Apple’s Federighi on M1 and why chip specs aren’t always indicative of real-world performance

According to Apple’s software engineering head Craig Federighi, technical specifications are no longer a good predictor of actual real-world performances that depend on the task at hand.

The comment was made in an interview with Om Malik, who sat down with Federighi, marketing boss Greg Joswiak and technologies head Johny Srouji to discuss the company’s new M1 chip that has upended the industry with its fast speed and low power consumption.

Here’s the full quote:

The specs that are typically bandied about in the industry have stopped being a good predictor of actual task-level performance for a long time. Architecturally, how many streams of 4K or 8K video can you process simultaneously while performing certain effects? That is the question video professionals want an answer to. No spec on the chip is going to answer that question for them.

Srouji added:

We are a product company and we built a beautiful product that has the tight integration of software and silicon.It’s not about the gigahertz and megahertz, but about what the customers are getting out of it.

Federighi on the M1’s high degree of integration and specialized execution engines:

It is difficult to put more transistors on a piece of silicon. It starts to be more important to integrate more of those components closely together and to build purpose-built silicon to solve the specific problems for a system. Being in a position for us to define together the right chip to build the computer we want to build and then build that exact chip at scale is a profound thing.

The M1 is a laptop-optimized chip powering the new MacBook Air, 13.3-inch MacBook Pro and Mac mini. Apple will release additional chips to power its higher-end computers, like the 16-inch MacBook Pro, along with the iMac all-in-one desktop and the iMac Pro workstation.

Federighi said:

It seems like some of these people were people who don’t buy that part of our product line right now are eager for us to develop silicon to address the part of the product line that they’re most passionate about. You know that their day will come. But for now, the systems we’re building are, in every way I can consider, superior to the ones they’ve replaced.

Srouji on Apple’s unique business model:

We’re developing a custom silicon that is perfectly fit for the product and how the software will use it. When we design our chips, which are like three or four years ahead of time, Craig and I are sitting in the same room defining what we want to deliver, and then we work hand in hand. You cannot do this as an Intel or AMD or anyone else.

The shift to the M-series chips will take as long as two years, Apple has said.

According to a back-of-an-envelope calculation by IBM’s AI strategy lead Sumit Gupta, the M1 chip costs Apple an estimated $40-50 versus about $200 for Intel’s Core i5 processor in the MacBook Air (TSMC is Apple’s manufacturing partner for chips). Overall, using its own in-house designed silicon should save Apple $2.5 billion this year alone, Gupta wrote in a Medium post.