CPO 2.0 arrives for AI / ML with high density optical interconnect
At this week’s OFC 2022, Ranovus demonstrates what it claims is the industry’s first adaptive compute acceleration co-packaged optics platform for artificial intelligence / machine learning (AI / ML).
The Ranovus Odin Analog-Drive CPO 2.0 eliminates the need for the retimer which results in a 75 per cent smaller footprint and 40 per cent cost and power consumption savings in the optical interconnect, claims the company.
Ranovus demonstrated the co-packaged optics using a Xilinx Versal ACAP (Adaptive Compute Acceleration Platform) and Ranovus Odin 800Gbits per second CPO 2.0. The joint demonstration highlighted the arrival of the CPO 2.0 or AI/ML platforms that demand power efficient, high throughput and high density optical interconnect.
Ranovus’ Odin is a low latency, high density, and protocol-agnostic optical engine that delivers “massive optical interconnect bandwidth”, scaling 800Gbits per second to 3.2Tbits per second in the same footprint by leveraging Ranovus’ 100Gbits per second per lambda monolithic electro-photonic IC (EPIC) IP, laser platform and advanced packaging technologies.
The Odin Analog-Drive CPO.20 for Ethernet switch and module applications was announced and the company offers multi-disciplinary IP cores to accelerate the adoption of Analog-Drive CPO in data centres.
CPO provides Nx100Gbits per second PAM4 optical I/O channels for Ethernet switch and ML / AI silicon in a single packaged assembly. This “significantly reduces” the cost and power consumption of the complete system, explained Ranovus.
“We’re proud of our collaboration with Ranovus that helped achieve record performance levels while at the same time reducing power and overall footprint of the complete solution,” said Dan Mansur, vice president, Adaptable and Embedded Computing Group, AMD (formerly Xilinx). By integrating the optics into the package using Versal Premium ACAPs, AMD and Ranovus are able to reduce power, simplify board routing, and reduce cost.
“This CPO demonstration highlights the versatility of the Versal GTM SerDes to operate over anything from long-reach copper to directly driving the Ranovus Analog-Drive CPO 2.0 optical engine. Co-packaging Ranovus Odin with Xilinx Versal is a significant advancement which enables data centre customers to build highly efficient and cost-effective systems for next-generation workloads,” added Mansur/
Visit the AMD booth (5711) or the Ranovus booth (5917) at OFC 2022 in San Diego (6 – 10 March).
