Highly parallel optical links with power efficiency of 0.1pJ/bit, bandwidth density of 10Tbps/mm2 and reach of up to 10m promise to smash current interconnect bottlenecks in distributed compute systems
Avicena Inc., a privately held company in Mountain View, CA, unveils LightBundleTM, a highly parallel optical interconnect technology targeting up to 10 meters reach for chip-to-chip interconnects in distributed computing, processor-to-memory disaggregation, and other advanced computing applications. LightBundleTM is based on arrays of novel GaN high-speed micro-emitters, leveraging the microLED display manufacturing ecosystem, and is fully compatible with high performance silicon ICs.
Recommended ITech News: cnvrg.io AI Operating System Teams Up with Supermicro to Deliver End-to-End AI Experience
Reinventing Chip-to-Chip interconnects; Avicena’s LightBundle™ interconnects achieve 10x – 100x power and density improvements over any alternative for 1 cm – 10 meter reach. Check out our new press release!
Interconnects are becoming the key bottleneck in compute and network systems. Highly variable workloads are driving the evolution of densely interconnected, heterogeneous, software-defined clusters of CPUs, Graphical Processing Units (GPUs), Data Processing Units (DPUs) and shared memory blocks. Exploding Artificial Intelligence (AI) and Machine Learning (ML) workloads are exemplary of emerging applications driving an accelerating need for interconnects with extremely high density, low power consumption and low latency.
Recommended ITech News: Marvell Extends PAM4 DSP Cloud Data Center Leadership with Industry’s First Integrated Solution
