Opinion These are uncomfortable times for Intel and its investors. All the cool kids are talking about AI chips, an area where Intel has no real story to tell. Low power, embedded, and mobile have long been annexed by Arm. Now it looks like Arm is coming for Intel's most iconic territory - the PC.

Last week, Nvidia, AMD, and Qualcomm were each reported to have Arm-based PC platform plans in the works. These are no idle threats. Each company has a long history of designing with and integrating Arm, AMD in particular having an inexhaustible appetite for eating Intel's lunch. Apple Silicon has demonstrated that Arm can be best-of-breed across the PC landscape. Is there anything Intel can do? No, no, and thrice no - and the evidence is in Intel's deep history of innovation failure.

This is beautifully exemplified in a lengthy interview released last week on YouTube, in which
ex-Microsoft engineer Dave Plummer talked to octogenarian wunderkind Dave Cutler. Cutler is the father of DEC's VMS OS, after which he changed partner and parented its half-sibling, Windows NT. He's still at Microsoft more than 30 years later. If you want to hear the real deal about why XP was so very, very buggy, the tech Xbox stole from Azure, and which internal MS system was called SLIME, go and watch it all - perhaps skipping the motor racing stuff. The bit about the stuffing for puffer jackets, though - that matters.

Yet it is the story behind Windows NT that illustrates Intel's flair for failure, and why the barbarians are now pouring into its heartlands.

NT was initially developed not on x86 chips but a new Intel architecture called the i860 - codenamed N-Ten or NT for short.

The i860 was Intel's second go at a RISC chip, as well as its first at a 64-bit design, and was much vaunted as being very efficient and very fast. It could indeed be both, and had it succeeded then it could have dominated the nascent RISC market of which the young Arm was just one of many potential prey species.

As Cutler and his team discovered, though, while the i860 could be a fast and efficient graphics processor, it was a three-legged dog for multitasking operating systems like NT. Intel expected its dominance to see a bad system through enough iterations for the market to both help to fix things in software and be won over.

Instead, the market found alternatives that worked already - mostly Intel's own 80486. Windows NT, designed from the outset as multi-platform, got ported to Intel's good chips and a few others that later died on the vine. Intel considered RISC a niche market, licensed Arm for its XScale embedded processors, and went after the big iron with its 64-bit Itanium.

Microsoft dutifully ported its core Windows NT server code to Itanium to find it too was a three-legged dog, only perhaps with not quite so many legs and laughable backwards compatibility. Never mind, thought Intel, the market will wait for a few iterations, help build the code base, and be won over.

Meanwhile, AMD sidled up to Microsoft and asked if it was interested in 64-bit extensions to IA-32. Ones with a full complement in the legs department and total compatibility. Cutler and co. signed up on the spot. The result was so successful that Intel was forced to adopt its rival's innovation and let Itanium sink without trace.

That's two successive examples of Intel completely misreading the market. For the hat-trick, we need only look at the Atom, Intel's next portable processor. This happened when the company realized that the mobile market was exploding and Arm was riding the shockwave.

Ditching its own Arm license - because who needs lifeboats? - Intel declared that its chip could be as lean and mean as Arm, and running x86 code was exactly what the market needed.

Alas, the Atom proved nowhere near as mean and far less lean. Performance and power consumption never met expectations - familiar, huh?

Nor could they, not just because of the Atom's inherent limitations but because the market wanted to integrate mobile cores into SoCs - which made the entire system faster, lower power, and cheaper. x86 compatibility didn't matter because none of the codebase made sense on phones. Intel's worldview couldn't adapt to either in time - and now we have Arm SoCs that beat Intel's mainstream designs in areas that the company once owned outright.

Microsoft, of course, is as happy about that as Apple has been. Windows NT, long since transmuted into the heart of every flavor of the OS, has been ported to Arm and will merrily run on all the new non-Intel PCs powered by Nvidia, AMD, and Qualcomm. x86, once Intel's primary card, is now a distinct turn-off. Literally, if your Intel laptop drains its batteries faster than Arm.

Nor can Intel get back into that game. It's too late - the dynamics that lost it the mobile market are spreading to commodity PCs and laptops. They'll spread up the curve too. Intel once drove nearly everyone else out of town, nuking the workstation and the big iron processors because it had all the software and could use that dominance to force the pace of processor development. It no longer has all the software, having ceded mobile, and the market is pouring its money into TSMC's model.

Intel knows it must be a partner, not a dictator, and that collaboration and flexibility are all. When we'll see concrete results is the most important question. The history of the past 30 years, and of the past week, is that the market is never in a mood to hang around and find out. ®

https://www.theregister.com//2023/10/30/arm_intel_comment/