In what is said to be an industry-first, the silicon building blocks in this node include 112G XSR SerDes (serializer/de-serializer), Long Reach SerDes, PCIe Gen 6 / CXL 3.0 SerDes, and a 240 Tbps parallel die-to-die interconnect.
Part of Marvell’s strategy to develop a comprehensive silicon IP portfolio for designing chips that will radically increase the bandwidth, performance, and energy efficiency of next-generation data infrastructure, these new ‘building blocks’ also support all semiconductor packaging options from standard and low-cost RDL (Redistribution Layers) to silicon-based high-density interconnect.
SerDes and parallel interconnects serve as high-speed pathways for exchanging data between chips or silicon components inside chiplets. Together with 2.5D and 3D packaging, these technologies are seen has helping to eliminate system-level bottlenecks to advance the most complex semiconductor designs. SerDes also help reduce pins, traces and circuit board space to reduce cost. A rack in a hyperscale data centre might contain tens of thousands of SerDes links.
The parallel die-to-die interconnect, for example, enables aggregate data transfers up to 240 Tbps, 45% faster than available alternatives for multichip packaging applications, and, to put that in perspective, the interconnect transfer rate is equivalent to downloading 10,000 HD movies every second, although over a distance of only a few millimetres or less.
Marvell has incorporated its SerDes and interconnect technologies into its flagship silicon solutions including Teralynx switches, PAM4 and coherent DSPs, Alaska Ethernet physical layer (PHY) devices, OCTEON processors, Bravera storage controllers, Brightlane automotive Ethernet chipsets, and custom ASICs.
Moving to a 3nm process enables engineers to lower the cost and power consumption of chips and computing systems while maintaining signal integrity and performance.
“Interconnects are taking on heightened importance as clouds and other computing systems grow in size, complexity and capability. Our advanced SerDes and parallel interfaces will play a significant role in providing a platform for developing chips with best-in-class bandwidth, latency, bit error rate, and power efficiency for meeting the demands of AI and other complex workloads,” said Raghib Hussain, president of products and technologies at Marvell. “We are proud to be able to deliver such advances on TSMC's 3nm technology and take semiconductor designs to the next level for our customers around the world.”
“Bandwidth is the lifeblood of the cloud. Service providers are growing their network capacity by approximately 50% per year in the cloud and by over 100% for AI applications,” said Alan Weckel, co-founder of 650 Group. “Marvell’s successful production of 3nm SerDes and interconnects marks the latest step in helping cloud service providers to stay ahead of the ever-escalating demand for higher speeds and more traffic.”
