NVIDIA L40S 48GB PCIe Gen4 Active Cooling GPU for HPE ProLiant

The NVIDIA L40S 48GB PCIe Gen4 Active Cooling GPU for HPE ProLiant (P65892-B21) is built on the NVIDIA Ada Lovelace architecture, delivering a highly versatile accelerator engineered for demanding enterprise AI inference, training, and professional visualization workloads. With 48GB of GDDR6 ECC memory across a 384-bit memory bus, the L40S provides the large memory footprint required for running large language models, generative AI applications, and complex simulation workloads within a standard PCIe form factor — enabling deployment across a wide range of HPE ProLiant server platforms without requiring proprietary interconnects.

The L40S integrates third-generation RT Cores, fourth-generation Tensor Cores, and NVIDIA CUDA cores to support concurrent AI compute and graphics rendering in a single GPU. With a peak FP8 throughput of 1457 TOPS and FP16 Tensor Core performance of 362 TFLOPS, the card is positioned for high-throughput inferencing of transformer-based models alongside 3D rendering and simulation tasks. Its PCIe Gen4 x16 interface ensures broad server compatibility and sufficient bandwidth for enterprise-scale AI and visualization pipelines.

Factory-integrated into HPE's ProLiant ecosystem, this option kit ships with HPE-validated drivers, firmware, and thermal management via an active cooling solution suited to datacenter rack environments. The GPU supports NVIDIA MIG (Multi-Instance GPU) is not applicable to the L40S; instead it supports NVIDIA MPS and time-slicing for workload consolidation. Designed for UAE, GCC, EMEA, and APAC enterprise datacenters, the P65892-B21 is available through Omnixon Global for organizations accelerating AI infrastructure and high-performance compute deployments.

Ideal for

• Large language model (LLM) inference serving for enterprise AI applications requiring high throughput and large VRAM capacity
• Generative AI workloads including image synthesis, video generation, and multimodal model deployment in production datacenters
• Professional 3D visualization and real-time rendering for engineering simulation, digital twin, and media production environments
• High-performance computing (HPC) workloads such as computational fluid dynamics, molecular dynamics, and scientific modeling on HPE ProLiant platforms
• AI-assisted medical imaging analysis and healthcare analytics requiring GPU-accelerated inference with large model support
• Virtual workstation GPU acceleration for remote design, CAD, and visualization via GPU sharing across enterprise user pools

Technical specifications

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