In a post earlier today, I asserted that Intel wanted to get rid of the x86 platform. It occurred to me that a lot of folks may not understand this claim, but there are good reasons for it. They have made at least three attempts at doing so, in fact, starting as early as 1986, but the momentum of the existing hardware and software base has made it impossible for them so far.
Why would Intel want to kill the goose that laid the golden egg? Simple: it isn't the cash cow most people think it is. Most of their real business isn't in the desktop systems, it's in embedded systems (including mobile systems, these days), and they never intended to get into the small computer market in the first place. In 1977, when they first began designing the 8086, the home computer market was tiny, the reserve of crazy electronics enthusiasts who mostly built their own systems from kits (there were a handful of pre-built systems on the market, but even those were mostly seen as a novelty). That market was too problematic and too small for them to waste much time on; as far as they were concerned, they were making microcontrollers, thank you very much, and this whole business of 'microcomputers' was a passing fad. It probably would have been a fad, if Apple hadn't come along, but even then Intel wasn't very interested, as Apple was buying their microprocessors from a competitor. It wasn't until IBM decided to use the 8086 for the PC in 1981 that Intel really noticed that market, and even then they saw it as a small part of their business, and didn't expect it to be much of anything.
It wasn't until a couple years later, as they started designing the 80386, that they started to see that it could be a major business, but by that time everyone was pretty much stuck with a design that was meant for a completely different purpose and was already seen as outdated by most of the industry. RISC was the new byword for CPU design, but Intel was committed to the x86 architecture and couldn't seem to pull themselves out of the quicksand.
Even if they had foreseen the rise of PCs, the x86 is an aging design, and has become increasingly difficult to maintain and improve. They would much rather be producing a CPU that doesn't use most of its silicon real estate for underutilized features - they could put four RISC cores in the space taken up with just one x86-64 core, even considering the facts that for both designs most of the CPU space is taken up by cache memory, and that RISC cores have no microcode store and thus need each and every instruction implemented in hardware. The x86-64 design is a huge resource sink for the company, and they'd prefer to work with a more efficient and easily-maintained design if they could, but their hands are tied.
It's not that they haven't tried to promote something newer. The Itanium was the most public effort to do this to date, but not the only one. The problem was that by 1994, when they announced the Itanium, there was a huge and ever-growing software base for the x86, and they would have to find some way get that shifted over to any new CPU they came up with. The Itanium was a compromise: it wasn't x86 compatible by any means, but it was intended to make emulating the x86 fairly easy, but in practice this never panned out. This, combined with what seemed like an endless series of design missteps and release delays, caused the 'Itanic' sank without a trace.
By 2003, they were warming up for another attempt to bury the x86, when AMD blind-sided them with the EMT64 design. Pretty soon, they were playing catch-up on their own platform (this had happened before, when the Zilog Z-80 came out back in the 1970s, but it wasn't nearly as important back then). They have been duelling it out with AMD ever since, and while they're back on top of the heap (for now), they are at best lukewarm about the IA-64, and dearly wish they could just scrap it and start fresh.
