Intel turns 50 today, marking a major landmark for the processor company that has, perhaps more than any other company, become synonymous with the chips that are the beating hearts of the electronic devices we use today. Anyone familiar with Intel should be aware of Moore’s Law, which served as the underlying philosophy that has driven the chip company for the last half-century.
Written by Intel co-founder Gordon Moore, Moore’s Law is commonly quoted as saying that every two years, the number of transistors that could fit on a microchip would double. But in Moore’s original paper that was published in Electronics magazine, his prediction was actually that transistor count would double approximately every year for the 10-year period from 1965 to 1975. The two-year time frame came from a revised estimate published by Moore in 1985.
Still, whether or not it was intentional, Moore’s Law and the looming goalposts it set have been a core part of Intel’s identity. The entire pace of progress for the company — and, in essence, the rest of the computing world — has been driven by it. In recent years, Intel operated in a tick-tock process that sees a smaller architecture size (that ups transistor count accordingly) released one year, an improved version of chips built on that architecture the following year, and then things start all over again the year after that.
Unfortunately, Moore’s Law is starting to fail: transistors have become so small (Intel is currently working on readying its 10nm architecture, which is an atomically small size) that simple physics began to block the process. We can only make things so minuscule. While things aren’t completely hopeless, further efforts to get transistor sizes down (and, consequently, transistor count up) are going to come slower and slower with higher costs as we move forward.
The pace is already beginning to decline. In 2015, then-CEO Brian Krzanich said, “The last two technology transitions have signaled that our cadence is closer to two and a half years than two years.” Intel’s new 10nm architecture has continued to hit delays and is now expected to release in 2019, marking over a three-year cadence since the previous architecture node.
Intel has released a revised generation of 14nm+ chips (Kaby Lake R) and an extra generation of 14nm++ Coffee Lake products that are both based on the previous architecture in the meantime. They are instead focusing on things like improved battery life and boosting the numbers of cores offered rather than a raw transistor count upgrade.
While there are a variety of solutions to this — including innovating new types of transistors, using new materials, or even rethinking how our computers work on a fundamental level to squeeze out more time (I’d recommend this excellent piece from The Economist detailing some of those options) — ultimately, we’re going to hit a wall. Like it or not, change is coming to Intel. The company has already missed a crucial shift in computing with the rise of smartphones, and it faces a mobile landscape where Qualcomm’s Snapdragon chips reign supreme, with nary an Intel smartphone to be found. There’s also the looming ghost of Spectre and Meltdown that could continue to haunt the company for years to come. All the while, upstarts like AMD and Qualcomm are elbowing back in on Intel’s turn with Ryzen processors and ARM-powered Snapdragon desktops. There are also rumors that even Apple may be looking to design its own chips for computers like it already does for its iOS devices. Lastly, Intel now has to replace its CEO after Brian Krzanich resigned after violating company policy over a relationship with an employee.
If you were feeling prone to the dramatic, one could make the argument that Moore’s Law is failing Intel just when the company needs it the most.
This isn’t necessarily a bad thing. The slower pace is already giving companies like Intel more time to refine and optimize their existing architectures, and companies like the newly resurgent AMD are getting more time to catch up and compete. Ultimately, this benefits everyone.
But it does mean that the future of Intel will have to change and evolve, instead of just relying on a continuous cycle of iteration. We’re already seeing this with products like Intel’s eighth-generation lineup of chips, which have seen quad-core and hexa-core processors become the standard across parts of both desktop and laptop lineups: doing more with the existing technology and transistor count we have with parallel multicore processing, instead of squeezing more raw power in.
Intel’s unprecedented partnership with AMD earlier this year saw a hybrid CPU/GPU chip that combined Intel’s Core line of processors with AMD’s Radeon graphics. Once again, Intel is leveraging existing processor technology in a new way to offer better performance and battery life in laptops that can now be smaller than ever, without fighting the warring demands of Moore’s Law and molecular physics.
It may not be today, or even within the next 10 years, but sooner or later, Moore’s Law is going to fail. What comes next — and what Intel does in the next 50 years without it — could change everything about computing again.