Lundy I think you are really close to the truth. And there is ample evidence to support your hypothesis.
First of all we have to recognize that Steve Jobs has a big ego and he wants to be second to none. I think thats true. I dont think its a bad trait as long as that ego can be tempered with reason and wisdom. I also think Steve has big ideas: he wants and needs certain freedoms and independence from the PC industry at large to implement his ideas.
Intel has to have some plan that really impressed Steve to switch teams like this. Many including myself dont believe its only because of power consumption.
Second fact. Intel wants to move on from x86. I havent been reading much about Intel for the past few years, so Ive had to catch up the past couple of months. X86 essentially is an old technology, largely the only reason it is still used is because of backward compatibility. The disadvantage of this as new technology is added to x86 little of the old technology is taken away. That has ultimately made x86 very inflexible. X86 cores have millions of transistors. As more transistors are added the chip consumes more power and produces more heat. Then there is the on chip x86 instructions decoder which takes a significant amount of room and power.
Over time billions in R&D are being put into x86 design to work around the problems of x86, and keep it going. There are in fact newer chip architectures which have significantly far more room for growth and innovation. Architectures such as IBMs Power and Intels Itanium.
Intel has also been in a low profile struggle with Microsoft. The two companies dont seem to be headed in the same direction any more. Ive read that new markets such as China are not openly embracing Intel because, to embrace Intel brings Microsoft.
As Lundy said Itanium is based on VLIW (Very Long Instruction Word). VLIW uses software compliers to translate x86 instructions into VLIW instructions. Over the past ten years Intel has invested billions of dollars of development into Itanium architecture. But the Itanium has proven to be largely a failure in the PC market. At this point Itanium is the scrappy step child that most kick around, even calling it the Itanic (Titanic).
Mostly from what I see the Itanium is so scorned because of its attempt to move the PC industry away from x86. I read that its first version was very poor, but has greatly improved over the years. The Itanium is currently used in high end workstations, mainframes, and super computing.
Intel exclaims Itanium is its answer to RISC processors most namely Power PC. Intel needed an answer to the Power 4 architecture which the Mac G5 is based on. Itanium was supposed to be that answer but no one supported it.
CEO of Intel has said:
Long term, the architecture Itanium needs to aim at is [IBM's] Power line. We have nothing in our existing 32-bit line capability that can compete with Power. It's a very high performance line.
When Itanium was introduced Intel billed it has its next evolution in chip architecture and a new instruction set (IA-64). Intel was partnered with HP on Itanium design and implementation. Nearly all of the major PC world announced support for Itanium: Microsoft, IBM, SGI, Compaq, SUN.and the list goes on.
In reality all of those companies bailed on Itanium support. IBM obviously because of its competition with Power. The rest largely because of continued support for x86. Even HP is pulling out its support. All of HPs Itanium designers have now gone to Intel, and HP may be bailing on Itanium support altogether.
Intel did not want to implement 64 bit computing into its x86 architecture because it wanted to move the industry over to Itanium which is natively 64 bit. That misstep left AMD room to implement its own AMD64 bit standard, which Microsoft adopted as its 64 bit instruction set in XP 64. To stay competitive that forced Intel to adopt AMD64 in its future line of x86 chips, which it calls EM64T.
Intel is lukewarm on its support for AMD64, calling it a feature enhancement for x86. They still promote Itanium and IA-64 as a fundamental change in architecture.
Today the few working Itanium stations are mostly running Linux. Microsoft has weak server support for Itanium, and cancelled its plans for an XP version running on Itanium. This leaves Intel holding a billion dollar boondoggle.
However what may save Itanium or more specifically VLIW architecture is the advancement of Power PC. IBM can adapt Power architecture in ways that are nearly impossible with x86. That is the reason Xbox 360, Playstation 3, Nintendo are using it. Itanium (VLIW) is the only other major architecture that can keep pace.
Intel has made some interesting acquisitions. The original VLIW designs that inspired the Itanium were from a Russian company called Elbrus. Intel bought Elbrus last year, and is integrating its designs into Itanium. HP engineers from the Itanium team were transferred to Intel. And Intel has gained software that allows x86 software to run on RISC microprocessors, which is related to running x86 binaries on Itanium.
X86 cores need out of order execution, have small architectural registers, and the x86 instruction decoder takes a significant amount of energy. Intel Pentium chips have extremely long pipelines, which draw more power and produce more heat.
VLIW chips can be designed with fewer transistors than commonly used in x86 currently, by only using transistors that significantly increase performance, x86 decompiling is done in software that does away with on chip decoders, has a short pipeline, strictly in order execution, and no branch prediction. VLIW chips can use multiple low power cores, larger file registers (SSE can be made to compete and surpass Altivec), and support hyperthreading.
This architecture allows for processors that will run at 2 to 3 Watts and are just as fast. Same as Jobs has described.
To hear this sounds as though Intel is continuing to support and enhance Itanium, inspite of the fact that the PC industry at large is ignoring it.
If you have a great technology that none of the mainstream OEMs will support (because it is too new) who do you call?
Where does Apple come into this? Tom Yager from Info-World describes it.
"IBM and Freescale, Apples previous CPU suppliers, both said the same thing about Apples departure: Thank God, now we can get back to serving our other customers. Apple pushed IBM and Freescale to advance PowerPC technology further and faster than the majority of their customers expected or required. I hope Apple makes itself another pain in Intels ass."
At this point Intel is being beaten by Power 5 architecture in the supercomputer scale, and is being beaten on the desktop and lower server scale by AMD. Intel needs a partner to help them innovate.
Strangely enough John C Dvorak predicted this two years ago, and gives as a good explanation as any to why Apple is going with Intel.
Itanium. What will be radical is the company's choice of processor. Apple will announce its use of the Itanium chip, which can be used in such a multiprocessor design and will become the first desktop use of the chip. The choice of the Itanium is suggested by four factors.
First, there is zero evidence that Apple is talking to AMDand it would if it were staying with the x86 legacy chips.
Second, Apple likes to make jazzy announcements in which it claims to be the first or the most aggressive in a market. The Itanium fills the bill perfectly, because Jobs can lord it over current PC makers with all sorts of performance claims.
Third, if Apple optimizes the OS X kernel for the Itanium, the likelihood of the Apple OS being ripped off by normal PC users is nil. And finally, by choosing the Itanium, Apple will have an ally in Intel, who will put its design team to work for Apple and perhaps even invest in the company, knowing AMD is not in the picture.
There is no doubt that a MacIntel machine could supplant the Wintel platform. And most likely, the entire hungry Linux community could port all the x86 Linux code to the MacIntel OS within weeks, creating a huge flood of good products.
A Macintel running an Itanium processor would be a huge shift in the computer industry.
First and most obvious would be Apple and Intel working together with Intel processors inside of a Macintosh. Thats weird enough in itself.
Steve Jobs will have teamed with a larger supplier and the ability to have common Intel motherboards using common PC components. Apple will also have the ability to be on the ground floor of pushing another powerful CPU architecture.
With the major PC industry at large treating Itanium as a laughing stock. Apple using Itanium would validate Intel and its investment in VLIW and IA-64. A Macintosh and OS X powered by the Itanium is huge validation and marketing for the platform.
Its doubtful Apple will use the current Itanium design as it was not made for the laptop or low power consumption. Plus the current Itanium line costs $900 to $4000. I imagine Intel will have to design a new line that scales from laptops, desktops, to servers.
As far as the time table I would guess Apple will need to bring laptop and desktop Itaniums right away. They may introduce x86 computers first, but that forces Apple to support three different OS types. PowerPC, x86, and VLIW.
The current developer boxes are to be returned as Apple has said they will be useless in the future. My guess is they will need to switch to Itanium with Tiger in 06.
One potential problem Im not sure of. If OS X is VLIW native to gain optimum performance applications will need to be complied in VLIW instruction set. Of course the chip can decode x86 to VLIW, but the app would be faster if it were VLIW native. It should be easier to recompile from x86 to VLIW than it is to recompile from PPC to x86.
To look at the whole situation it makes sense.
But we will have to wait and see how it all plays out.