The Lenovo ThinkSystem SR670 delivers optimal performance for Artificial Intelligence (AI) and high-performance computing (HPC) workloads while maintaining a low total cost of ownership (TCO). The ThinkSystem SR670 allows up to either four large or eight small GPUs per 2U node, and is suited for computationally intensive workload requirements for Machine Learning (ML), Deep Learning (DL), or Inference.
Built on the latest Intel® Xeon® processor Scalable family CPUs and designed to support high-end GPUs including NVIDIA Tesla V100 or T4, the ThinkSystem SR670 delivers optimized performance for AI training and accelerated HPC workloads. Capabilities include:
Full-width 2U enclosure
Up to 2x Intel® Xeon® processor Scalable family CPUs per node
Up to 768GB using 24x 2667MHz TruDDR4 DIMMs per node
Up to 3 PCIe adapters: 2x PCIe 3.0 x16 + 1x PCIe 3.0 x4 slots
Up to 4 dual-slot, full-height, full-length GPUs (each PCIe 3.0 x16 slots), or up to 8 single-slot, full-height, half-length GPUs (each PCIe 3.0 x8 slots)