Johny Srouji and the Apple A7 Chip

Read Silicon is “Unforgiving,” Says Apple’s Chip Chief Johny Srouji (CTECH - www.calcalistech.com)
In an exclusive interview with Calcalist, Apple’s senior vice president of hardware technologies discussed Apple’s ascent up the supply chain ladder, and the fine balance between what the company plans to buy from vendors and what it plans to develop in-house in the future
Former Sarnoff Corporation coworker, engineer and local Kington resident Jeremy sent me this recent note along with a link to an article on Johny Srouji.

Johny Srouji is SVP Hardware Technologies at Apple (AAPL), heading up the development and design of Apple’s custom A-series systems on chip (SOCs). It was Srouji who led the design of Apple’s first A4 SOC in the first iPad. Apple may have become one of the most important semiconductor companies on the planet, and Srouji has had an awful lot to do with that.

Srouji was born in the beautiful Israeli port city of Haifa and educated nearby at the Technion, the Israel Institute of Technology. He recently gave an interview with the tech site Calcalist, providing useful insights into the chip design process at Apple. He also had some interesting tidbits of information regarding the involvement of PrimeSense, the Israeli company Apple acquired in 2013, in the TrueDepth sensor design.

Johny Srouji on the long gestation of Apple chips:

Silicon is unforgiving. My team is already working on the chips you’re going to see in 2020. You make bets. We have the system and the software. We have better knowledge versus external chipmakers about where things are going to end up. Since we own the silicon, we own the software, the operating system and everything else, we deliver, always. We deliver the exact specification of iOS and nothing else. We don’t have to worry about other operating systems.

I’ve often tried to capture the importance and value of controlling both the silicon design and the operating system design, and Srouji sums it up beautifully. I’ve also argued that development lead times are very long, but I haven’t heard anyone from Apple confirm this directly before.

The implications for the current A11 Bionic are interesting. Apple began planning the chip roughly three years ago. Fairly early in the planning stage, Apple’s hardware and OS teams had to have realized that they wanted to implement machine learning and inference algorithms locally on the mobile device, rather than in the cloud.

In 2014 or 2015, no one was doing that, or even talking about doing it. Conventional wisdom was that it had to be done in the cloud. Those hardware and OS teams had to have realized that special purpose hardware was needed to make that work, which would need to be provisioned early in the SOC design process. This became the Apple Neural Engine inside the A11.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Feel free to comment on this story directly above, but you can also go to copies posted to social media on the left, and reply to or comment on them there. Your responses via Twitter, Facebook, and Google+ will appear below.

To respond on your own website, enter the URL of your response which should contain a link to this post's permalink URL. Your response will then appear (possibly after moderation) on this page. Want to update or remove your response? Update or delete your post and re-enter your post's URL again. (Learn More)