Broadcom’s new Sian3 and Sian2M physical layer digital signal processor (DSP PHY) are aimed at the connectivity needs of artificial intelligence (AI) and machine learning clusters.
The new devices expand the company’s line of 200G/lane DSPs. The Sian 3 is a 3nm DSP PHY with low power consumption for 800G and 1.6T optical transceivers over single mode fiber (SMF). The Sian 2M features vertical-cavity surface-emitting laser (VCSEL) drivers and is optimized for low-power short-reach multimode fiber links in AI clusters, according to Broadcom.
“Broadcom’s Sian family of DSP PHYs is foundational to the low-power, high-bandwidth optical connectivity needed for AI/ML clusters,” Vijay Janapaty, vice president and general manager of Broadcom’s physical layer products division, said in the company’s announcement.
“Our new 3nm Sian3 delivers over 20% power reduction for 1.6T optical modules, while Sian2M with integrated VCSEL drivers and 200G VCSELs brings cost and power efficiency to short reach links. These innovations enable our customers to scale AI clusters to meet the demands of growing AI workloads.”
This is not Broadcom’s first announcement directly related to AI. Last September, the company introduced the Sian2. At the time, Broadcom said the DSP was aimed at the demands of emerging AI technology.
The initial Sian2 offering provided 200 Gbps electrical and optimized per lane performance. It is aimed at 800G/1.6T links AI clusters. The company says that goal was to reach this capacity while maintaining low latency, power use and cost per bit.
The telecommunications industry is gearing up for the growth of AI in general and generative AI in particular. In February, Google Cloud said that AI shows the most promise in data migration, proactive network operations, empowering field technicians and creation of new revenue streams through APIs.
