Lenovo ThinkSystem SR675 V3 with AMD Instinct MI325X 256GB

The Lenovo ThinkSystem SR675 V3 (7D9QCTO5WW) is a 2U dual-socket AMD EPYC-based rack server engineered specifically to support high-density GPU accelerator configurations. Built on the AMD EPYC 9004 series processor platform, the SR675 V3 provides the PCIe 5.0 bandwidth, memory capacity, and thermal infrastructure required to fully exploit the AMD Instinct MI325X accelerator at scale. The chassis supports up to eight double-wide OAM-format GPU modules and is designed to meet the rigorous demands of modern AI training, HPC simulation, and large-scale inferencing environments in enterprise and hyperscale datacenter deployments.

The AMD Instinct MI325X 256GB accelerator at the heart of this configuration leverages AMD CDNA 3 architecture and delivers high-bandwidth memory capacity through HBM3 technology, offering 256 GB of on-package memory per GPU with aggregate memory bandwidth suited to memory-intensive AI and scientific workloads. The MI325X supports AMD ROCm open software ecosystem, enabling compatibility with popular ML frameworks including PyTorch and TensorFlow, and is interconnected via AMD Infinity Fabric for high-speed GPU-to-GPU communication within the node. This combination of processor platform and accelerator makes the SR675 V3 a capable foundation for generative AI model training, large language model fine-tuning, and molecular dynamics workloads.

As a Lenovo ThinkSystem platform, the SR675 V3 incorporates Lenovo XClarity management integration for centralized monitoring, firmware lifecycle management, and telemetry across fleets of servers. The system supports high-capacity DDR5 RDIMM memory across 24 DIMM slots, multiple NVMe storage options, and redundant power supply units scaled to the power envelope of fully populated GPU configurations. Its architecture is optimized for deployment in UAE, GCC, EMEA, and APAC enterprise datacenters requiring proven, standards-based accelerated compute infrastructure.

Ideal for

• Large language model (LLM) pre-training and fine-tuning requiring high aggregate GPU memory across multi-node GPU clusters
• Generative AI inference serving for enterprise applications demanding low-latency, high-throughput GPU compute
• High-performance computing (HPC) workloads including computational fluid dynamics, molecular dynamics simulation, and climate modeling
• Computer vision and deep learning pipeline development at scale within research and enterprise AI platforms
• Financial services risk modeling and Monte Carlo simulation leveraging GPU-accelerated numerical computation
• Healthcare and life sciences genomics analysis and drug discovery workloads requiring sustained GPU memory bandwidth

Technical specifications

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