.

Windows ran "great" under emulation on PowerPC --- as fast as on previous generation native hardware. But users saw it as slow relative to their current native Windows machines. And that was (and would be) unacceptable.

Jeff B | Apr 25, 2008 | 7:02PM

cringely is often wrong and seems to over dramatize stuff with wild sweeping statements.

I don't see this as a cpu for PCs (including macs). The Mac has a big advantage that it can run Windows and OSX - that is going to be factor for a long time.

I can see this as a great processor for the next level of AppleTV or other post-PC devices such as a larger form factor, higher power iTouch tablet. Also, I bet there is some IP that Apple wanted for miniaturization and low power operation.

david | Apr 25, 2008 | 7:28PM

Things like instruction decoding just don't take all that much in chip resources, and dropping unused instructions is unlikely to save much. There are definitely crufty corners of the x86 architecture which are a waste of space, but they probably don't amount to much.

Roughly speaking modern CPUs get their performance from getting good cache hit rates: a combination of having large caches (of all kinds: data, instruction, tlb, registers, etc), having smart replacement policies, having good speculation, having good prediction, and if all else fails, being able to fill a miss quickly.

All that stuff applies regardless of instruction set architecture: x86, PPC or anything else, and is plain complex to get right.

Getting low power is a whole other issue, and again, isn't very sensitive to instruction set architecture.

If Apple has any chance of getting good performance out of a PA chip, it's going to be massive multicore rather than a direct replacement for x86.

To make that worthwhile, you're talking about a software architecture which can take advantage of that. x86 emulation definitely isn't that application, nor are general purpose desktop apps.

Server stuff may be better suited, but in general things like graphics, dsp, etc are a better match. That suggests either embedded (AppleTV) or co-processor (vector/media accelerator). But co-processors are very hard to make use of in general...

Jeremy | Apr 25, 2008 | 7:50PM

The first point is that virtualization (Parallels, VMWare, etc.) is NOT the same as CPU/instruction set emulation (Virtual PC, Rosetta, etc.). Emulation will always be slower than running native code on native instruction processors, it can only ever provide a temporary bridge to a different architecture, no matter how many cores you devote to it (native code will always be able to make better use of extra cores).

There's relatively little an OS needs to do to enable it to be virtualized, given the the hypervisor and appropriate CPU support, so "virtualizing Windows" makes no sense - that's Intel and VMWare's job not Windows itself. Which is probably why MS didn't do it.

The second point is Apple could never keep up with Intel/AMD performance with such a small investment, even if they switched to a stripped down x86 processor family that threw away all the accumulated crud. It's much more likely they want to build accelerator chips/video processors and other chipset add-ons for Intel/AMD processor-based machines that will keep Mac OS X on Apple gear ahead of (and locked out of) any clones and keep their OS X portable and handheld s/w (currently ARM-based but possibly switching to Atom and it's descendants in the future) ahead of anything competitors can come up with using pure Intel/ARM chipsets.

David S. | Apr 25, 2008 | 8:12PM

I don't get where this obsession with Windoze performance comes from. People don't buy Macs to run Windoze on, even when they do appreciate the ability to either dual-boot or run it in a VMWare instance for when they happen to need it.

I don't know if Bob's theory is right or not, but the performance of Windoze on a hypothetical future chip is not likely to be one of Apple's major considerations. Or at least, I hope it wouldn't be, since I'd like Apple to keep itself in the game.

Matt | Apr 25, 2008 | 8:19PM

Here's another suggestion.

Perhaps as a foundation partner is the ARM joint venture Jobs has discovered how profitable licensing chip IP can be. Not only is it an ARM chip used to power the iPhone, but virtually every other mobile phone in the world. Almost 3 Billion ARM chips were shipped in 2007 from the 200+ IP licensees.

Paul | Apr 25, 2008 | 8:42PM

Bob, I think you are on to something. Sony finally announced the availability in the US for video on demand for the PS/3, about time. I can see Apple doing a media/TV/iTunes/game console. It would be the God machine for the home. And it is about cores these days.

J Peters | Apr 25, 2008 | 8:46PM

Given multi-core is the way of the future, how about a multi core running generic x86 on one core - to provide a backup 'can run windows' capability - and something else on the other cores?

jaycee | Apr 25, 2008 | 9:32PM

Apple just wants a quick in with the military. :)

Partners in Grime | Apr 25, 2008 | 10:21PM

So, to make software available for a new platform, ship a new version of XCode, recompile for a new Universal Binary, and a new architecture is available for all developers.

Getting everyone to switch over to XCode wasn't just about making the switch to Intel... it was to break away from permanent dependency on any hardware platform ever again.

Scott Advani | Apr 25, 2008 | 10:27PM

This time you are proposing something believable. I was thinking the same thing when I heard this news, after checking PA's web site: they are all about performance per watt, but at the high end, and they apparently do not have anything to compete with ARM. (Just as Intel's Atom also doesn't really match ARM's efficiency.) Not that they couldn't, if they applied themselves in that direction...

But you didn't distinguish virtualization (e.g. VMWare and Parallels) from emulation. If they ditch x86, they will probably have to continue to emulate it for a while, to support older MacOSX apps that have already shipped. Just like they had to emulate the 68040 on the PowerPC, which was a really impressive engineering feat back then. Well, I bet by buying a chip company, they can make that job easier: the new processor will be able to emulate an X86 much more easily, because it will be designed to do that! And Windows apps will be able to run on it, too. So the pain of switching architectures again will be less than it was the previous two times. On the other hand, to retain power efficiency they will want to minimize the amount of silicon that is added for the purpose of emulation. So it will probably be a hardware-assisted software emulation rather than a complete, performance-competitive x86 implementation. But after recompiling the apps to use the PA instruction set, they will indeed be more power-efficient. This chip can be a big win for the next-generation slim laptops and the tablet we are still waiting for. As for the iPhone, if they just branch out into higher-bandwidth radio technologies, keep innovating in the graphics space, and maybe pump up the GPU capabilities a bit as necessary, it will be fine for several years.

I hope PA's existing customers are not left out in the cold though. I hope none of these engineers' talent gets put "on the shelf", because their work has such wide applicability.

Shawn Rutledge | Apr 25, 2008 | 10:34PM

Bob, I've been an avid reader of your columns for a long while now. Always on the margin, never participating. This time I thought I could drop in.

First of all, please pardon my english (I'm an spanish speaker), so I beg you to bear with me.

I've been using Macs for about 15 years, but not exclusively as I also have used Linux based machines, and where I work, more than 90% of the devices are Windows based.

Now to the point. I'm a System Administrator, and I hardly need Windows for anything except to manipulate Active Directory (For the time being, until I find a tool to do that without the need of a Windows based OS installed). I have a MacBook Pro at work, which I only use to connect to Outlook Web Access and to open some attachment in an email. Other than that, I always have an SSH connection open to a server, which I use to perform my tasks. And here is where I'd like to draw your attention: I don't use MY computer as much as I use OTHER computer to work.

I've seen this trend in other areas (Such as the regular desktop user). With services becoming more and more dependant on a server somewhere else (Think Gmail, Facebook, Myspace, Salesforce.com and others).

I know that the 'local' applications would never dissapear from the desktop (People still like to hear music and see photos while disconnected from the Internet), but the fact that the workplace -and the social space, for that matter- is moving away from the desktop an into the net is not avoidable.

Add to that the fact that consoles are gaining momentum, and each time more people are prefering to have a dedicated console (Be it a PSP, PS3, Wii, Xbox or anything else) than a gaming computer.

I may be dead wrong, but for me, the OS (Any OS) is becoming less and less relevant. I could perform my daily task as well in OS X as in Linux or in Windows... so there goes the platform advantage. Yes, virtualization is also a matter here, but even that topic will fade away in the future when applications become less dependandt of the architecture it is running on (Maybe something like Java, but well done).

Just a couple of cents. Best regards to all.

Eduardo J. Gisbert | Apr 25, 2008 | 10:36PM

The reason Apple uses Intel and not AMD is design. Steve Jobs is obsessed with aesthetics (that's not a value judgment). The reason Apple is stuck making pudgy iMacs is that the 65nm CPU's are hot. Intel now has 45nm CPU's. Where the 65nm parts used 65 watts, the 45nm CPU's at the same pace use 27 Watts, and have more cache to boot. And you know that smaller parts take less platter space and so the yields go up (eventually) and the costs come down. Now that the 45nm parts are starting to become available in supply again Apple is about to intro a new iMac. Coincidence? No. AMD isn't even in production with 45nm chips. An AMD iMac is last year's iMac. Apple can't go there.

But PA is making some nice cool chips too. They're not destined for the Mac line, but something like the AppleTV is dying for a custom chip solution. AppleTV needs three things: Massive processing power, very low operating temperature, and very large scale integration. If PA's chips can make an AppleTV (all costed in-house) for $99 that's a huge win. If all of AppleTV can be put on a chip, even better. If it can fit in an OpenCable (nee CableCard) slot and connect wirelessly to the iTunes store - checkmate.

Bill McGonigle | Apr 25, 2008 | 10:41PM

Here's another take on the deal from zdnet.

http://blogs.zdnet.com/Apple/?p=1627

srinisr | Apr 25, 2008 | 10:57PM

Apple could be licensing PA technology to intel for their own chips. They have done this in the past.

ryan | Apr 25, 2008 | 11:08PM

"Apple sells a 'brand experience' to upscale customers with lots of disposable income. Intel has the brand ID with the non-technical sort who make up the bulk of Apple's base and AMD is seen as the upstart cloner by that demographic. Again let me try to avert the flamewars by noting that the truth of the matter isn't under discussion, only perception by the NON INFORMED matters.
John Morris | Apr 25, 2008 | 4:43PM"

This is wonderfully uninformed. Every good "brand" has upscale customers. By "Apple's base" the uninformed poster means, to grant credit where credit has not been earned, iPod owners. Apple's "base," however, has always been techies. And the graphic arts and entertainment industries. Always. Bought my first one 20 years ago on a wage just-above minimum.

As for the recent purchase under discussion, it may be about Apple leveraging pricing with Intel but Intel could just stare them down. More likely it's as the column suggests: a hedge against virtualization, and a tool for the future.

eddie ever | Apr 25, 2008 | 11:24PM

"Apple sells a 'brand experience' to upscale customers with lots of disposable income. Intel has the brand ID with the non-technical sort who make up the bulk of Apple's base and AMD is seen as the upstart cloner by that demographic. Again let me try to avert the flamewars by noting that the truth of the matter isn't under discussion, only perception by the NON INFORMED matters.
John Morris | Apr 25, 2008 | 4:43PM"

This is wonderfully uninformed. Every good "brand" has upscale customers. By "Apple's base" the uninformed poster means, to grant credit where credit has not been earned, iPod owners. Apple's "base," however, has always been techies. And the graphic arts and entertainment industries. Always. Bought my first one 20 years ago on a wage just-above minimum.

As for the recent purchase under discussion, it may be about Apple leveraging pricing with Intel but Intel could just stare them down. More likely it's as the column suggests: a hedge against virtualization, and a tool for the future.

eddie ever | Apr 25, 2008 | 11:25PM

I cannot believe the premise that Apple bought PA to threaten to compete against Intel. I don't understand how this could possibly make sense to Apple; competing with Intel in a field where they're market leaders and Apple are inexperienced would be suicide.

A more significant threat to encourage Intel to reduce prices would be to switch to AMD chips. Or at least to invest in them to keep Intel honest.

No, I would guess, as some others have, that the PA acquisition is in order to make custom chips. There are a couple of scenarios (off the top of my head) that would benefit from such hardware.

Today, one of the worst user experiences is to create DVD's. Transcoding video is a slow process. It takes many hours and pegs your CPU (no matter the number of cores) close to 100%. I suspect throwing custom hardware at this problem would be possible and valuable to Apple.

Further, watching videos on portable devices (iTouch, iPod & iPhone) is becoming ever-more common. But you've either got to a) scale high-res video to fit on the smaller screens in real-time (which takes up a lot of space on the devices) or b) you need to transcode it down to suit. Method a) today cannot be done - the devices just aren't powerful enough and, even if it could be done, would consume significant power. Option b) can be done but it takes hours. Custom hardware would alleviate the issue.

Anyway, that kind of custom hardware is where I'd be interested if I were steering the ship at Apple. Competing or threatening Intel has no business benefit IMO.

MattyT | Apr 25, 2008 | 11:45PM

Bob,

I don't think the answer is what you think it is. I don't know have a clue what it is myself, but I think 3-4 years down the road we will find out you were wrong.

Guess we will see.
\

Wayne | Apr 25, 2008 | 11:52PM

The MacOS X has actually been running on 4 CPU platforms.
1. PowerPC
2. Intel x86
3. ARM (a stripped down version for the iPhone)

I'm most impressed by Apple's ability to move the MacOS from processor to processor with out too many hiccups.

The MacOS actually has an advantage by being on x86 right now due to the perception that they can easily run Windows. Gives people who are switching over a safety net.

Stanley Wong | Apr 25, 2008 | 11:56PM

How is it more efficient to "just throw an additional core at it" for windows vs some additional unused silicon specialized for windows and not os x. A little bit of unused silicon is a lot cheaper, cooler and efficient then a whole additional core only there for occasional windows use.

Also what about economies of scale here. Yes many if these pieces of silicon are commodities, but commodity prices are usually cheaper then custom parts run at 1/20th the volume.

tink | Apr 26, 2008 | 12:56AM

The real reson? ALTIVEC, the PPC routines that hugely speed up media tasks. Intel can't or won't produce an equivalent instruction set in their X86 chips (SSE4 is weak imitation), so Apple is buying what they want and need elsewhere. Future Macs with P.A's ALTIVEC implementation will have a huge performance advantage.

TJK | Apr 26, 2008 | 1:09AM

Bob,

I don't think the answer is what you think it is. I don't know have a clue what it is myself, but I think 3-4 years down the road we will find out you were wrong.

Guess we will see.
\

Wayne | Apr 26, 2008 | 1:10AM

The current Macintosh (Mac OS X) probably has a ten year life before Apple must replace it with a totally new operating system.

Operating systems need more intelligence. They need to think. So hopefully, the next generation of operating systems will include artificial intelligence.

That will require a totally new chip architecture. If Apple starts today, it will probably take ten years to get the whole thing working. PA Semi sounds the ideal company for this task and I think that Steve will enjoy this ultimate challenge.

Robert Elliott | Apr 26, 2008 | 1:22AM

Cringely, you know I love you (well, maybe you don't), but this time you've been smoking the crack-pipe. As much as the PPC->x86 transition was amazingly smooth in relative terms, in absolute terms it must have been expensive and definitely wasn't transparent on the users.

Running Windows "multi-core virtualized" on AppleChip in 3 years' time is total Alice in Wonderland because there a huge (humungous, gargantuan, etc) between sorta-speedy hassle-light /virtualization/ and non-cross-bootable-dead-slow-completely-ungameable /emulation/.

I don't know why Jobs bought the fab, and it could be used as a leverage on Intel for price cuts, but it ain't because he /wants/ to switch CPU manufacturer. Surely not to conserve a few nanometers of silicon real-estate on some esoteric x86 instruction, which frankly, OSX is more likely to use than Windows as-it-is.

Yaniv | Apr 26, 2008 | 1:32AM

Down the road, every computer will be hand holdable
little monsters with integrated gpu's, radios and
other necessary whatnots, sorry but fail to see your argument sir as PA Semi has no such intellectual properties.
Skaff.

Skaffen Amtiskaw | Apr 26, 2008 | 1:35AM

Here's a little bit different possibility: Apple knows that Microsoft Windows is near end of life, which means that x86 is too. Microsoft will go Linux, and become the standard Linux distro. That would make PA Semi very valuable. PA and Apple are friends, and Apple wouldn't want PA to fall into the wrong hands.

zato3 | Apr 26, 2008 | 2:40AM

"I'm sure partisans on both sides would disagree, but my statement is still true."

This is a bad way to make an argument. If you feel your audience needs more convincing, then offer more evidence.

I stand with others that say emulation won't cut it. I believe Mac's surge has primarily been due to native Windows compatibility, because Mac/OS X didn't really start to move until after it became Windows compatible. (Also, it fits my local experience). Windows like all software will grow to consume the available hardware when running natively. This is because programmers can and will trade CPU cycles and memory for ease of development, and lord knows us programmers need all the ease of development we can get. That will seem to always put emulation at a significant disadvantage. It would be mind-blowing to think that Apple doesn't understand that Windows compatibility has been a huge boost for them, and that they would then cut it off before they hit critical mass, which if the current rate kept going, would be in about 10 years.

(If I have a 4 thread x86/Windows app that runs only acceptably on an Intel quad core, how can an equivalent Apple quad core run it in emulation acceptably? How is Apple going to throw more cores at the problem than Intel to overcome the emulation disadvantage)?

Pentalon | Apr 26, 2008 | 3:10AM

"CPUs optimized for OS X would be smaller, cheaper, and use less power than any Intel or AMD alternative simply because they could be simpler overall."

Wow, that would usher in a whole new era of computing with a reduced instruction set. ;-)

leeg | Apr 26, 2008 | 5:02AM

I'm not as sure :) I think going by themselves would cut them out of all the free research Intel will ensure is built into their chips. They're buying a company for under 300 million. They'd need a lot more to lock in a sustainable advantage, if it's even possible.

Richard | Apr 26, 2008 | 6:15AM

Intel has the twenty-to-fortysomething-cores with SSE-on-Steroids coprocessor more or less covered with their Larrabee product, which ought to appear around 2009-2010, so I guess Apple won't try to design anything highend-ish there: it doesn't need to. I think PA-Semi will provide low-power "systems on a chip" for everything above an always-on very low power device such as the iPhone (unless they build things around the ARM processor until they create a comparable PPC core.

Snafu | Apr 26, 2008 | 7:41AM

Bob, you begin your article with the notion that no company buys another company for what they might produce in the future, only for what they have on hand. This directly contradicts what Steve Jobs claimed was his reason for switching to Intel -- he said it was their product roadmap, not the currently-shipping chips. Sure enough when Apple switched, it was several months later, with newer Intel chips.

Then you seem to say that the only thing that would motivate Steve Jobs is cheaper parts supplied. You go on to say that these PA Semi chips are no good for anything Apple now makes. This somehow causes you to claim that Apple now has incredible bargaining advantages over Intel. 'Hey Intel, now I have chips that I can't use, so you better sell me your chips that I can use for a lot less money!' Huh? That doesn't make sense.

You go on to say that PA Semi can have no idea as to why Steve Jobs bought their company. Well, it seems to me that they would have a lot better idea than you would. (Whether their press releases embody all they know, or even the right path, is another question.)

You go on to say that when the next-gen PA Semi PPC chips are ready (and isn't that buying a company for what they are going to have, not what they have now?) 'Windows virtualization' will be no problem -- 'just throw more cores at it.' The misunderstanding here is between 'virtualization' which is what Apple (VMWare and Parallels) can use with x86-workalike chips, and 'emulation' which is all that PPC-family chips could use to operate MS-Windows. We had emulation with PPC chips before, and I can tell you, it was a dog. Rosetta is what Apple uses for emulation now, to run PPC-only apps on Intel-based Mac hardware, and even with the patented, new tech algorithms Rosetta's creators came up with, non-Intel apps just don't run as fast.

The moment Apple switches off x86 CPUs is the moment VMWare Fusion and Parallels cease to work.

Bootcamp, of course, will NEVER work using PPC-family chips.

Steve Jobs might not have anticipated how much 'run MS-Windows on your Mac' would help sell Macs. He might not have wanted to believe it. But he's no fool and he must see how much it helps sell Macs now. There is also a whole world of 'enterprise customers' who are tempted to get Macs now. This is a market Apple never had.

MS-Windows games would not be feasible under emulation. As it is, gamers are reporting that to play PC-games on Macs today, Bootcamp is the only way to go.

Don't you think Steve Jobs feels a special shudder of pleasure whenever some trade organ like PC Mag releases a report saying, 'The Fastest Windows Vista computer...is a Mac' ? Sure he does. And it helps sell Macs.

There is one more argument about costs of parts that seems to have escaped you here. PA-Semi is fabless. That means they must design their chips (which costs money) and then contract with somebody like IBM to manufacture them. That means that Apple will have to know in advance how many chips to order. And even then the cost of those chips is going to be higher than Dell will pay for Intel semi-commodity chips. Apple will be losing price competitiveness by going in-house, not gaining it. The Apple-PA-Semi chips will cost Steve Jobs more money than the regular list price of Intel chips. Add to that the headaches of supply, knowing how many chips to order, being stuck with extra chips when new products come online, and so forth.

And finally: in 5 years or even 2, when you say all this will come about, can we even be sure AMD will still be in business? So I can maybe buy some of your arguments as a price hedge against a monopolistic Intel as the sole source of x86 chips.

But the rest of it? You really struck out on this one.

pond | Apr 26, 2008 | 8:32AM

I think that everyone is missing the point, and yet it is a point they all bring up, it is about more efficient computing. PA probably can help Apple with design ideas for chip sets. The efficiency of processors is tapering off but there is the whole chip set, and remember it was wanting for a more efficient 3G chip set that kept Apple waiting for a 3G iPhone. I think it was Andy Grove that told Apple, use our processors and we will work to optimize them for you. So I think that this is all about Apple helping Intel design better more efficient chip sets. And maybe helping Intel design processors that are more centered around a Unix OS instruction wise.

Dude | Apr 26, 2008 | 9:43AM

Clearly to me this is Apple's attempt to get Intel designs optimized for OS X. Better 64 bit optimization, lower energy consumption, specialized cores and instruction sets that will differentiate Apple Computers, iPhones, iPods, AppleTV and new products in the pipeline.

Knowing Apple's longer term planning, I would bet that PA has already designed and Intel has implemented specialized designs for future products developed in the Apple labs.

ChampagneBob | Apr 26, 2008 | 11:32AM

My observations for what they are worth:

This really isn't about prices for intel chips. The threat of Apple moving to AMD is more than enough to keep Intel very price competitive. This is more about control and Steve's inability to run Intel's R&D programs.

I recall Steve's quote (of Alan Kay) from that D-5 interview last year. "People that love software want to do their own hardware" This is just a natural extension of the "we build the whole widget" and lets Apple do more innovative stuff. Apple is cash flush and can shovel another couple billion into PA and not even miss it.

But don't expect the fruits of this merger to be about conventional PC's. This is likely going to be for a new class of gadget, or at least iPhone/appleTV.

OSX is already platform agnostic (that ain't intel silicon in the itouch or the iPhone), so development for a new chipset is no big deal. Apple has done it several times now.

Apple could care less about Hackintoshes. They can squelch these for most users just by making osx security updates check the hardware before installing. Right now the amount of shrinkage these are causing is not an issue and just makes the osx platform a little more popular. This is not a play to have locked-down hardware.

mac84 | Apr 26, 2008 | 1:08PM

I believe everyone maybe missing a point here, Apple recently dropped Computer from its Corporate name, reason: they are diversifying their corporate environment. In other words, they are reaching out to a broader spectrum of clients. They have iTunes, Apple TV, iPhones, iPods, Desktops, Laptops, Servers and homegrown software. Think about the missing elements to a new broad spectrum Corporate Company. Currently, they service Consumer America, Corporate America, and now with the acquisition of PA Semi they can reach out to Government America. How about a rugged MacBook Pro for Military, CIA deployment. How about OS X on Naval vessels, communication systems and the like. Apple may say they don't care about the product line, but think about misdirection here; they are notorious for this line of thinking and approach. Think about what is missing, you may have your answer. The problem with going back to specialized chips, companies that write software for you would have to rewrite their software again. I doubt that will happen. That would give Microsoft the upper hand again and push developers back to Redmond. Apple is not giving up the high ground now! Think about it.

macwarrior322 | Apr 26, 2008 | 2:43PM


Cringley, this is speculative nonsense. And technical nonsense. the virtualization of Windows on Windows happens on Intel boxes, similarly the virtualization of Windows on OS X happens on Intel boxes. The reverse does not happen so easily because of licencing. Running VM on other machines is better described as emulation which is a lot slower. In any case none of this is game changing.

It is totally nonsense to say that being Unix the OS X kernel cannot be optimized for Intel. It is like saying that beef cannot be produced from Cows because there are Orange trees - not only does it not follow it is wrong. OS X was originally, in it's NEXT days, written for Intel at the Kernel level and in fact has some advantages in that environment ( specifically related to method dispatching in Obj C calls) - but in general the OS is chip agnostic , which is why the iPhone works.

PPC is not coming back. Apple is not becoming a chip maker - except possibly for low end devices and application specific chips. Intel will blow them away if they start that, and Intel spends billions, and employs thousands in every one of its fabs. It is sucicidal to compete.

You've convinced me that this is about IP. I must look up what they have.

Eoin | Apr 26, 2008 | 4:21PM

Seems a little risky here. They don't own enough of the PC market to play like that. Intel could raise the price and pump cash into insuring Linux compatibility (wifi, especially), forcing more price pressure on the Mac (they're good, but what makes them so is the Linux base + peripheral support; Linux peripheral support is horrible for the non-hacker consumer).

dhimes | Apr 26, 2008 | 4:57PM

A couple minor corrections:

1. For macwarrior322 : changing architectures doesn't inherently mean rewriting code. It may just mean recompiling code, or even a Rosetta-like translation layer. Remember, Intel Macs are running plenty of PowerPC code through translation, and aside from slower launch times, it runs just fine.

2. For Eoin : NEXTSTEP was originally written for NeXT workstations, which used Motorola 68K CPUs. The ports to Intel, SPARC and HP PA-RISC came later, and foreshadowed Apple's current PPC/Intel "universal binaries."

Dan | Apr 26, 2008 | 5:05PM

Apple is a centripetal, introverted company, and will be for as long as Jobs is at the helm.

That means it will work to express its internal values and control quality over as much of its product as possible. And that means acquiring the ability to design if not build all components.

Apple buys where it must and builds where it can.

stefn | Apr 26, 2008 | 5:07PM

J. Peters. . . you are the 'Thinking Out Of The Box' winner. . . the PA Semi acqusition will result in a 'Game Console' product. Remember, Sony's 'Playstation3' runs on the PowerPC chip, Micro$soft's 'Xbox2' is powered by a PowerPC chip, and Nintendo's 'GameCube' runs on a PowerPC chip. The PA Semi purchase gives Apple access to lots of PowerPC licenses and patents, and best of all. . . the design team and the third party game software library.

Poohpa | Apr 26, 2008 | 5:15PM

If it is true that Intel is out in 2 or so years and Apple is using them for that period, then I see no reason for Intel to lower their prices. Why lose $278 million if your history?

gjdw1 | Apr 26, 2008 | 6:38PM

If it is true that Intel is out in 2 or so years and Apple is using them for that period, then I see no reason for Intel to lower their prices. Why lose $278 million if your history?

gjdw1 | Apr 26, 2008 | 6:39PM

If it is true that Intel is out in 2 or so years and Apple is using them for that period, then I see no reason for Intel to lower their prices. Why lose $278 million if your history?

gjdw1 | Apr 26, 2008 | 6:42PM

If it is true that Intel is out in 2 or so years and Apple is using them for that period, then I see no reason for Intel to lower their prices. Why lose $278 million if your history?

gjdw1 | Apr 26, 2008 | 6:44PM

If it is true that Intel is out in 2 or so years and Apple is using them for that period, then I see no reason for Intel to lower their prices. Why lose $278 million if you're history?

gjdw1 | Apr 26, 2008 | 6:45PM

If it is true that Intel is out in 2 or so years and Apple is using them for that period, then I see no reason for Intel to lower their prices. Why lose $278 million if you're history?

gjdw1 | Apr 26, 2008 | 6:47PM

PA Semiconductor website has been Cringelied.

hseldon | Apr 26, 2008 | 7:08PM

The ZDnet article provides an interesting perspective also.

Thomas Paine | Apr 26, 2008 | 9:15PM

I don't believe Apple would risk itself to became another SUN (propietary CPUs with integrated hardware-software). Intel has enormous resources in this area and its economies of scale cannot be matched.
Building custom chips (for video encoding?) or help Intel with tightly integrated solutions, or both, seems more likely.
One more thing: as someone have pointed up, to lock the hardware, it's possible even with stock x86 CPUs, custom wiring and BIOS integration. But... what about adding in-house developed (and obscure) chip into the mix?

a_fun | Apr 27, 2008 | 12:51AM

I think Apple's going to use this chip for an entirely new gaming platform. Radically different than what's on the market today. One that will easily use all the games that will be developed for the iPhone using Open GL, Al etc., with minimal code changes. Recent patent applications tend to support this strategy. Think about, why make another XBOX or PS3- that avenue is somewhat exhausted.

lrd | Apr 27, 2008 | 7:58AM

I think Apple's going to use this chip for an entirely new gaming platform. Radically different than what's on the market today. One that will easily use all the games that will be developed for the iPhone using Open GL, Al etc., with minimal code changes. Recent patent applications tend to support this strategy. Think about, why make another XBOX or PS3- that avenue is somewhat exhausted.

lrd | Apr 27, 2008 | 8:00AM

So will apple rule the worl like the 1984 ad they put out against IBM? if you are right cringely, then that is in the future!
Apple can rule the mobile phone universe with variants of Iphone (just like blackberry) in 5 years if it desires, with PC and Mac convergence, we could all be using macs soon if they are inclined to cut $500 of their $1000 cream profit, and it is not unthinkable for apple to buy a major gamer (seeing that the rivers of blood there will kill some of them)

so if steve doesnt die in the next 10 years you never know what can happen

kbentil | Apr 27, 2008 | 10:47AM

What are you talking about, Cringely? ISA stands for Instruction Set Architecture when discussing processor architecture. "Industry Standard Architecture" was an I/O bus used prior to the introduction of PCI.

In addition to this obvious factual error, you also assert that modern processors are all about clock rate and core count, which is an appallingly simplistic view of modern processor design. The Core2 processor is allowing Intel to beat AMD black and blue because of superior architecture, not superior clock rates or core counts. Intel and AMD are competing on instructions per clock and power usage, not clock rates and how many cores they can add.

Why do you have a column analyzing an industry you clearly do not understand?

Chase Johnson | Apr 27, 2008 | 3:24PM

Bob's premise is wrong; it doesn't make a difference what PA is doing right now; chip engineers can make anything, their training is general and skills are transferable. I've programmed 20 kinds of DSP and embedded systems and can program anything, it doesn't make a difference whether I am processing sonar or moving a robotic arm or handling a communications protocol.

I'm in with DavidLBC and the several people that believe it is for various accelerators, video, etc. Maybe even security, or parallelization apps. Hardware assist on what are quickly becoming de facto standard apps can be a huge competitive advantage, and generalized Microsoft can't really compete. I would be astonished if Apple built a proprietary processor.

Processors are commodities and Bob's reasons for Apple to start producing a commodity are totally weak. Don't forget that includes constantly updating them to fix bugs and compete (because his machines will still be competing against Windows and Unix running on Intel and AMD). Saving a few watts doesn't cut it; pour the money into proprietary battery technology, there is a ton of room at the top there.

Besides, some low voltage hardware assists can relieve the load on the main processor and let it run slower, that will save the watts you need.

Tony C. | Apr 27, 2008 | 7:08PM

Bob's premise is wrong; it doesn't make a difference what PA is doing right now; chip engineers can make anything, their training is general and skills are transferable. I've programmed 20 kinds of DSP and embedded systems and can program anything, it doesn't make a difference whether I am processing sonar or moving a robotic arm or handling a communications protocol.

I'm in with DavidLBC and the several people that believe it is for various accelerators, video, etc. Maybe even security, or parallelization apps. Hardware assist on what are quickly becoming de facto standard apps can be a huge competitive advantage, and generalized Microsoft can't really compete. I would be astonished if Apple built a proprietary processor.

Processors are commodities and Bob's reasons for Apple to start producing a commodity are totally weak. Don't forget that includes constantly updating them to fix bugs and compete (because his machines will still be competing against Windows and Unix running on Intel and AMD). Saving a few watts doesn't cut it; pour the money into proprietary battery technology, there is a ton of room at the top there.

Besides, some low voltage hardware assists can relieve the load on the main processor and let it run slower, that will save the watts you need.

Tony C. | Apr 27, 2008 | 7:08PM

I think it's an interesting and thoughtful take on the PA Semi move. The part that interests me the most is the idea that Steve could release a hardware product with a new processor that is more optimized to the software than what they have today with Intel.

Intel seems like just the right move for this moment in the market. It gives them a lower cost proven platform while helping them win over Windows users because many of the applications can have the same core code base.

But I remember NeXT and the many of the ideas that were before their time. NeXT was a huge failure because nothing ran on it and while it had productivity features far advanced from alternatives it didn't get the adoption that it needed due to a raft of issues.

But that was then and in five years the conditions for a software optimized HW platform, at least on the client side, could be interesting.

Beyond all this speculation it just makes sense to have more processor IP inside your company if you are a systems maker. Apple will want to be smart where they can whether the products are servers, laptops, iPhones or whatever else they dream up. As long as Steve doesn't end up getting too closed up again and driving innovation away from his platform, he should be able to do do some proprietary magic here and there.

Kris Tuttle | Apr 28, 2008 | 2:29AM

I think Steve just made a $278 million personal Chip design fantasy fulling move...

Kiran | Apr 28, 2008 | 2:58AM

I think Steve just made a $278 million personal Chip design fantasy fulfilling move...

Kiran | Apr 28, 2008 | 2:59AM

I think Steve just made a $278 million personal Chip design fantasy fulfilling move...

Kiran | Apr 28, 2008 | 3:03AM

Intel already makes customizations for Apple (see Macbook Air). Apple needs their own counterparts for Intel engineers, who can do part of the customization in house, and work closely with the Intel engineers.

This will improve communications between the two companies, increase development speed, drive down costs, and may even net Apple some (new) intellectual property in the process.

Allerbe | Apr 28, 2008 | 4:19AM

Intel already makes customizations for Apple (see Macbook Air). Apple needs their own counterparts for Intel engineers, who can do part of the customization in house, and work closely with the Intel engineers.

This will improve communications between the two companies, increase development speed, drive down costs, and may even net Apple some (new) intellectual property in the process.

For Apple to some day go it alone, the way Cringely seems to indicate, is a Very Bad Idea. Intel brings (and will continue to bring), far too much to the table. Remember: during the PPC to Intel transition, Intel had 1000 engineers working exclusively for Apple!

Allerbe | Apr 28, 2008 | 4:27AM

Intel already makes customizations for Apple (see Macbook Air). Apple needs their own counterparts for Intel engineers, who can do part of the customization in house, and work closely with the Intel engineers.

This will improve communications between the two companies, increase development speed, drive down costs, and may even net Apple some (new) intellectual property in the process.

For Apple to some day go it alone, the way Cringely seems to indicate, is a Very Bad Idea. Intel brings (and will continue to bring), far too much to the table. Remember: during the PPC to Intel transition, Intel had 1000 engineers working exclusively on the Apple account!

Allerbe | Apr 28, 2008 | 4:28AM

Intel already makes customizations for Apple (see Macbook Air). Apple needs their own counterparts for Intel engineers, who can do part of the customization in house, and work closely with the Intel engineers.

This will improve communications between the two companies, increase development speed, drive down costs, and may even net Apple some (new) intellectual property in the process.

For Apple to some day go it alone, the way Cringely seems to indicate, is a Very Bad Idea. Intel brings (and will continue to bring), far too much to the table. Remember: during the PPC to Intel transition, Intel had 1000 engineers working exclusively on the Apple account!

Allerbe | Apr 28, 2008 | 4:29AM

Sorry for the double posts! I got an error message, asking me to go back in the browser and post again! And again... And again...

Allerbe | Apr 28, 2008 | 4:44AM

Here is my reply. Apologies for it being off-site.

http://gfair.livejournal.com/11900.html

Also, Bob, the survey is nuts. Don't tell me 82% of your readers forget what prices, feature lists, and speeds were like on PowerPC Macs. That doesn't even factor in market share, profit, and stock price gains. What Intel has done for Apple is so tremendous that it can be summed up in three words: bigger than Dell.

gfair | Apr 28, 2008 | 6:41AM

Another thing: As a high performance computing researcher, let me point out the typical OS on an x86 (including Windows and Linux) is running at about 5% of peak efficiency, and with some fairly easy profiling and tuning can be written to run about 10 times faster, at about 50% of peak efficiency for the x86. If Apple really wants to kick ass on the wattage and performance, they can do what they have always done: Write better programs. They can just think their way to better battery life and less wattage. Another thing PA could really do is provide some wildly ambitious and intelligent hardware control over the power usage; that would be a great piece of Apple IP.

Tony C. | Apr 28, 2008 | 7:00AM

As a long time Mac software develope (mainly realtime media processing code) I'd certainly enjoy a good old PPC architecture.

Many Mac users and developers DO appreciate the elegance of their machine's making to the degree of affecting their purchase decisions.

You could compare this with the mechanical spec of a Car's engine, which people appreciate even when it has no impact on their use of the car.

Motti Shneor | Apr 28, 2008 | 7:41AM

I don't care about all the technical arguments here. As usual the serious tech-heads think everything comes down to which processor is better for what, when in fact nothing depends on that at all. As long as Apple continues to make good stuff, it doesn't matter if it runs on vacuum tubes. They're not going to turn out any system that's sluggish, that's all we need to know as far as running OS X is concerned.

*However*, I am wondering what would happen to gaming on the Mac (I know, "What gaming for the Mac?" har har). As someone mentioned earlier, the only way to have access to all the games that are Windows-only is to use Boot Camp (it's basically the only real use for Windows, as long as developers snub OS X). I have already found that virtualization just doesn't cut it when it comes to gaming, but of course Boot Camp works flawlessly because my Mac actually becomes a Windows machine.

The downside of the current situation is that it hurts game development for the Mac even more (why bother when the Mac users are emulating/virtualizing/installing Windows?). So what happens if we move to a platform that doesn't support Windows, or does, but with a big performance hit, like back in the PPC era? Does that help us, or hurt us? If the Mac's market share continues to climb (and it's got a long way to go before this happens), game developers may start to care about the platform again. Otherwise, with things as they are now, we'll go on having only a handful of games for OS X and suddenly be unable to acceptably run all the Windows-only games.

Oh well, gaming is for consoles, right? Let's face it, most Mac users don't care about using their Macs for gaming, they just buy PlayStations.

Hampster(TM) | Apr 28, 2008 | 10:04AM

"I am wondering what would happen to gaming on the Mac"

Gaming will come to the Mac when the Mac is a "big enough" platform. You're not worried about "when gaming will come to the iPhone" are you? No, because you know that is inevitable. Well, at some point the Mac market share (combined I suspect with user demographic) becomes compelling enough for games. As with all things - a few will be early pioneers, the rest will follow.

Mister Snitch | Apr 28, 2008 | 11:51AM

BTW, this is NOT meant as an insult to Mr. Cringely, but the comments are the best part of his columns. Indeed, I'd say it is a testimony to his ability to attract a certain type of intelligent reder.

Mister Snitch | Apr 28, 2008 | 12:28PM


You have a situation now where Samsung is
providing the key ARM-based embedded devices
in the iPhone, while also offering its own
so-called iPhone killer, the Instinct
(available from Sprint sometime in June).

So this is about ARM/Samsung, not Intel/AMD
(also low-power Gigabit TCP/IP Offload and
other protocol accelerator engines) and doing
to ARM Cortex what DEC achieved with StrongARM.


N.C. Brown | Apr 28, 2008 | 1:54PM

Kevin Watters writes:
"If the processor design was so "bad" comparatively, then why can Intel keep it competitive?"

Simple. And this is something that Bob glossed over.

Intel can apply economies of scale that other chip makers only dream of.

That's why Intel won the Intel/PowerPC war. The PowerPC is a vastly superior architecture over the x86. In the early 90's, everyone would have predicted that it would get and stay way ahead of x86 -- because of the price / performance issue. The x86 had to commit far more silicon to get the same level of performance. So, it should cost more.

But no, Intel sold enough product that the economy of scale worked in its favor. In the end, IBM couldn't keep up.

Bill Coleman | Apr 28, 2008 | 5:00PM

I read about 30 of these comments and don't see any one asking what Intel or their customers are going to do with 16 or 32 cores? 12 core server chips are only 4 months away. What are you going to do with 16 cores on a cell phone or laptop? Even if the software becomes compatible you are still looking at a huge waste on processes that must be solved sequentially. Perhaps Apple and Intel are already planning to make chips much more heterogenous. Why not include custom silicon for your operating system, games, graphics, 3d or sound processing. Any process that can benefit from a custom silicon solution can now have its own processor. Your brain is organized with a few billion processors made in a few hundred different formats. Why not copy a system we know works?

The kicker for this speculation is that it would explain how apple could have its cake and eat it too. One chip could have 4 x86 cores for backward compatability, 8 power pc derived cores of various types for specialized functions and 4 cores that run system functions for os x. Apple could liscense designs for cores to Intel and keep some for itself and still use Intel as their fab with it's huge economies of scale benefits. Have you noticed how much the newest multimedia extensions on Core 2's sound like Altivec knockoffs? Maybe Apple is already liscensing chip tech to Intel in exchange for design support.

Patrick McMaster | Apr 28, 2008 | 10:40PM

Patrick McMaster: I know researchers right now that are working on parallelizing problems to many cores, and there is already a large body of work on distributed processing.

The typical Windows/Linux graphical desktop has over a hundred active tasks and is handling at least 1000 interrupts (the clock) and up to 15,000 interrupts per second. The typical simulation (both scientific and game) is massively parallelizable, and 32 cores would be welcome, 256 cores would be more welcome. If you think game graphics are cool now, wait until the processors can handle Pixar-movie-quality ray-trace rendering ON THE FLY. That is another massively parallel operation. The guy in the office next to me has spent the last year of his life figuring out how to automatically and efficiently distribute code on multiple cores, with great success. On standard benchmarks, he can recompile and get a 3X speedup using 4 cores (which is damn efficient since there is coordination overhead). Even what look like simple Windows programs can be parallelized; route browser animations to one guy and mouse handling to another. The amount of work going into this area is huge.

Tony C. | Apr 29, 2008 | 8:20AM

I'll tell you what i can do with 16 or 32 cores.... but i can't right now, because my machine is grinding on converting my footage from my Sony EX-1 and some other HDV cameras into H.264 for a Blu-Ray Disc. it is monumentally PAINFULLY slow. BTW: most HD cameras are now h.264 or HDV, and so my mom will have one soon, and she'll want to make movies to email to people... and she has a G4 Mac Mini.. so she is screwed.



If i had 32 cores, i wouldn't need to have a second computer for "day to day" things... i could simply run my apps on other cores.

theotherstevejobs | Apr 30, 2008 | 6:07PM

I'm pretty sure everything you said about the x86 architecture was wrong, but I'm too tired to correct you at the moment...

astrange | Apr 30, 2008 | 8:31PM

Seeing all the trend, I believe Apple does not like the others to take over their part, and that is happening at the moment making the Mac with an x86 architecture. It happened before and still happens. If I'm looking to the green area of what people push with their OS X, making it to run over a standard x86 CPU I understand where the Apple is aiming to. And is their own rights.
With a bit of effort, you do not need a Mac to run the OS X and their apps. Maybe some incompatibilities with the hardware drivers might exist, but people make it running. It wouldn't be the same if Apple will get and develop a NoN-x86 CPU specifically designed for their own needs (Mac-Needs). Everything will be designed in such a way to work just smoothly throwing away all the NoN-necessary bits and keeping just what is needed (this lead me to think to the RISK architecture). Look to the tests people do today comparing the Macs with the PCs both running on OS X. I see why S.Jobs is getting that company. I hope he'll succeed to build and make more innovations pushing the technology beyond the limits we (the end consumers) trend to have just cos we are comfortable. In this way the end consumer will have only to benefit/win (hopefully) and have at an affordable price (hopefully) again) all the tools the corporate world uses.

JuBal | May 01, 2008 | 8:08AM

PA Semi acquisition is also about revamping Apple's IC design team; IC design has never been a strong suit in Apple's technology portfolio. There's a reason that we never saw a G5 portable back in the PowerPC days; Apple couldn't produce devices which were efficient enough to allow reasonable battery runtime.

Apple has always preferred to "buy" versus "make", and so the x86 migration made sense from a vendor-sourcing perspective. The PA Semi team would allow Apple to fake a "make" punch at vendors, but in the end Apple corporate culture doesn't support internal development, and PA Semi's engineers are so top-shelf that the chances of them sticking around in a "buy"-culture company are low. The majority of PA Semi's engineers will stick around about as long as it takes for the ink to dry on their retention bonus checks, and not a minute longer. In the end this saga will end up being referred to as SiByte/Broadcom, Part II -- The Adventure Continues.

TheInsider | May 01, 2008 | 2:17PM

Um, apple + OSX is nothing but an overpriced overblown and irrelevant bit of fluff. In the big scheme apple is utterly inconsequential. IBM dropped them as a big waste of time. Intel only bothers wasting its time precisely because of apple's terrific margins selling to fools looking for ways to part with their money. But take note: Apple gets only 'off-the-shelf' product choice. That's why it is so easy to clone an intel mac.

Grunchy | May 02, 2008 | 1:31AM

Um, apple + OSX is nothing but an overpriced overblown and irrelevant bit of fluff. In the big scheme apple is utterly inconsequential. IBM dropped them as a big waste of time. Intel only bothers wasting its time precisely because of apple's terrific margins selling to fools looking for ways to part with their money. But take note: Apple gets only 'off-the-shelf' product choice. That's why it is so easy to clone an intel mac.

Grunchy | May 02, 2008 | 1:32AM

Grunchy has a good basic point. The Intel transition was said to be about 'options', but in reality, Apple is buying everything Intel sells. They are locked in; Intel is the mafia of chip companies. You will never see a non-Intel Mac from now on. Not even an AMD, just watch.

Rick | May 02, 2008 | 9:29AM

Nope, I'm not buying your line.

Some time in the not too distant future, when Apple is done taking over the consumer market, the enterprise market will become Apple's next growth engine. As instrumental as the switch to Intel has been in convincing consumers to take the Mac plunge, it's even more important in the enterprise market.

Furthermore, you just mentioned that AMD is right there ready and willing to supply chips to Apple. Why does Jobs need to fire a 278 million dollar shot across Intel's bow when AMD can furnish the cannonball gratis?

al | May 02, 2008 | 12:47PM

What a lot of people don't realize is the fact that Jobs has always promoted and favored "garage shop" operations segregated from the traditional corporate environment, allowed to think outside the box, free to invent the "NEXT" successful gadget, and not tied down by corporate policies stifling innovation. This was exemplified by the development of the original Apple I, Macintosh, Pixar, Next, etc. Current Apple corporate campus is set up in a more traditional way, focused on supporting and improving existing products as opposed to bringing up something new and revolutionary. This acquisition puts a ready made organization in Steve's hands fully under his control and away from the main thrust of Apple Inc. It will also be easier to keep a smaller organization under wraps until they are ready to reveal their new product which can be a challenge at the main campus given its current size, as I'm sure it was when the iPhone was developed.

Joe | May 04, 2008 | 9:52AM

"making a lot of Intel hardware simply unnecessary" please state one piece of the x86 architecture that is unnecessary for OS X.

Incog | May 08, 2008 | 1:08PM