Lenovo ThinkSystem SR780a V3 GPU Server — Product Overview & Specifications

Product Overview

The Lenovo ThinkSystem SR780a V3 is a 5U rack-mounted GPU server purpose-built for demanding Artificial Intelligence (AI) and High-Performance Computing (HPC) workloads. It combines dual 5th Gen Intel® Xeon® Scalable processors with eight fully interconnected NVIDIA Tensor Core GPUs via high-speed NVLink, delivering massive computational throughput for large language model (LLM) training, generative AI, machine learning, simulation, rendering, and scientific research.

What distinguishes the SR780a V3 is its Lenovo Neptune™ hybrid liquid-cooling architecture. Direct liquid cooling is applied to the CPUs, GPUs, and NVLink switches, removing up to 95% of generated heat and enabling the system to sustain maximum performance without thermal throttling—all within a compact 5U form factor rather than the larger chassis traditionally required for 8-GPU systems. This design not only improves energy efficiency but also reduces cooling costs in water-cooled data center environments.

The server is engineered for maximum GPU-to-GPU and GPU-to-CPU communication bandwidth, with PCIe 5.0 connectivity throughout and front-accessible PCIe slots directly linked to the GPU complex via PCIe switches. It is positioned as Lenovo’s flagship Intel-based AI server, designed entirely in-house for best-in-class modularity, thermal performance, and reliability.

Technical Specifications