Intel’s Xeon server chips dominate hardware in data centers, and now they could also end up powering robots on factory floors.
The new line of Xeon D chips, announced Monday, are designed primarily for servers and network appliances, but as industrial automation grows, Intel believes the chips can all add processing muscle to robots that handle complex manufacturing tasks.
Simple robots that do mundane work can run on basic, low-power processors, but faster chips are being plugged into advanced robots for more sophisticated tasks.
Xeon D is the first server chip from Intel based on the Broadwell architecture. It’s already being used in PC chips, but it’s graduating to servers, appliances, and now perhaps robots.
The chip has features that could benefit robots, such as on-chip security to protect them from hackers. It also has high-reliability features that are increasingly in demand where failure isn’t an option, said Dean McCarron, principal analyst at Mercury Research.
An example is optical inspection, used for quality control in manufacturing and packaging. Objects are scanned and compared against a database, and flagged if a unit looks defective or doesn’t match patterns.
“Throwing more compute power gives you more options,” McCarron said. “With the compute power comes fancier algorithms and more advanced ways of doing the job.”
Xeon D will let robots to execute Internet of Things tasks by letting them connect to cloud services for information and giving them stronger on-board processing to parse and analyze data.
Intel said Xeon D can be customized and made available in power efficient designs, which means special chips can be developed for robots. But the first chips in the lineup, the quad-core 1520 and eight-core 1540, are targeted at microservers and networking appliances.
Intel already sells low-power chips called Xeon E3 and Atom chips code-named Avoton. But Xeon D will provide more horsepower while being relatively conservative on power consumption. The dense servers and appliances will be used for web hosting and cloud services, and for processing data generated by the growing number of mobile and sensor devices, said Raejeanne Skillern, general manager for cloud services at Intel’s Data Center Group.
An estimated 50 billion devices will be transmitting information by 2020, and the emergence of the Internet of Things highlights the need of power-efficient servers and edge appliances that can deliver information and scale performance quickly, Skillern said.
Xeon D is effectively a system-on-chip, with a combination of components including I/O and controllers for networking and storage appliances. Intel was due to ship the Xeon D chips by late 2014, but had to shift dates following manufacturing glitches affecting Broadwell processors.
The Xeon D will draw a minimum of 20 watts—higher than Avoton’s 15 watts. An eight-core Xeon D chip was 3.4 times faster in Web serving than an eight-core Avoton, 2.5 times faster in storage, and around 3.1 times faster in networking, according to Intel’s internal benchmarks. The tests were carried in Hewlett-Packard’s Moonshot and Supermicro’s Superserver dense servers.
Per watt of power, users will get 70 percent more performance with Xeon D compared to Avoton, Intel claims.
There may be some overlap between Atom workloads and Xeon D workloads, but Intel wants to provide a wider range of options for performance, density and cost, said Lisa Spelman, director of datacenter product marketing at Intel’s Data Center Group.
The Xeon D improvements will support 128GB of DDR3 and DDR4 memory, 10GB ethernet and PCI-Express 3.0.
Intel has also said it offers reprogrammable FPGA circuits with Xeon D chips, but the company will share more details about those options in the second half of the year.