Intel and x86 have their days counted
The ball just went out of their court
Just as Apple killed the aging PowerPC architecture from mainstream by migrating to Intel x86 powered chips, the Cupertino based company is doing it again in favour of ARM.
The once considered leader of x86, Intel, slept on the job for a decade. You wanted performance, you would go over x86. No more.
The 40-old architecture didn’t show its age until Apple invested highly on ARM by buying PA Semi in 2008, the same folks that gave some light to PowerPC after it went “open” bacause IBM was already out of ideas. From there, the company made iPhone chips powerhouses by themselves, with each iteration dropping support for older designs and features — something that x86 never could possible do without the risk of breaking the core of the world itself, as almost all software has been made for x86. They know mobile was the future, bet on it, and paid off. This is x86 big demise.
This is a history lesson on neglecting the future.
Winter came for Intel and they didn’t see it
Today, Intel problems are three: no market share on the mobile markets, losing against a competitor (AMD), and competitive but not disrupting products. They have money, but no amount of it can buy you the time they lost doing the same. As Tony Stark once said:
No amount of money ever bought a second of time.
Add on top of that their roadmap gets very hazy in the medium term. There are Xe graphics that won’t blow nothing out of the water, and they’re still behind AMD and Apple process nodes. Intel Optane has been a niche gimmick. The acquisition of Intel Enpirion by Mediatek is just another way to say Intel lost money and wanted to recoup losses.
But Intel problems came before Apple made its “We’re switching to Apple Silicon” announcements. Much before Apple cut ties with them after the disastrous Skylake architecture for their Macs mid 2015. It was their neglection on mobile, something that started happening in 2006, and culminated in 2019.
History lesson on neglecting the future
2006 was the year of “Water in Mars” and climatic change awareness. Intel would sell their “communications” divisions for anything considering the money spent never returned.
Apple announced a year before transition from PowerPC to Intel, and quickly became the spotlight as media rumoured. While there was exceptisism about putting an old Pentium 4 instead of an Athlon 64, common sense says that Apple knew about the Conroe architecture months before anyone, and Intel offered great supply. In 2006 the Mac Mini came with Core 2 Duo, confirming the latter. Subsequent new processors from Intel kept AMD irrelevant until 2017, even forcing them to sell Imageon to Qualcomm for a dime to keep afloat. During these years, CPU progress stagnated thanks to the lack of competition, and Apple was the collateral.
The problems became big when the iPhone was introduced in 2007, aiming to become the next big thing. It was. It even broke the mind of Steve Ballmer, CEO of Microsoft of that time. There was no more need to bring your old phone or aging laptop to check your email, or create notes on the go, no keyboard needed; you could do it in a way it didn’t suck. Apple acquisition of PA Semi proved ARM was up to big things.
From there, Intel tried to push Centrino Atom processors to the market. These x86 powered chips were energy efficient, aimed the rising portable market, but made the critical mistake of being x86. In other words, Intel, in their cruzade to keep control on their golden eggs goose and push to the market faster with a well known ecosystem, offered to the public cheaper and smaller portable machines. Clients ended up with the new Celerons of old time: processor ecosystems that became infamous for their atrocious, terrible performance and battery life. Atom powered netbooks we’re just unreliable 9-inch laptops. Tablets where so awful bad both in hardware and software-wise nobody remembers them.
Apple put on the market the coup de grâce called iPad in 2010, which as of today, is still the best tablet in the market, if not the only one. Netbooks, Tablets and Atom died in the mainstream. Intel got a big sign saying the market was growing but Apple was leading it. No tablet or other offering came close to what Apple did. Even Windows 8, with two years to check what made the iPad so successful, failed.
Yes, at that point, the writing was on the wall: mobile was the future, and ARM was the architecture of choice… on mobile. It baffles me how Intel lost the opportunity to just buy ARM Holdings in 2010 considering they bought McFee the same year for almost 8 billion dollars. Not even a competitive ARM chip with the Intel name on it to offer, or an alternative architecture considering the talent they held. Nope. Nada. So much for financial horsepower.
Change is bad when you’re not the one changing
ARM desktop processors never had a chance on high performance desktop, notebook and server ecosystems. High performance was still exclusive to x86, and even Microsoft and Sony put their gaming division on AMD mostly thanks to a good deal from the technology firm. It seemed that ARM would be forever relegated to mobile chips.
That was until Apple closed the gap with their offerings next years.
The “ARM can’t reach x86 performance” lasted until people started to note mid 2016 when iPads and iPhones started to show suspicious performance gains. Tasks that took ages on PC, like transcoding videos or editing large images, where snappier on these tiny machines — many of them thanks to ARM-specific optimizations and co-processors introduced by Apple. You even had the infamous “What’s a computer” 2017 ad, where Apple tried to push iPad Pro as the notebook replacement. ARM was not something just “efficient”, its performance and feature richness became greater when Apple introduced ML acceleration cores.
In Skylake years, Apple knew Intel was aging worse than PowerPC considering the market was not PC-focused anymore, but now they had their own kingdom to serve. Now that they’ve jumped ship, there is no turn around for Intel, and surely iMac are the next in line to be replaced once they figure out how to make the next “Apple M2” faster for professional workloads. May be introducing a new dedicated pro-GPU? Widening their PCI Express lanes to allow faster storage and devices? Making their Memory Controller support up to 256GB of memory? After that, they only have to look for someone who wants to make a fortune making millions of Apple Silicon chips, like TSMC.
The Macbook M1 and Mac Mini M1 are Apple not bets, but products of the future, today. A good user experience across the board, not enough to kill Intel and AMD, but to warn them. They’re still behind on the SoC department if you think about the high performance market, where electricity is still not a problem. If you have a RYZEN desktop, or an Intel iMac, it will still be relevant for the this decade. Also, remember that companies change machines once they die, which explains why so many people still runs Windows XP, so this may take a while.
Intel demise is near, AMD will follow
This will be a domino-esque scenario: if the so called “Apple Silicon” proves that the best computing experience — tools, ecosystems, apps, battery life, performance — is not on x86 systems, and offers a competitive price, people will switch. The best part is that they won’t notice the architectural change, apart from switching from Windows to mac OS. Good luck trying to synchronize work between a PC and a Mac: it can be done but is a PITA.
To paraphrase Clayton Christensen, Intel stood with the present, the x86 architecture, but never decided to invest on the future, or even a real competitor to ARM. Now it’s too late, the competitor to ARM is called RISC-V, something that will be very popular once companies understand they won’t need to meddle with NVIDIA royalties and their slowness to push an efficient product to the market. Apple has an architecture license, meaning that NVIDIA will benefit slightly from per-chip royalties, if else.
The x86 architecture was way up Intel rectum and tried to force it where it didn’t belong. AMD took the only card they could instead of investing in other architectures. There is still gas on x86, note that, but you can see the bottom of the barrel.
Media may say that Apple M1 is the x86 first nail for its coffin, but it’s not. The first thing to die will be the PC ecosystem, the openness and modularity, because smartphones have proven people don’t give a crap about that. Once it does, x86 as an architecture will die from mainstream, which may bring Intel down if they don’t play their cards right — seeing how they’ve done it recently, that sure don’t know how to play the game anymore.
The future for x86 will become niche, not mainstream. The next fight may be ARM vs x86 vs RISC-V for who can offer more with less, in terms of licencing and performance per watt. The server market is still a thing, and the companies that pay the electricity bill will have an eye on those ecosystems that lower their costs as long that their software can keep up with the new hardware.
This seems like PowerPC vs x86 all over again.