This isn’t a cataclysmic end for the chip architecture that has carried us from the early days of personal computing in business to client-server, mobile computing, datacenter and cloud. See also: Everything Apple just announced and how to buy it all. Rather, x86 has now been assigned its role as an elder statesman. Any Foundation fans out there? Allow me to describe this in terms of Apple’s epic sci-fi series running on its TV Plus service – of which I am a huge fan. In Foundation, the 12,414-year-old Galactic Empire is ruled by brothers Dawn, Day, and Dusk. Dawn is the youngest. Intel x86 has been Emperor Day for a long time, but he is no longer in his prime. Yet he’s not quite ready to be “Emperor Dusk” – relegated to doddering unimportance or prepared to be “ascended” in a disintegration chamber on the final day of his life. However, Brother Day is no longer identified as the most vigorous leader of the Empire or seen as the future of our industry. Emperor Dawn is the Arm architecture. Until now, he was good enough for the little tasks – first, for embedded microcontrollers. Then for IoT. Then smartphones and tablets. But he is still not strong enough to take the reign from his big brother, x86, which remains the workhorse of business desktop computing, demanding creative computing workloads and server-based computing in datacenters and hyperscale clouds. Last year at this time, we saw the Apple Silicon processor move on from iPhone and iPad to its use on the MacBook Pro, the Mac Mini, and in early 2021, on the iMac, using the M1 chip. This made it good enough for entry and mid-level business desktop computing tasks, as it had a maximum RAM footprint of 16GB. At that moment, Emperor Day x86 looked at his little upstart brother biting at his heels and laughed nervously. On Monday, October 17, 2021, Emperor Dawn flexed his muscles, and he put Emperor Day x86 on notice. The new Macbook Pro systems, outfitted with M1 Pro and M1 Pro Max, can achieve – if Apple’s claims are to be believed, and benchmarks will soon confirm – Intel mobile PC (high-end laptop) workstation-levels of performance with about 100W less power consumption, if we are including equivalent GPU performance as well. See also: The new MacBook Pro highlights what’s gone wrong with Windows laptops. The M1 Pro, with up to 10 CPU and 16 GPU cores, has a memory bandwidth of approximately 200GB per second, whereas M1 Max, with up to 10 CPU and 32 GPU cores, can hit up 400GB per second, with a shared RAM pool using its unified memory architecture. That’s extremely impressive. M1 Pro and M1 Max can be outfitted with as much as 64GB of RAM, which is more than good enough for capturing a large percentage of moderately demanding creative content workloads. Indeed, there are still areas where x86 can outperform it, where much larger amounts of memory (in the terabytes) and many more CPU cores/threads are needed, such as on the most powerful Intel Xeons. Now, this doesn’t mean that M1 Pro or the M1 Max suddenly will displace the PC in business computing. These are expensive systems, in base configurations starting at $1,999 for the 14-inch model, and we are a far cry from replacing the $1,000 office PC or the $500 entry-level business laptop. Large enterprises aren’t moving to Macs anytime soon. But let’s not kid ourselves – Apple is not the only company working on Arm-based systems. Microsoft has the Surface Pro X, a very capable Windows 11 business laptop that costs $999. Is it as fast as M1 Pro or even last year’s M1? No. But Qualcomm’s SQ2 SoC inside is very power efficient and runs Windows business applications nicely. And it doesn’t cost a fortune. Qualcomm will continue to iterate the SQ series with Microsoft, and they will gradually get faster and faster. NVIDIA is also developing desktop-class chips, and with its pending acquisition of Arm, it will also be driving the future of the architecture. In September, AMD stated that it too is ready to go down the Arm path. And Intel has stated that as part of its expanding business strategy, it will manufacture Arm processors for other companies – as well as open source, license-free RISC-V chips, which are well on their way to being the future Emperor Dawn. RISC-V chips are already used in some applications that Arm used to be used heavily in, such as support chips on disk drive controllers, integrated within Arm processors for less computationally heavy tasks, and other IoT-type things. How long will Emperor Day x86 continue to lead? That’s hard to say. The cloud is still heavily reliant on his power. Still, over the next five years, we could very well see some cloud-native applications migrate to Arm as chips optimized for those environments begin to prove themselves, and hyperscale providers such as AWS and Azure see them as more power efficient and capable of running enterprise and SaaS workloads. Emperor Day x86 has many years before he is no longer of use to us and becomes “Brother Darkness.” But the roadmap to his ultimate demise is now visible – and he may be less than a decade from becoming Emperor Dusk, where Emperor Dawn – aka Arm64 – takes its rightful place as the new Emperor Day.
