NVIDIA L40S PCIe GPU 900-2G133-0000-000

Brand: NVIDIA | Category: GPUs

SKU: NVIDIA-900-2G133-0000-000 | Part #: 900-2G133-0000-000 | MPN: 900-2G133-0000-000

Contact for Pricing — Request a Quote

The NVIDIA L40S PCIe GPU (part number 900-2G133-0000-000) is a full-height single-slot, passively cooled accelerator built on the NVIDIA Ada Lovelace architecture for enterprise AI and high-performance computing environments. With 48 GB of GDDR6 memory and 864 GB/s memory bandwidth, the L40S delivers 1,457 TFLOPS of peak tensor performance in TF32 precision, making it suitable for demanding workloads that require both capacity and throughput without active cooling overhead. The GPU connects via PCIe 4.0 x192 and draws up to 350 W, powered by dual 6-pin 12V server-class connectors that integrate seamlessly into standard enterprise infrastructure.

NVIDIA's L40S is the preferred choice for AI infrastructure teams deploying large language model training, inference pipelines, and mixed HPC workloads at scale. The passive air-cooled design eliminates blower noise and maintenance concerns, while the single-slot form factor maximizes GPU density in multi-GPU server configurations. Every NVIDIA L40S (900-2G133-0000-000) includes a three-year manufacturer warranty, with extended coverage options available. To source this accelerator for your enterprise deployment, contact Omnixon Global for an RFQ.

Typical Workloads & Deployment Scenarios

• Large Language Model (LLM) training and fine-tuning with high-capacity memory pools
• Real-time inference serving for generative AI applications and transformer models
• High-performance computing (HPC) simulations requiring sustained memory bandwidth
• Multi-GPU cluster deployments leveraging single-slot passive cooling for thermal efficiency
• Data center acceleration where noise reduction and space optimization are critical constraints