Intel licenses ARM’s physical IP and signs LG as foundry customer
There was much excitement at the news that Intel had signed a licence for physical IP from arch-rival ARM, but this does not signal a return to designing ARM processors – something Intel abandoned when it sold its XScale business to Marvell in 2006. Instead, it is a sign that the company is serious about its foundry business, and about building on its acquisition of FPGA maker Altera.
The company has never been as religious about x86 as sometimes portrayed. It did buy the StrongARM platform, complete with ARM architectural licence, from Digital Equipment, making it the basis of XScale and one of its first failed attempts to get into smartphone processors. And more recently, it has acquired ARM-based chip designers on the network infrastructure side, such as Picochip and part of LSI.
Now it will license ARM’s physical IP in order to optimize its 10nm and 14nm processes for ARM-based chips – an essential move as CEO Brian Krzanich looks to expand the foundry business as a key new revenue stream, to offset the decline in PC chip sales.
Intel also announced that LG Electronics will be one of its first foundry customers, as it looks to take some of the contract chip manufacturing business dominated by TSMC, Samsung, Global Foundries and UMC, as well as emerging Chinese giants like SMIC. LGE has hired Intel Custom Foundry to make a mobile processor in its 10nm process. Existing Intel foundry customers, as well as its own Altera unit, are the smaller FPGA maker Achronix and Chinese partner Spreadtrum.
As most foundries move to 10nm and start to think about 7nm, Intel needs to build on its process lead urgently. And now its 10nm design platform will offer access to ARM Artisan physical IP, opening the doors to companies such as Apple – which has often been rumoured to be interested in working with Intel. Krzanich recently announced that his firm was, in effect, abandoning the attempt to be a major smartphone player (except in modems, where Apple recently signed up for chips from the former Infineon Wireless unit at Intel). Making smartphone chips for others could be a compensation for the failure of the mobile strategy.
Krzanich also told the Intel Developer Forum last week that he promised to accelerate Altera’s growth rate, which would involve using ARM cores. “More than 50% of the value of the acquisition was actually tied to growing Altera at a faster rate than it was growing,” he said. Altera’s flagship Stratix 10 FPGAs will ship by year end, he pledged, with Intel branding and made using Intel’s 14nm process.
“There is no plan to yank ARM out [of FPGAs] and stick in Intel architecture and make you all change your programming models,” he told FPGA developers at IDF. “There are some advanced products that may like an IA core but my guess is the majority of them won’t. That would be wasted money on our part and trouble for our customers. The default processor will be ARM.”
He added that Intel was already shipping a co-packaged product with Xeon alongside FPGAs, for high end applications such as carrier networks, but he saw this option as being one for the upper end of the data center market only.
As for ARM, the general manager of its physical design group, Will Abbey, said in a statement: “Today’s announcement represents what we expect to be a long term, mutually beneficial partnership with Intel Custom Foundry. One of the strengths and differentiators of the Artisan platform is the availability of ARM core-optimized IP – what we call ARM PO technology. 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.”