Lenovo ThinkSystem SR680a V4 Server

• CPU: Two Intel Xeon 6700-series
• GPU: Eight NVIDIA B300 1100W GPUs
• Memory: Up to 32 DDR5 memory DIMMs
• Storage: 8x 2.5-inch hot-swap PCIe 5.0 NVMe drive bays
• Network: 4x PCIe 5.0 x16 FHHL slots; 1x OCP PCIe 5.0 x16 slot

Download Data Sheet

Example Use Cases

• Generative AI model training and fine-tuning — The server is purpose-built for generative AI, with dual Intel Xeon 6 CPUs, eight NVIDIA B300 HGX GPUs, NVLink GPU interconnects, and very high GPU memory bandwidth for large-model training workloads.
• High-throughput AI inference and reasoning — Lenovo positions it for advanced AI inference, noting up to 11x more inference performance versus the previous GPU generation, making it well suited for production LLM serving, agentic AI, and reasoning workloads.
• High-performance computing and scientific simulation — The guide explicitly calls out simulation and scientific research, and the platform includes HPC-oriented operating modes plus dense GPU acceleration and high-speed interconnects for compute-intensive technical workloads.
• Large-scale data analytics and in-memory databases — The system supports up to 4TB of DDR5 memory and Xeon 6 accelerators such as Intel IAA and DSA, which are intended to improve analytics, data movement, and in-memory database performance.
• Rendering and real-time video generation — Lenovo specifically lists rendering as a suggested use and notes the NVIDIA HGX B300 is designed for complex workloads including real-time video generation, which fits advanced media, visualization, and content-generation pipelines.

Would you like configs, estimated performance, or deployment recommendations for any use case?