Marvell Technology, Inc. showcased its 3nm PCIe Gen 7 connectivity at the 2024 OCP Global Summit in San Jose, highlighting its progress in scaling compute fabrics within AI server platforms and disaggregated infrastructure. PCIe Gen 7 doubles data transfer rates compared to its predecessor, PCIe Gen 6, driving better performance in AI and machine learning workloads. This new standard enables faster communication between CPUs, GPUs, and accelerators, addressing the exponential growth in data processing demands, especially for AI models that double in complexity every six months.
Leveraging its expertise in PAM4 modulation, Marvell’s interconnect technology now extends beyond traditional Ethernet and InfiniBand connections into PCIe and CXL systems. PAM4, a crucial component of PCIe Gen 7 and 6, allows for higher bandwidth efficiency, which is essential for today’s AI-driven infrastructure. With the introduction of PCIe Gen 7, Marvell aims to optimize power, speed, and latency in the increasingly dense AI data centers.
PCIe Gen 7 operates at 128 GT/s per lane, significantly improving data center efficiency by lowering power consumption and latency. Its use of 3nm fabrication technology ensures reduced energy usage while supporting the large-scale infrastructure required for next-generation AI clusters and high-performance computing systems.
Key points:
• PCIe Gen 7 doubles transfer speed compared to PCIe Gen 6.
• Operates at 128 GT/s per lane.
• Designed for AI, ML, and HPC workloads in cloud data centers.
• PAM4 modulation supports both optical and copper connections.
• Uses 3nm fabrication technology for improved efficiency.
“AI workloads are driving the evolution of server interconnects, and our PCIe Gen 7 technology is engineered to meet the performance and scalability needs of next-generation AI data centers,” said Venu Balasubramonian, vice president of product marketing at Marvell.
