Lenovo ThinkSystem SR675 V3 with NVIDIA B200 SXM6 192GB GPU

The Lenovo ThinkSystem SR675 V3 (7D9QA010NA) is a purpose-built, high-density 4U rack server engineered to host NVIDIA B200 SXM6 192GB GPUs within Lenovo's NVLink-optimized infrastructure. The system is designed around the AMD EPYC 9004 series processor architecture and is configured to support NVIDIA's GB200 NVL-class GPU pairing paradigm, delivering exceptional memory bandwidth and compute throughput for the most demanding AI and HPC workloads. The integration of the NVIDIA Blackwell B200 SXM6 GPU — featuring 192GB of HBM3e memory per GPU — enables large-scale model training, scientific simulation, and real-time inference at a scale previously unavailable in standard rack-mount form factors.

The NVIDIA B200 SXM6 GPU at the heart of this configuration delivers up to 20 petaFLOPS of FP4 AI performance per GPU, with the SXM6 form factor providing the high-bandwidth interconnect headroom required for multi-GPU training jobs and distributed inference across large language models. The ThinkSystem SR675 V3 chassis is purpose-engineered to accommodate the thermal and power envelope of Blackwell-generation SXM GPUs, with high-efficiency power subsystems and precision airflow management that supports the GPU's demanding TDP requirements. NVLink 5.0 interconnect capability enables GPU-to-GPU communication at up to 1.8TB/s bidirectional bandwidth, removing memory and bandwidth bottlenecks in transformer-based model workloads.

As an enterprise-grade GPU server, the SR675 V3 with B200 SXM6 is positioned for deployment in AI supercomputing clusters, national research infrastructure, cloud service provider GPU pools, and large enterprise AI Centers of Excellence. It supports full integration with Lenovo's XClarity management suite for lifecycle management, firmware orchestration, and health monitoring across large-scale deployments. The system is available through Omnixon Global for enterprise buyers across the UAE, GCC, EMEA, and APAC regions, offering a globally accessible procurement path for this premium GPU compute platform.

Ideal for

• Large-scale generative AI and large language model (LLM) training, including models exceeding 100 billion parameters requiring high-bandwidth multi-GPU memory pooling
• Enterprise inference serving for real-time AI applications such as retrieval-augmented generation (RAG), AI copilots, and multimodal inference pipelines
• High-performance computing (HPC) workloads in scientific research, computational fluid dynamics, molecular dynamics simulation, and climate modeling
• Deep learning-based computer vision and medical imaging analysis requiring high-throughput GPU memory and low-latency batch processing
• AI-accelerated data analytics and digital twin simulation within industrial and financial enterprise environments
• GPU cluster node deployment in sovereign AI infrastructure, cloud-native GPU-as-a-Service platforms, and national AI supercomputing initiatives

Technical specifications

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