Freescale CEO on Apple, Intel and IBM

[AidenShaw]

Umm... I think the main reason the switch is happening is because the old 32-bit G4 (and its crappy FSB) simply cannot deliver the goods in today's laptop world. And anyways, the G5 can't go much faster in an iMac due to heat issues and Jobs said the transition would be complete in 2007 (Hint hint Power Macs). The pros outweigh the cons by a mile.

I think that your response missed the context....

I wasn't saying not to switch - I was saying that waiting a few more months and switching to 64-bit x64 might make more sense.

As it is, Apple will be forcing 2 transitions on its software partners - one to 32-bit OSx86, then later another one to 64-bit OSx64....

Perhaps it would be better to only do one switch, to 64-bit OSx64 with true 64-bit support (GUI/Cocoa/Carbon). That way, all MacIntel software would be fully 64-bit.

And remember that the usual "ballpark" figure is that x64 mode is 10-20% faster than x86 mode on the same hardware.

 

[ksz]

The PowerPC never really outshined Intel's processors.

Actually it clearly DID at times--and for some important tasks, it still does. The Intel change is, once again, about the future as much as the present. (The PRESENT need is for laptops--the G4 needs a successor.)

Just as AMD's current processors are considered better than Intel's in a variety of ways, the lead is not very significant and Intel is not nearly out of the game. Similarly, there were times when PPCs were faster by a significant margin and boasted innovations like AltiVec and fostered debates on RISC vs CISC, but those leads and those debates quickly evaporated as Intel gained steam and IBM/Motorola lost steam -- and lost it in a big way. Excitement over the PPC Roadmap that peaked on that historic day when Jobs introduced the PowerMac G5 has now become a dim memory, while continued use of the 7447 with a 167/333 MHz bus in the PowerBook is an ongoing reminder of being stuck in the past.

They were unable to increase performance and unable to innovate with new on-chip features and unable to produce parts quickly in needed quantities (a slow yield ramp particularly for the 90nm process at IBM's Fishkill plant).

The PowerPC may have left the gate running, but since the late 130nm period to the current early 65nm period, the pace has slowed to a crawl. They are not competing aggressively in a market that demands aggressive progress. The PPC is not outshining or outdistancing the competition in any significant way.

 

[nagromme]

I wasn't saying not to switch - I was saying that waiting a few more months and switching to 64-bit x64 might make more sense.

Would x64 run in some way on 32-bit Intel chips too? I hadn't heard that.

Or are you saying that within a few months, all PCs even on the low end (like the Mini) will have 64-bit processors? No more Dothans or Yonahs within a few months?

 

[Fukui]

As it is, Apple will be forcing 2 transitions on its software partners - one to 32-bit OSx86, then later another one to 64-bit OSx64....

Perhaps it would be better to only do one switch, to 64-bit OSx64 with true 64-bit support (GUI/Cocoa/Carbon). That way, all MacIntel software would be fully 64-bit.

While I would agree with that premise, I think customers wouldn't stand for having to re-buy or re-download thier apps just to get 64-Bit and just to run them on thier new Macs...

Perhaps apple has a strategy similar to Rosetta that will let 32-bit apps run along side, when apple does intend to ship x64 systems.

This may be better for developers because they can choose to move to 64-bit when they need/want to.

 

[generik]

While I would agree with that premise, I think customers wouldn't stand for having to re-buy or re-download thier apps just to get 64-Bit and just to run them on thier new Macs...

Perhaps apple has a strategy similar to Rosetta that will let 32-bit apps run along side, when apple does intend to ship x64 systems.

This may be better for developers because they can choose to move to 64-bit when they need/want to.

Erm.. 32 bit programs can run on a 64 bit system, however device drivers that operate in kernel mode are a different story altogether.

 

[Fukui]

Erm.. 32 bit programs can run on a 64 bit system, however device drivers that operate in kernel mode are a different story altogether.

Hmm, Ok but why then is there a big stink about making everything 64bit?
I thought X64 could only run in 64-Bit only mode or only 32-bit?

 

[shamino]

Hmm, Ok but why then is there a big stink about making everything 64bit?
I thought X64 could only run in 64-Bit only mode or only 32-bit?

A lot of people seem to believe that 64-bit means 32-bit incompatibility. But it's not true. At least not in the AMD-style 64-bit model (which Intel is supporting in their newest chips.) This is very different from older 64-bit chips (like Itanium) which are not compatible with 32-bit code.

As for making everything 64-bit, it could be done, but there's really little point to it. Unless your program requires large registers or more than 4G of address space, moving to 64-bit is counterproductive. You gain no additional performance, and increase the memory footprint (since memory pointers all become twice as large.)

Apple's solution with the G5 (support both 32- and 64-bit code) is best. Let the developers choose which model to code their apps to. (But I do hope they eventually ship 64-bit equivalents to all of the APIs, so 64-bit code doesn't need to shunt all UI calls though 32-bit code.)

Sun used a similar solution for Solaris - a 64-bit kernel that supports both 32- and 64-bit apps, so it's not like this is a terribly unusual concept.

 

[shawnce]

As for making everything 64-bit, it could be done, but there's really little point to it. Unless your program requires large registers or more than 4G of address space, moving to 64-bit is counterproductive. You gain no additional performance, and increase the memory footprint (since memory pointers all become twice as large.)

Not fully true in the case of AMD's (Intel's) 64b ISA since when in that mode they took additional liberties to change the ISA giving the compiler/programmer access to more named registers. So you can see a performance increase even for applications not needing 64b addressing.

In the case of PPC you have the same number of registers regardless of using 64b or 32b addressing. So the negatives of moving to 64b addressing isn't offset by gaining access to more named registers.

...at least that I how I understand things the last time I looked this over...

Apple's solution with the G5 (support both 32- and 64-bit code) is best. Let the developers choose which model to code their apps to. (But I do hope they eventually ship 64-bit equivalents to all of the APIs, so 64-bit code doesn't need to shunt all UI calls though 32-bit code.)

Yes what Apple is doing for G5 (PPC ISA) is a good thing and I would love to see full support for 64b addressing across all frameworks / libraries on Mac OS X.

 

[shawnce]

Or are you saying that within a few months, all PCs even on the low end (like the Mini) will have 64-bit processors? No more Dothans or Yonahs within a few months?

He is saying (and what I generally have been expecting myself but...) if Apple waits a little bit they could skip Dothans and Yonahs and go direct to first of the next generation cores which all have 64b addressing capabilities. The amount they have to wait is really close to being able to have product shipping by WWDC 2006, as they implied at WWDC 2005.

I will say that I expect the Intel replacements for Power Mac to be using a 64b capable core hence those as expected will likely be near the last to switch to Intel.

 

[Kai]

The story probably goes like this:
IBM makes the G5 because they themselves can put it in smaller blades, which the Power4 is unsuited for. They provide the chip to Apple at no design cost, also as a show of good faith.
Then Apple wants a Laptop G5. IBMs beancounters say "well, we can't make use of this chip ourselves, so this will cost you $xy Mill for design!"
Then Jobs decides to throw a tantrum because the millions of blind followers (like many on this forum for example) that believe his every word actually inflated his ego into perverse proportions and made him believe he is god and he can do whatever he likes and everyone has to bow to his will and do stuff for him at no cost. Even though he has $8 billion in the bank now, mind you.
IBM doesn't balk, Apple refuses to pay for design, then IBM decides to focus on the people that DO pay for their custom designs: Sony/Toshiba and M$.
Apple feels neglected and El Turtleneck decides to redefine the concept of a hissy fit by switching to Intel. He tells a few lies on stage about the IBM that couldn't to make aforementioned blind followers recite his "point of view" in forums the world over to take his sweet revenge on IBM, because they wouldn't bow to The Diva's will!

Unfortunately Mr. Jobs hasn't spent a single thought on the consequences in his blind rage. For example that Intel has just blown the whole P4 and is lagging badly behind AMD, that the transition to 64bit is just fullscale underway on x86 (and here - contrary to PPC/64bit - there is actually some nice performance to be gained through extra registers, the last "free lunch" before they all have to multithread their code for Multicore-CPUs!) or that he will be producing exactly one single line of non-64bit-capable Intel-Macs that will require testing and compiling of 32bit-versions of ALL PROGRAMS atleast until 2010!
What el Stevo also forgot was that Apples own Pro-Suite and ********s of Audio-Software makes heavy use of Altivec. And no: SSE is *not* a substitute, because a 64bit ALU still remains a 64bit ALU!... A 3.6 GHz P4 with Hypnothreading is 3 times slower in Altivec-Emulation than a 1.5 GHz Powerbook G4...

And what most Macrumors-Readers fail to realize is that sites like this will mostly cease to exist. As will the appeal of Keynotes for anything that's not software... Because: Who exactly is holding their breath to see Apple releasing Computers with tech that Dell and others have been selling for the past 3 months?

Please check which vendor will get the latest Intel-chips first:

[ ] the vendor from Texas with 18% global marketshare
[ ] the vendor from California with 3.5% global marketshare

You may hold your breath for a fancy new case-design, i know i won't... Optics are a nice bonus, but putting a pwetty bow on a pile of shyte won't make me crave it!

And hey Intel-Lovers, you're in for a big surprise: OS X will still be slower on Intel than Windows or Linux. Why? Because Windows and Linux have had YEARS OF OPTIMIZING on x86! And they're both still using a GUI that is generations behind OS X (well, we can only hope for Vista to slow down Windows - a lot!). And Software-Developers will STILL optimize their stuff for Windows first and Mac only as an afterthought!

And you're in for even more surprises: Game developers will still develop for DirectX which means the OpenGL-Mac-Versions of Games will still be stuck with not-as-well-optimized Ports from Mac-Gamecompanies. Windows will still get new hardware way earlier than OS X because NOTHING in the drivers can be re-used for OS X' completely different driver model & APIs.
NOTHING will actually change for the better! Apple may be using the same Architecture as the other 95% of the planet, but OS X will STILL HAVE 3.5% MARKETSHARE!
Okay, *something* will change: You'll hear ALOT more "sorry, we won't do a Mac-Version - why don't you just install Windows?" And you'll be able to run Windows-Programs in Emulation much faster than before. It completely defeats the purpose of using a Mac in the first place, but hey, who's gonna be picky? Hooray, the future is bright! \o/

We will simply see no more hardware innovations like Firewire, Gigabit-Ethernet (comes in pretty handy now with Xsan and Xgrid, right, Steve?) or Airport long before everyone else got them. Why? Because in his hissy fit Uncle Steve didn't consider what consequences it has when a vertical vendor loses control over the hardware completely....

"Macs are working so well because Apple makes everything" - yeah right, not anymore, dude! Wake up and smell the coffee!

Phew, but that Yonah surely rocks, yeah! Why don't you start thinking for only a fraction of a second of consequences before you blindly reurgitate Steve's lies?

 

[Flynnstone]

Seems like a pretty weasely move on IBM's part--

Who's the weasel? Mayer?

 

[Kai]

And anyways, the G5 can't go much faster in an iMac due to heat issues

Complete BS. The new iMacs are much much more silent and don't even start their fans when you put them under load. Makes you wonder why when the G5 is operating at its limit, right?

The pros outweigh the cons by a mile.

More BS. Read above. I'd love to see you refute this!

 

[Flynnstone]

What el Stevo also forgot was that Apples own Pro-Suite and ********s of Audio-Software makes heavy use of Altivec. And no: SSE is *not* a substitute, because a 64bit ALU still remains a 64bit ALU!... A 3.6 GHz P4 with Hypnothreading is 3 times slower in Altivec-Emulation than a 1.5 GHz Powerbook G4...

True. Altivec is superior to SSE. I think Apple has a trick up its sleeve. I believe all Apple computers have a Real video card. Real graphics cards are hardware accelerators. The GPUs have more horsepower than the CPU. AMD/Intel/PPC, doesn't matter. But GPUs are more like Altivecs (&SSE). The trick is to get the data from the CPU/memory to the GPU efficiently.

Also I don't believe too much of what this Mayer guys says. He's not really in the picture anymore. Also I don't know of any 16bit PowerPCs as he says.
Perhaps he's just a snake oil salesman.

 

[Kai]

Of thinking Apple's decision NOW is based on factors from years ago that Apple just didn't see. Intel's future (Pentium M and beyond) really IS different from its past (Netburst.) PowerPCs have done VERY well, but they're not going any further. Intel is.

What makes you think IBM will cease its CPU-Development? With all their "Power now" efforts and the licensing and all? And do you actually believe the new consoles will be the last generation of consoles ever? And that all of them will be switching (back in M$' case) to Intel with the next generation?
Wait for IBM to show some sample of the Power6 on next year's MPF.

Oh, have you noticed IBM already has an integrated memorycontroller in the Power5 and there's NONE to be found AT ALL on Intels just recently castrated roadmap?

What Apple did do is keep their options open--which took no small effort! Very good planning.

Yeah, just like Intel made Yamhill with x86-64 in case the Itanic failed. However, having a Plan B doesn't mean that Plan B is the better choice. Especially if you're solely responsible for Plan A failing!
And as for "keeping options open": Apparently Steve doesn't like to have choices all that much anymore these days, otherwise he wouldn't have wed himself exclusively to Intel and would not SWITCH to x86, but sell PPC alongside x86!

Anyway, the PowerPC had great potential, as some G5 quad owners can attest! I hope it goes on in other forms--Cell, etc.--even if it can't have a future in computers.

Just think of a G5 with a Cell as a co-processor for FCP, Motion, Shake, Logic, Core Image and Compressor, together with a nice API 3rd party developers could make use of. Just think for a second and start weeping. I don't know how Sony will manage to make use of this chip for a game console, but i do know for certain that it's absolutely perfectly suited for said programs/tasks!

 

[skunk]

A 3.6 GHz P4 with Hypnothreading is 3 times slower in Altivec-Emulation than a 1.5 GHz Powerbook G4

If it's using Hypnothreading, I'm not surprised. Falling asleep on the job.

 

[Kai]

Just as AMD's current processors are considered better than Intel's in a variety of ways, the lead is not very significant and Intel is not nearly out of the game.

"Not very significant". Yeah, right..

The PowerPC may have left the gate running, but since the late 130nm period to the current early 65nm period, the pace has slowed to a crawl. They are not competing aggressively in a market that demands aggressive progress. The PPC is not outshining or outdistancing the competition in any significant way.

Incase you didn't notice: The Slowdown in clock scaling and performance-scaling was true for the WHOLE industry during the past 2-3 years! I remember when AMD effectively advanced 70 MHz (!) in one year! And when Intel was stuck AGES on 3-3.2 GHz before the Prescott came out (which was a dud basically and totally didn't deliver on power promises!)

Which is why everybody is going multicore now, and IBM wasn't exactly very late with the 970MP, not in respect to their own roadmap neither in respect to Intel and AMD (infact Intel released their Dualcore-Xeon (again a huge dud!), the only contender for a 970MP because the Pentium D doesn't do Dual-CPU also in November AFTER the first 970MP Macs shipped!)

 

[Kai]

If it's using Hypnothreading, I'm not surprised. Falling asleep on the job.

I do sincerely hope you don't actually think this was NOT a joke on purpose.. Otherwise I'd really ultimately lose belief that humor on the Internet ever works!

 

[skunk]

I do sincerely hope you don't actually think this was NOT a joke on purpose..

I'm very glad to hear it. Obviously I'm not up to speed with PC humour.

I'll get my coat...

 

[Kai]

A lot of people seem to believe that 64-bit means 32-bit incompatibility. But it's not true. At least not in the AMD-style 64-bit model (which Intel is supporting in their newest chips.) This is very different from older 64-bit chips (like Itanium) which are not compatible with 32-bit code.

Hmm, i kinda wonder what that big chunk on the Itanic-Die labelled "IA32" is for then! ;-)

As for making everything 64-bit, it could be done, but there's really little point to it. Unless your program requires large registers or more than 4G of address space, moving to 64-bit is counterproductive. You gain no additional performance, and increase the memory footprint (since memory pointers all become twice as large.)

This may be true for PPC and other ISAs that were designed with 64bit in mind right from the start. However: AMDs x86-64 is a completely different ballgame. You don't just have twice the registers now (keep in mind x86 is traditionally register-starved!) but also other nice enhancements that definately speed up ANY program made for it instead of x86-32!

 

[Donm]

Who exactly is holding their breath to see Apple releasing Computers with tech that Dell and others have been selling for the past 3 months?

Please check which vendor will get the latest Intel-chips first:

[ ] the vendor from Texas with 18% global marketshare
[ ] the vendor from California with 3.5% global marketshare

The one thing I can comment on is this...
Dell and apple will get access to the latest chips at the same time. What they choose to do, and when is another matter. I'm actually a PC guy, but have recently experienced macs for the first time. From this experience I can tell you that I would not buy another Dell. The quality is not there.

I'm sure you probably know this already but apple is actually growing much quicker than any other PC OEM, with higher margins and the most loyal following. Being on the same platforms as Dell and the others forces their hand to some degree because people will be able to take intel roadmaps and determine what apple should be doing, but at the same time apple is an innovator compared to any other PC OEM, and they will take intel technologies and design new things that the old pc oem's would never. Why? because they(Dell, HP, ect) wait for intel to make the standards, design the reference platforms and build white boxes, so all the have to do is validate, add their logo and box (ok, so I'm over simplifying, but there's some truth to it). Apple will design new usages around new technologies, like VT, like AMT, like LT, etc. Intel will use apple as the golden child against their other customers, showing them how closely they should be working and that the oem should be pushing the envelope, not just build black or white boxes. anbd imagine the leverage on MS as apple adopts hardware technologies into the OS that MS didn't want to, or didn't plan on it for another refresh. All of the sudden MS is on the defensive when it comes to platform technology adoption and integration.

 

[Kai]

The one thing I can comment on is this...
Dell and apple will get access to the latest chips at the same time. What they choose to do, and when is another matter. I'm actually a PC guy, but have recently experienced macs for the first time. From this experience I can tell you that I would not buy another Dell. The quality is not there.

I'm sure you probably know this already but apple is actually growing much quicker than any other PC OEM, with higher margins and the most loyal following. Being on the same platforms as Dell and the others forces their hand to some degree because people will be able to take intel roadmaps and determine what apple should be doing, but at the same time apple is an innovator compared to any other PC OEM, and they will take intel technologies and design new things that the old pc oem's would never. Why? because they(Dell, HP, ect) wait for intel to make the standards, design the reference platforms and build white boxes, so all the have to do is validate, add their logo and box (ok, so I'm over simplifying, but there's some truth to it). Apple will design new usages around new technologies, like VT, like AMT, like LT, etc. Intel will use apple as the golden child against their other customers, showing them how closely they should be working and that the oem should be pushing the envelope, not just build black or white boxes. anbd imagine the leverage on MS as apple adopts hardware technologies into the OS that MS didn't want to, or didn't plan on it for another refresh. All of the sudden MS is on the defensive when it comes to platform technology adoption and integration.

Hmm, looking at your location makes me pay extra attention to your words, that much is certain! ;-)

I see where you're coming from, and this sounds like a plausible scenario, in fact this more or less exactly was my first (then still positive) idea of what the future might be like with Intel and Apple. However, there are some things to consider: Neither M$ nor Dell will be content to play the second fiddle to Apple at Intel, so they will take measures to ensure this is not happening. Dell could f.ex. switch to AMD, which would hurt Intel bigtime (Biggest vendor, 18% Marketshare, all Intel!). M$ could do the same and choose to support AMDs versions of said Intel technologies instead of the Intel versions f.ex.
Another thing to keep in mind is that Intel is by far not the sole advancer of technology in the PC world. Alot of other companies (e.g. Nvidia/ATI for GPUs etc.) contribute to the general advancement just as much, and they will continue to focus on Windows first. This means that for anything that's not straight Intel tech Apple would still be way behind Windows! And how often does Intel release really new technology that would be worth to build custom-made stuff for? Once every 2 years? I can't even remember the last one, i think that would be SSE3 (whose usability beyond SPEC is questionable) or Hypnothreading!... No, I don't count the NX-Flag and x86-64 as Intel-Tech! ;-)

And thanks, i'd rather prefer not to see what "innovative" uses Apple finds for LaGrande! ;-) I don't like my own computer locking me out of things on my harddrive!...

 

[nagromme]

He is saying (and what I generally have been expecting myself but...) if Apple waits a little bit they could skip Dothans and Yonahs and go direct to first of the next generation cores which all have 64b addressing capabilities. The amount they have to wait is really close to being able to have product shipping by WWDC 2006, as they implied at WWDC 2005.

I know, but that depends on two things that would be news to me:

1. Merom and Conroe being ready in EARLY 2006 instead of late 2006-2007 as was the consensus, I had thought.

2. CHEAP, LOW-end Merom-series chips being ready at the same time. Otherwise, you can't have ALL Intel Macs be 64-bit. You've left the Mac Mini and iBook out in the cold.

I had thought Yonah variants (like Yonah1/Core Solo and Sossaman) are expected months after Yonah (Yonah2/Core Duo) comes out. So I don't see how they'd immediately dump Yonah1 AND Dothans, replacing them all with new 64-bit CPUs that are cheap enough for bottom-end machines. Not in the "few months" time frame.

And if we're talking a year or more, that's too long to stick with G4.

And do you actually believe the new consoles will be the last generation of consoles ever? And that all of them will be switching (back in M$' case) to Intel with the next generation?

The fact we've been discussing, though: IBM isn't getting out of the PowerPC, they are focussing on PowerPC for game consoles instead of for computers. The needs are NOT the same for computers and for consoles. Especially when it comes to laptops. So yes, of course there will be more consoles in future, very likely non-Intel ones. That won't help Apple make a PowerBook G5.

The emotional rollercoaster that results from the Intel change is a short ride for some, a longer one for others Either way, enjoy the ride! Big changes are fun to watch, and this one happens to be for the better.

 

[~Shard~]

Hmm, Ok but why then is there a big stink about making everything 64bit?
I thought X64 could only run in 64-Bit only mode or only 32-bit?

Hopefully by the time I buy my next Mac in 2007 everything will finally be really, truly, honestly, 100%, really, actually 64-bit.

 

[Donm]

Hmm, looking at your location makes me pay extra attention to your words, that much is certain! ;-)

I see where you're coming from, and this sounds like a plausible scenario, in fact this more or less exactly was my first (then still positive) idea of what the future might be like with Intel and Apple. However, there are some things to consider: Neither M$ nor Dell will be content to play the second fiddle to Apple at Intel, so they will take measures to ensure this is not happening. Dell could f.ex. switch to AMD, which would hurt Intel bigtime (Biggest vendor, 18% Marketshare, all Intel!). M$ could do the same and choose to support AMDs versions of said Intel technologies instead of the Intel versions f.ex.
Another thing to keep in mind is that Intel is by far not the sole advancer of technology in the PC world. Alot of other companies (e.g. Nvidia/ATI for GPUs etc.) contribute to the general advancement just as much, and they will continue to focus on Windows first. This means that for anything that's not straight Intel tech Apple would still be way behind Windows! And how often does Intel release really new technology that would be worth to build custom-made stuff for? Once every 2 years? I can't even remember the last one, i think that would be SSE3 (whose usability beyond SPEC is questionable) or Hypnothreading!... No, I don't count the NX-Flag and x86-64 as Intel-Tech! ;-)

And thanks, i'd rather prefer not to see what "innovative" uses Apple finds for LaGrande! ;-) I don't like my own computer locking me out of things on my harddrive!...

you know your stuff, I'll give you that.

 

[ksz]

"Not very significant". Yeah, right..

Yeah, right. There are several pages of benchmarks here published by Tom's Hardware. AMD's processors have a nice performance edge almost across the board, but the difference is "not very significant". Your link shows 2x improvement over Intel, but that's not evident in Tom's benchmarks. You can always twist benchmarks just as you can twist a "study" to suit your foregone conclusions.

Incase you didn't notice: The Slowdown in clock scaling and performance-scaling was true for the WHOLE industry during the past 2-3 years! I remember when AMD effectively advanced 70 MHz (!) in one year! And when Intel was stuck AGES on 3-3.2 GHz before the Prescott came out (which was a dud basically and totally didn't deliver on power promises!)

Which is why everybody is going multicore now, and IBM wasn't exactly very late with the 970MP, not in respect to their own roadmap neither in respect to Intel and AMD (infact Intel released their Dualcore-Xeon (again a huge dud!), the only contender for a 970MP because the Pentium D doesn't do Dual-CPU also in November AFTER the first 970MP Macs shipped!)

Intel developed two generations of the Pentium M as well as an ultra-low voltage version of the chip. They introduced HyperThreading and will have Virtualization Technology on the desktop with Yonah. Intel has also pushed aggressively on new process technologies and have had better success than IBM.

To their credit, IBM has produced noteworthy advances (alone or in partnership) in process technology including the copper damascene process, strained silicon, SOI, double-gated FinFETs, improved junction properties with high-k dielectrics, etc. etc. The problem the industry as a whole encountered 2-3 years ago was power management and high leakage currents. These two issues brought conventional scaling to a virtual standstill as the average power density increased to about 13 Watts / cm^2. A steam iron, by comparison, dissipates 5 Watts per cm^2. The industry as a whole rode the CMOS power curve up to its very limits, and is actively searching for new materials and techniques to continue to increase both performance and packing density while managing heat dissipation and leakage. This is a very difficult problem, which is a key reason for the paradigm shift away from raw Mhz to increased function.

In effect, if you cannot continue to jam more speed, then you must jam more features. This is the driving force behind dual and multi core processors, VT, additional FPUs, improved vector units, more L1 cache, etc. More features are going on-chip because the customer is not going to pay top dollar without a good reason. Clock speed has been the historical justifier for top dollars, but that's changing.

I don't fault IBM for technological incompetence. I do fault them for problems with execution in the time to ramp yield, in the time to introduce more differentiation based on features, and in the time to introduce low-power mobile models.