Intel will start producing ARM chips to boost foundry business

Chip maker Intel and British semiconductor IP company ARM announced an agreement that could help boost the chip giant’s custom foundry business.

The deal, revealed today at the the Intel Developer Forum in San Francisco, will allow Intel Custom Foundry to make ARM processors for third parties.

Intel’s latest earnings announcement made it clear that the company is in the midst of a shift, and it needs to gear up for the Internet of Things. The agreement with ARM Holdings, which was acquired by Japanese tech giant Softbank a month ago, could be the first step in this direction.

“This will allow Intel to compete with Taiwan Semiconductor Manufacturing and Samsung foundries for the business of the likes of Qualcomm and Apple,” Abhinav Davuluri, equity analyst for Morningstar in Chicago, wrote in an online note.

As part of the collaboration, ARM and Intel Custom Foundry will accelerate the development and implementation of ARM SoCs on Intel’s 10-nanometer process. Specifically, Artisan Physical IP, ARM’s intellectual property, will be made available in the process.

“Optimizing this technology for Intel’s 10 nm process means that foundry customers can take advantage of the IP to achieve best-in-class PPA (power, performance, area) for power-efficient, high-performance implementations of their designs for mobile, IoT and other consumer applications,” Zane Ball, co-general manager of Intel Custom Foundry at Intel, wrote in a blog post.

Ball added that Intel has partnered with ANSYS, Cadence, Mentor Graphics and Synopsys previously. Intel also is making chips for customers such as Netronome and LG Electronics, which recently announced a new flagship device that will be the first handset to ship with Android 7.0.

Happy 10th birthday to Intel’s Core 2 Duo

Ten years ago July 27, Intel launched the Core 2 Duo. The company’s then-new 65nm Conroe core set new records for efficiency and high performance. It gave Intel a serious response to AMD’s Athlon 64 family, which had spent nearly three years making hash of the Pentium 4 — and it set the stage for Intel’s ultimate dominance of the x86 CPU market.

The Core 2 Duo wasn’t Intel’s first dual-core — that was Smithfield, the Prescott P4 chip that debuted the previous year — but it was the first CPU from Intel to show the promise of the dual-core approach that AMD had already debuted. It’s hard to remember now, but back then the cost difference between a single-core and dual-core system was enormous.

The roots of the Core 2 Duo go back much farther than 2006, though, to the P6 (Pentium Pro) architecture that Intel launched in 1995. Pentium Pro begat Pentium II, which begat Pentium III (Klamath, Coppermine, and Tualatin). After the P3, Intel made the decision to switch to the Netburst architecture and leapt from 2GHz CPUs in January 2002 to 3.2GHz (with Hyper-Threading) on June 23, 2003. That’s a 60% clock speed jump in less than two years, with the addition of simultaneous multi-threading on top of that. AMD, meanwhile, was still flogging the original Athlon core — and while K7 had competed extremely well against early iterations of the P4, Intel’s 130nm Northwood refresh had pumped a heck of a lot of gas into the Pentium 4’s proverbial tank.

But Intel didn’t just lean on its market position — it invested in a long-term engineering program that became the future of its CPU efforts. At every step, Intel focused on building a power-efficient CPU that evolved and became more powerful without sacrificing its intrinsically better power efficiency or execution capabilities. When it became clear that Prescott would never hit its performance or power targets, Intel was caught flat-footed in the short-term — the Pentium M wasn’t ready to take over on the desktop yet. By July 27, 2006, Conroe was — and benchmarks show just how potent the new core was.

Anandtech has a deep dive into the Core 2 Duo’s architecture and capabilities that I highly recommend reading if you want to brush up on how Intel evolved the P6 core and what the long-term ramifications were. Conroe didn’t just sweep the Pentium 4’s performance out of the market — it kicked off a chain of events that left AMD unable to compete against Intel in raw CPU performance.

In retrospect, Core 2 Duo was the herald of difficult things to come. The signs had been there for years — in late 2004 I wrote an article about using Dothan on a desktop and warned that if AMD didn’t pay attention to the performance Pentium M could put on the board it would leave K8 in a world of hurt. Even so, I didn’t expect C2D to be the beginning of a trend that would leave AMD flatly unable to compete with its rival for nearly half a decade.

Hopefully Zen will be the beginning of reversing that trend, but the graph from Anandtech above shows one other problematic fact — AMD is fighting its way back from the deepest competitive trough that’s ever existed between the two companies. The gap between the top-end Athlon 64 X2 and the top-end Core 2 Duo is much larger today than it ever was back in 2006.