When Intel announced that it was canceling its tablet and smartphone products and effectively exiting the Android-x86 market, it was generally seen as a sign that the company had given up on attempting to compete against low-power ARM products manufactured by Samsung, TSMC, and GlobalFoundries. Now Intel announced that it would begin building 10nm hardware for companies that want to license standard ARM cores.
ARM and Intel released joint blog posts with additional information. According to ARM, this deal will cover future ARM products, dubbed ARM POP technology (POP is not an acronym). ARM writes:
The value of POP technology for an ARM core on the Intel 10nm process is tremendous, as it will allow for quicker knowledge transfer, enabling customers to lower their risk in implementing the most advanced ARM cores on Intel’s leading-edge process technology. Additionally, POP technology enables silicon partners to accelerate the implementation and tape-outs of their ARM-based designs. The initial POP IP will be for two future advanced ARM Cortex-A processor cores designed for mobile computing applications in either ARM big.LITTLE™ or stand-alone configurations.
Meanwhile, Intel’s post notes that it has been collaborating with companies like ANSYS, Cadence, Mentor Graphics, and Synopsys to build advanced IP tools for customers, as well as working with ARM to bring its products over to Intel’s 10nm node. Intel writes:
Our 10 nm design platform for foundry customers will now offer access to ARM® Artisan® physical IP, including POP™ IP, based on the most advanced ARM cores and Cortex series processors. 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.
The ARM Artisan platform includes:
High Performance and High Density Logic Libraries
POP IP (for future ARM premium mobile cores)
Having leading IP providers in our portfolio will accelerate ecosystem readiness while providing greater flexibility and time-to-market advantages to our customers.
LG is the only customer currently confirmed for Intel’s 10nm node, and this deal seems limited to companies that want to build standard ARM cores as opposed to designing custom architectures to be built on Intel’s fabs, like Qualcomm or Apple.
Intel’s decision to start fabbing hardware for ARM customers isn’t as surprising as it would’ve been this time last year, before it killed its tablet and smartphone Atom cores. Building hardware for the companies you intend to compete against is an obvious conflict of interest that wouldn’t have sat well with Chipzilla or its ARM customers. Now that Intel is out of the smartphone and tablet business, building chips for other companies is a logical step to take. Intel still retains a license to design its own ARM cores should it ever decide to do so, and it’ll be interesting to see what happens as a result of its entry into the ARM business. In theory, adopting Intel’s 10nm technology could give LG a leg up on its competitors.
It’s worth noting, however, that Intel’s previous foundry partnerships don’t seem to have resulted in any high-profile success stories. Altera signed an agreement to build FPGAs with Intel before Intel acquired them outright, but it’s not clear if doing so bought Altera any significant capabilities compared to its primary competitor, Xilinx. Spreadtrum, Intel’s other 14nm partner, is a low-end Chinese fabless semiconductor company. LG is the highest-profile win that Intel has secured, though the company’s smartphone sales have not been good in recent years.
One rumor we’ve heard in the past is that while Intel’s foundry technology is second to none, it’s also tuned and designed specifically for Intel’s own design rules. Presumably the Custom Foundry business has been either moving away from this approach — giving Intel more generic capacity that would allow it to compete effectively against the likes of TSMC, GlobalFoundries and Samsung — or has found a way to make it easier for companies to come in and build chips on its own technology. Pricing constraints and fab costs were a major problem for Intel when competing against ARM with its own Atom processors, and it’ll have to hit some aggressive price targets to build hardware for third parties that can compete effectively.
Exact details on all these points is obviously confidential, but it’ll be fascinating to see if LG’s future chips are measurably more power efficient or faster than their 10nm counterparts built at other foundries. This kind of comparison could give us the first apples-to-apples comparison of Intel’s foundry prowess against its competitors that we’ve ever seen.