Intel Corp. (Nasdaq: INTC)'s new form factor reference designs (FFRD) for cable modems and gateways are based on the company's new Puma 7 chip, one of the first Intel has fabricated at the new 14 nm node.
Being rendered at a brand new node is an honor Intel typically reserves for its mainstay microprocessors (and indeed, it is working on 14 nm processors). Simply moving from one node to the next brings with it a set of well-known improvements that include smaller size, faster operation, and lower power requirements.
The Puma 7, to pick an example almost entirely at random, is five times faster than its predecessor, with the same thermal profile.
Intel's FFRDs for cable CPE are by no means a sideshow, but the iPhone market is even bigger. Let's review some numbers for perspective. The global market for all broadband CPE (cable, DSL, FTTH, mobile CPE) is between $11 billion and $12 billion a year (IHS/Infonetics; from Q2 2015). The global market for Apple iPhones -- not all smartphones, just the iPhone -- is worth roughly $31.4 billion a quarter (Statista; from Q2 2015).
We know that Intel dominates the market for chips going into personal computers. We know the PC market has been slowing down. We know that Intel would like to get much deeper into the smartphone business.
We know that Apple Inc. (Nasdaq: AAPL)'s latest iPhone 6s didn't knock people's socks off. We know that Apple dominates not just because it is hipper and cooler, but because it pushes smartphone technology in a way the Android community does not (and perhaps cannot). Apple needs a knock-yer-sox-off iPhone next year.
What can Apple do to imbue its next iPhones with some step-function improvements? How about build them around chips fabbed at the next node, at 14 nm? A node that maybe four semiconductor manufacturers in the world will be capable of anytime during the next iPhone development cycle -- or two, or maybe even three? Apple currently relies on Qualcomm Inc. (Nasdaq: QCOM), which had been outsourcing its production to TSMC, but Qualcomm is reportedly switching to Samsung to gain access to the latter's 14 nm processing capabilities (originally reported by re/code).
All of this lends credence to Venture Beat's report that Intel has an enormous team working on a project aimed at getting Intel LTE modems designed in to the next iPhone model, presumably the iPhone 7, sometime in 2016.
Venture Beat and others who have picked up the report say Intel is trying to entice Apple with some customized version of its current 7360 LTE modem. Why would anyone need 1,000 people working on an existing product? What might be more understandable is 1,000 people working on an LTE modem implemented in the same insanely difficult 14 nm process used for the Puma 7 cable modem.
Again, the Puma 7 cable modem is no side show. I first sighted Intel's demonstration model in September at a CableLabs interop event, where Intel engineers passed it around to cable operators who seemed eager to examine the new device (see DOCSIS 3.1 Is Right on Schedule). We got a closer look (and the accompanying photo) last week at Cable Tec Expo.
While Intel has implemented a cable modem, it is a demonstration model only; the company has no intention of making CPE. Instead, it is supplying OEMs with several FFRD options: a simple cable modem, an eMTA (with two voice ports), and a "kitchen sink" model that includes a WiFi router and four Ethernet ports, said Jim Crammond, a Ph.D who works in Intel's Connected Home Division, in an interview with Light Reading at Cable Tec Expo.
The modem-only model that Intel had on display measures roughly 4.5 x 4.5 x 1 inch. The form factor is out of favor in US markets, but is employed elsewhere, including China and Australia, among other markets, Crammond noted.
Intel has tested the Puma 7 reference designs for DOCSIS 3.0 performance, and are "solid" on D3.0, Crammond asserted. The chip is currently in CableLabs' DOCSIS 3.1 certification process.
It is sampling, and will ramp into production next year.
The hurdles associated with the 14 nm node are incredibly high, including new lithography processes, new resist technology, approaching the physical limits of optical metrology, and increasingly rigorous test regimes that are becoming thoroughly integrated into the entire design and fabrication process. Some wonder if the semiconductor industry can keep pushing silicon processes much further at anywhere near the same pace the industry has been keeping for decades.
Intel is having none of it. "We wanted to let people know that Moore’s Law is not dead," Crammond said.
— Brian Santo, Senior Editor, Components, T&M, Light Reading