Lenovo ThinkSystem SR670 V2 NVIDIA H100 NVLink GPU Tray

The Lenovo ThinkSystem SR670 V2 NVIDIA H100 NVLink GPU Tray (4C17A81847) is a purpose-engineered compute tray that integrates NVIDIA's H100 Tensor Core GPU into the ThinkSystem SR670 V2 server platform via NVLink interconnect topology. Built on NVIDIA's Hopper architecture and manufactured on a 4nm process node, the H100 GPU delivers a substantial leap in AI training and inference throughput compared to prior generations, leveraging the Transformer Engine with FP8 precision to accelerate large language model and deep learning workloads at datacenter scale.

The H100 GPU incorporated in this tray utilizes NVLink 4.0, enabling high-bandwidth, low-latency GPU-to-GPU communication within multi-GPU configurations. With 80GB of HBM3 memory and a memory bandwidth of 3.35 TB/s, the GPU sustains the data movement demands of memory-intensive workloads including scientific simulation, genomics processing, and large-scale model training. The SXM5 form factor used in this tray configuration delivers higher power headroom and thermal dissipation than PCIe-based alternatives, allowing sustained peak performance in demanding datacenter environments.

Designed for enterprise IT and hyperscale datacenter operators, the 4C17A81847 tray integrates seamlessly into the ThinkSystem SR670 V2 chassis, which supports up to four such GPU trays in a validated, thermally managed configuration. The solution is qualified under Lenovo's server ecosystem, ensuring firmware compatibility, systems management integration via Lenovo XClarity, and validated interoperability for organizations deploying AI infrastructure across cloud, on-premises, and hybrid environments in regions including the UAE, GCC, EMEA, and APAC.

Ideal for

• Large language model (LLM) training and fine-tuning at enterprise scale, leveraging FP8 Transformer Engine precision for accelerated iteration cycles
• High-performance inference serving for generative AI applications, including real-time natural language processing and multimodal model deployment
• Computational fluid dynamics and finite element analysis simulations in engineering and energy sectors requiring double-precision floating-point throughput
• Genomics sequencing pipelines and molecular dynamics research in life sciences environments demanding high memory bandwidth and capacity
• Seismic data processing and subsurface imaging for oil and gas exploration workloads running at scale in HPC clusters
• Enterprise AI platform build-out within sovereign datacenter infrastructure across GCC and EMEA regions requiring validated, rack-integrated GPU compute

Technical specifications

Available from Omnixon Global. Submit an RFQ and our team will confirm configuration and availability for your order.