Fabless chip company Integrated Device Technology Inc. announced Wednesday (Feb. 25) that it has collaborated with Nvidia Corp. and Orange Silicon Valley to develop a distributed computing platform that can analyze 4G and 5G data in real-time.
The platform links clusters of Tegra K1 mobile graphics processors with IDT chips for RapidIO interconnect, and can be housed in network operators basestations, IDT (San Jose, Calif.) said. RapidIO is a high-performance, packet-switched interconnect standard and is overseen by the RapidIO Trade Association.
Nvidia's processors are intended not to render graphics but to perform general-purpose computing, deep learning algorithms and analytics. The idea is that by installing the hardware at multiple, geographically dispersed basestations along the edge of the network, analysis of locally generated data can be delivered more quickly. It also gives network operators the opportunity to generate revenue from the data that is flowing through their network.
IDT gives the example of social media content analysis. "For example, if you tweet that you’ve just seen a movie and are headed out to dinner, some nearby dining options could pop up on your screen," said Sailesh Chittipeddi, IDT’s vice president and chief technology officer, in a statement. "Or the technology can be used for mass transit; if you’re standing at a bus stop, you can check your phone for the precise location of your bus."
The distributed computing platform is described by IDT as "supercomputing at the edge" and is designed to serve high-performance computing and support IoT appliances and wireless access networks handling 4G and higher traffic.
The platform is based on server computers from ProDrive Technologies (Eindhoven, The Netherlands) that contain computing cards from Concurrent Technologies plc (Colchester, England). Each computing card contains four Tegra K1 processors connected with a RapidIO network interface controller (NIC) circuit from IDT capable of 20Gbps. The platform can provide up to 12 teraflops of performance per RapidIO server blade, IDT said.
Because the technology is deployed and analysis conducted at local basestations—at the edge of the wireless network, rather than in a central location—it removes a bottleneck between the basestation and the core of the network. "The solution was developed with an architecture designed to handle the emerging market for geographically distributed analytics, deep learning and pattern recognition in real time," Chittipeddi said.
Questions or comments on this story? Contact firstname.lastname@example.org.