Apple would have transitioned to the Cell processor had it stayed with PowerPC, but too expensive

Seth Weintraub
|
June 15, 2009
|
Share

CNET, on the fourth anniversary of the switch from PowerPC to Intel, interviewed an ex-IBMer who had been familiar with the IBM PowerPC-Apple relations at the time.  He/She(/Papermaster?) had some interesting insights into the situation at the time. 

The generally accepted reason for the big switch was that Intel's Power/Watt ratio with the Core Duo crushed anything that IBM/Motorola could come up with (plus it ran Windows).  This person offers some different scenarios:

Apple wanted better pricing, according to this person. Apple was paying a premium for IBM silicon, he said, creating a Catch-22. IBM had to charge more because it didn't have the economies of scale of Intel, but Apple didn't want to pay more, even though it supposedly derived more from an inherently superior RISC design as manifested in the PowerPC architecture.

 For IBM, the business with Apple was a financial sinkhole because the company had to invest a lot of money in chipsets, compilers, and other supporting technologies but could only take about 5 percent of the overall PC processor market, he said. So, in the end, it was impossible to make money. 

Why 5 percent? Apple insisted on double sourcing (IBM and Motorola). So, from the start, this left IBM with about half the market it could have had. This, he said, was an enormous financial burden. Paraphrasing the ex-IBMer: Intel was a single company with the lion's share of the market. While two companies--IBM and Motorola--had to divvy up a much smaller share of the market, while still investing, individually, tremendous amounts of money. And Apple played one against the other, according to this person.

Perhaps most interestingly, IBM planned to migrate Apple to the Cell processor, where the economics of scale could then be utilized.  The Cell platform is shared with Sony Playstation and others.  With the release of Snow Leopard, which is not PowerPC compatible, that hope is all but dead.

IBM had hoped to amortize the cost of PowerPC on Cell, the PowerPC-based chip design now used in the Sony PlayStation, some IBM severs, and IBM Roadrunner supercomputers. Big Blue was hoping to move Apple to Cell and then get the economies of scale there, according to this person.

Comments (23)

9to5Mac Noob (not verified) | June 15, 2009 - 8:00am

Is this headline in English?

admin | June 15, 2009 - 8:03am

Sorry, sometimes the caches don't clear after initial edits.

9to5Mac Noob (not verified) | June 15, 2009 - 8:42am

Wait wait wait ... Apple was so stubborn over paying a few extra bucks for each proc that they decided it'd be better to put everyone (them / users / devs) through the wringer and switch over to Intel?

Not sure I'm buying this one.

9to5Mac Noob (not verified) | June 15, 2009 - 8:53am

Good luck putting that in a laptop...... G5 power book anyone? Lol

9to5Mac Noob (not verified) | June 15, 2009 - 9:09am

After IBM stiffed SJ with the 3Ghz G5 the writing must have been on the wall especially when a lot of engineers were developing stuff for MS, Sony and Nintendo. Where was the love for Apple?

Well Intel had no console work to worry about.

MadDog (not verified) | June 15, 2009 - 10:36am

Snow Leopard not PowerPC compatible? uh-huh... just like all versions of Mac OS X prior to 10.4 Tiger weren't Intel compatible?

I'd bet a fair chunk of change that in the same way Mac OS X lead a secret life as an Intel build in a lab for years, it will continue to be compiled for PowerPC for some time to come. Apple just won't release it, that's all. It's a hedge, and we all know how Steve loves to grow his hedges. In his own words, OS X is "cross platform by design" and they'll keep it maintained for PowerPC just like they have to keep it maintained for the ARM processor in the iPhone and any other platforms that future Apple products may use.

9to5Mac Noob (not verified) | June 15, 2009 - 12:11pm

This is certainly not the case.
The release of 10.6 is much more deep than you think.
Snow Leopard contains technologies that are useless on PowerPC.
It's so deep that if Apple didn't drop support for the PowerPC processors on Snow Leopard, they would have been unable to move forward in the direction they are heading.
Grand Central Station, is Apple's technology to allow developers to add multicore support into their software, thus taking advantage of the Core architecture and it's optimizations AND their multiple processor cores.
The only PowerPC chip that Apple released were the last iterations of the PowerMac G5; the 2.5 Quad, and the Dual Core 2.7 G5.
How many do you suppose are in use out there?
I bet not a very significant number.
Next, 64 Bit support from kernel to core Mac OS X applications like Mail, iCal and Address Book.
G4s are unable to run 64-Bit code.
So, again... what PowerPC processors fill the requirements to take advatage of such technologies?
Multicore G5's. And that's it.
Like I said, that is not enough to justify a whole new OS release.
Besides, it's also a political move.
Is no secret that Steve Jobs has been acting truly disdainful toward the PowerPC platform ever since the announcement of the Intel transition.
Remember that they shipped the first Intel Mac 4 months ahead of schedule?
And remember that they transitioned the entire Macintosh line in less than a year?
Jobs was in a hurry to get rid of PowerPC. And that wasn't entirely necessary.
He could have let the transition go as planned; the way he had promised on 2005.
In fact, on that keynote he said Apple would continue to support the PowerPC platform for "a long time" because of the large installed user base... he even said that they had "great PowerPC products in the pipeline yet to be introduced"... needless to say, those products never saw the light of day.
And the fact remains that it's only been 3 years!
No, my friend. Apple is seriously done with PowerPC.
Snow Leopard is an "Out with the old, in with the new" release if there ever was one.
The inclusion of the aforementioned technologies, and a bunch of new stuff like OpenCL, and other factors like the rewrite of the Finder from scratch and the optimization of the code in core OS assets like the new Quicktime X; makes it very clear that this release is the foundation of what's to come.
This is the code version that future versions of OS X will be based on.
Apple had to do this eventually. Clean up the OS code. Get rid of the PowerPC instruction set. And in the process reducing the OS footprint.
The real question is... after only 3 years, was it too soon to kill PowerPC?
That, remains to be seen.

9to5Mac Noob (not verified) | June 15, 2009 - 12:48pm

saved me saying this. I'm betting Snow Leopard's purring on cell processors (and iPhone OS on who knows what, PA Semi notwithstanding) even as I write.

msiller | June 16, 2009 - 5:01pm

Blhe... whatever then...
Some people believe in aliens and UFOs, some believe in Santa Claus...
If you want to ignore the facts and go with what you "feel", then by all means... be my guest.
I bet you're one of those people, though, that think we didn't really land on the moon, and Hitler's still alive and such... right?

ulfoaf | June 15, 2009 - 10:59am

Intel was not the best,  just the most POPULAR. That made it cheaper ...

You get the feeling the G series processors were just not that much of a moneymaker for IBM. That is a pity. I have read that power consumption and heating were the biggest issues, and Intel had already solved it.

I hate to see the processor world become uniform. It doesn't encourage innovation if Intel is a monopoly. If it wasn't for AMD's AMD64 and IBM's G5, we would probably all have 32 bit processor machines still.

NotSo Noob (not verified) | June 15, 2009 - 11:10am

MadDog nails it. SJ never leaves himself w/o an alternative or three. I wouldn`t be surprised to see, in a few years, laptops with both Cell & Intel chips in them. Portable "supercomputers".

MadDog (not verified) | June 15, 2009 - 12:13pm

Personally I'm not sure we'll ever see a Cell/Power derivative in a desktop or laptop again (unless something *really* drastic happens in Intel's world) but I wouldn't be at all surprised if some other product/platform showed up with one in somewhere down the line.

Dave (not verified) | June 15, 2009 - 11:12am

It is pretty clear to me that the PoerPC offerings of the time sucked performance wise. Honestly this isn't even debatable an is backed up by some of Apples own benchmarking.

The fact that they sucked is one thing, the fact that they burned up energy faster than anything at the time is another black eye. The only thing PowerPC had going for it was Alt-Vec and that was of limited use.

All that aside I hope everyone remembers how pleased the new adopters where with the performance of the new intel machines when they first came out. A lot of that due to extremely good integer support on intel processors which to an OS and many user apps is far more important than Alt-Vec type processing.

Now looking at Cell Apple could have gone that route if they had OpenCL back then to exploit the vector units. They didn't and it would have been of limited usefullness anyways. The problem with the Cell was that the PPC unit was a huge step backwards in performance that would have been obvious until Apple could have fully integrated the vector units. That and the question of how do you evolve Cell. It was no time for Apple to take a step backwards just when intel was just about to take a giant leap forward.

Sadly all I see in this report is BS. It just sounds like somebody is trying to polish IBMs tarnished image.

Dave

9to5Mac Noob (not verified) | June 15, 2009 - 11:37am

To DAVE:
PowerPC didn't suck at all. I have a octocore (2xquad-core 2.8 GHz) Mac Pro, and it is only about 50% faster than the Quad-core PowerMac G5 it replaces. And that Quad G5 is 3 yrs older!!

Sure, the G5 roadmap was stagnating, the chips were so power-hungry that they required liquid cooling, and they couldn't be fitted into a laptop. These are very real disadvantages. But performance being an issue? Please. These G5 PowerPC procs performed very well, thank you. And the P4's they were competing against were just no match.

Peter (not verified) | June 15, 2009 - 12:49pm

I'd also like to point out that the last machine Apple "updated" was the PowerMac G5, and the initial Macintosh Pros weren't all that much faster than the PowerMacs they replaced (but they did give you lots of space for internal drives without having to have all that space for cooling).

The funny thing about this, though, is that at the low-end (Mac mini) and the high end (Macintosh Pro), the new machines cost more than their predecessors.

Patz (not verified) | June 15, 2009 - 11:38am

To DAVE:
PowerPC didn't suck at all. I have a octocore (2xquad-core 2.8 GHz) Mac Pro, and it is only about 50% faster than the Quad-core PowerMac G5 it replaces. And that Quad G5 is 3 yrs older!!

Sure, the G5 roadmap was stagnating, the chips were so power-hungry that they required liquid cooling, and they couldn't be fitted into a laptop. These are very real disadvantages. But performance being an issue? Please. These G5 PowerPC procs performed very well, thank you. And the P4's they were competing against were just no match.

George (not verified) | June 15, 2009 - 1:56pm

Apple 'failed' when they launched the original PowerMacs.
The initial release should have been 6100/80, 7100/66 x2 processors, and 8100/60 x3 processors. Yes, there would have been little advantage, besides running Photoshop, but it would be nice to build an installed base of multiprocessing machines. [for the anticipated release of Gershwin]
Apple should have taken advantage of processor efficiency, and let Intel play a losing game of maximum clock speed.

The initial PowerMac G5 could have been 1x2.0, 2x1.8, 3x1.6 [hey the case was large enough]
and then when the 970fx was released, the speed could have been bumped by 500MHz
and when the 970mp came out, no increase of clock speed would have been necessary or desired, due to more cache and second processor core.

The last iMac G5 should have been 2 GHz 970fx on 1 GHz bus (for the 17"), and a 1.5GHz 970mp on 1.5 GHz bus (for the 20")

and saying that it was "impossible" to have a G5 PowerBook was such a crock-a 1.5GHz 970fx would have been more than capable.

9to5Mac Noob (not verified) | June 15, 2009 - 2:36pm

Apple considering Cell? That's preposterous. Cell is a watered down PowerPC core with no out-of-order execution. Cell is to the G5 what a dual-core desktop Atom chip and eight GPUs would be to the Core 2 Quadro.... Huge, huge step down in performance.

Would Cell have been a step up from a G4? Maybe, but nowhere near the performance of the Intel CPUs. Beating G4 performance is a pretty low bar....

Cell is *highly* tailored to running games, and it does that well. However, because it does not do any out-of-order execution, Cell would almost certainly be an absolute dog in terms of desktop application performance. Moving a desktop app to Intel is actually easier than moving a desktop app to Cell if performance is a consideration at all.... I think the odds of Apple having considered Cell seriously are slim to none unless IBM had plans to do some sort of Super-Cell chip or something.

9to5Mac Noob | June 16, 2009 - 4:55am

Methinks you don't know much about processor architecture. The cell isn't inherently geared toward games. Even in the PS3, it's not geared toward games.

9to5Mac Noob (not verified) | June 15, 2009 - 2:43pm

Doesn't matter. It's the same thing Anyway. Apple wanted cheaper chips. IBM couldn't deliver because not enough scale (and 10% wouldn't have helped it anyway). Intel WAS getting economies of scale, which they dumped back into R&D, which was helping the Intel platform catch up and negate any advantage of RISC.

The only true moral is not why they switched, but that they did switch. I'm no apologist, but it's pretty clear the switch to intel is one of the two biggest reasons Apple has exploded in the market (the other being the mass market seeding with the iPod and now the iPhone).

nonyabusiness (not verified) | June 15, 2009 - 9:16pm

What nobody has asked is what about the future of cell? Since Sony is the only one using it and Sony is TANKING, I think the future for cell is bleak indeed. Smart move for Jobs to not have fallen for the Cell processor.

It's more of the usual at IBM - Great technology that is either too complex, too expensive, or too difficult. And I should know - I worked for IBM for 10 years.

As for processors, I like AMD myself...

9to5Mac Noob (not verified) | June 16, 2009 - 1:28am

I wouldn't say they are tanking, they might be having a very bad few years but tanking is a stretch. The PS3 will continue to gain market share as the 360 stops being able to compete, their royalties for BluRay technology will continue to bring them money, and OLED will help them turn around their TV market. Sony might have misstep there but they'll comeback

9to5Mac Noob (not verified) | June 16, 2009 - 12:47pm

Sony is actually loosing marketshare, and continues to loose money on every PS3 it sells. Further, it's online presence has been pretty much universally maligned, and it's motion controller is a me-too version of the WiiMote. At the end of the day, Sony as a consumer electronics company doesn't really know what it's doing when it comes to the software environment, particularly of a complex device like the PS3.

What's interesting about the current crop of consoles, however, is that they're all based on the PPC. The Wii is essentially a G3. The xBox is a 3 core version of the G5, and the PS3's main core is also derived from the G5.