ARM officials are continuing their push into the data center through an expanded partnership with chip foundry Taiwan Semiconductor Manufacturing Corp. to develop a 7-nanometer FinFET chip manufacturing process.
The two companies already are working together on 16nm and 10nm processes that have used ARM’s Artisan foundation physical IP. The development of a 7nm process would move the work the two companies have done together from mobile systems-on-a-chip (SoCs) to data centers and next-generation networks, according to officials with both ARM and Taiwan Semiconductor Manufacturing (TSMC).
“Existing ARM-based platforms have been shown to deliver an increase of up to 10x in compute density for specific data center workloads,” Pete Hutton, executive vice president and president of product groups at ARM, said in a statement. “Future ARM technology designed specifically for data centers and network infrastructure and optimized for TSMC 7nm FinFET will enable our mutual customers to scale the industry’s lowest-power architecture across all performance points.”
ARM designs processors and then licenses those designs to manufacturers like Qualcomm, Samsung, Nvidia, Apple and Applied Micro, which then put their own technologies on top of the architecture and build the chips for their own use or to sell to device makers. Most of the chips in smartphones and tablets are based on ARM designs.
The company, for several years, has been eyeing the data center, with officials saying their low-power designs are a good fit for many emerging workloads running on servers and over networks. ARM is looking to carve out space in a data center environment dominated by Intel, which owns more than 95 percent of the market.
A number of chip makers—including Applied Micro, Cavium and old Intel rival Advanced Micro Devices—offer 64-bit ARM-based processors. There’s also a lot of work going on building up the software ecosystem around the architecture, though adoption is not expected to ramp for at least another year or more.
All chip makers are working on shrinking their chips down to at least 7nm, leveraging the FinFET technology that essentially creates a 3D transistor design. FinFET helps chip makers increase the performance of their SoCs while keeping power consumption and leakage down. In July 2015, IBM unveiled a 7n test chip, the result of five years of development work by an alliance of IBM Research, chip foundry Globalfoundries and Samsung at SUNY Polytechnic Institute’s Colleges of Nanoscale Science and Engineering.
IBM officials said the chips will be used for the company’s Power servers and mainframes to run modern workloads, including big data analytics, cloud services and mobile applications.
TSMC officials said the Taiwanese foundry, through its 7nm FInFET process, is expanding beyond mobile and into high-performance compute. The process will “deliver more performance improvement at the same power or lower power at the same performance as compared to our 10nm FinFET process node,” Cliff Hou, vice president of R&D at TSMC, said in a statement.
The two companies have worked together to bring ARM’s Cortex-A72 to 16nm FinFET and 10nm.