Omnixon Global
Lenovo ThinkSystem GPU Module NVIDIA L40S 48GB PCIe

Lenovo ThinkSystem GPU Module NVIDIA L40S 48GB PCIe

Brand: Lenovo | Category: GPUs

SKU: LENO-4X67A84633 | Part #: 4X67A84633 | MPN: 4X67A84633

Contact for Pricing — Request a Quote

Request a Quote Contact Us

About the Lenovo ThinkSystem GPU Module NVIDIA L40S 48GB PCIe

The Lenovo ThinkSystem GPU Module NVIDIA L40S 48GB PCIe brings the full capability of NVIDIA's Ada Lovelace architecture to validated ThinkSystem server platforms, delivering a purpose-built solution for organizations deploying large-scale generative AI inference, high-performance computing, and professional visualization workloads. Built on the third-generation Ada Lovelace GPU architecture, the L40S integrates 18,176 CUDA cores, 568 fourth-generation Tensor Cores, and 142 RT Cores within a 48GB GDDR6 frame buffer, yielding up to 91.6 TFLOPS of FP32 compute and 733 TOPS of INT8 inferencing throughput. The high-capacity onboard memory, paired with 864 GB/s memory bandwidth, ensures that large language models, diffusion models, and complex simulation workloads remain resident in GPU memory without costly data staging overhead.

Validated as a Lenovo ThinkSystem add-in card, the L40S is fully integrated into Lenovo's system firmware, management tooling, and thermal designs, ensuring predictable performance characteristics in rackmount server environments. The module supports PCIe Gen 4 x16 connectivity and is compatible with multi-GPU configurations via NVLink bridge (where platform topology allows), enabling scaling of AI inference capacity proportionally with deployment needs. Its 300W TDP is managed through active cooling solutions validated by Lenovo engineering, maintaining stability across sustained enterprise workloads.

Launched in Q2 2025, this ThinkSystem-validated module targets enterprises that require a single-slot, PCIe-native accelerator without the overhead of NVSwitch fabric infrastructure, making it particularly attractive for inference-optimized server clusters, virtual workstation infrastructure via NVIDIA vGPU software, and hybrid AI-plus-visualization deployments. The combination of Ada Lovelace Tensor Cores, dedicated AV1 hardware encode/decode engines, and NVIDIA's mature ecosystem of AI frameworks — including TensorRT, Triton Inference Server, and CUDA — positions the L40S as a versatile, production-grade accelerator within Lenovo's enterprise AI portfolio.

Ideal for

  • Generative AI inference serving for large language models (LLMs) such as LLaMA, Mistral, and GPT-class architectures requiring large on-device memory footprints up to 48GB per GPU
  • Enterprise GPU virtualization using NVIDIA vGPU software to provision multiple concurrent virtual workstation or virtual compute instances from a single physical card
  • Scientific computing and HPC simulation workloads including molecular dynamics, computational fluid dynamics, and finite element analysis benefiting from high FP32 and FP64 throughput
  • Professional visualization rendering for CAD, digital twins, and media production pipelines where RT Core-accelerated ray tracing and large frame buffers reduce iterative render times
  • AI-assisted video transcoding and content processing leveraging dedicated AV1 hardware encode/decode engines in broadcast, media, and streaming infrastructure
  • Multi-model AI inference consolidation in data center environments where multiple inference workloads are co-located on ThinkSystem servers to maximize GPU utilization and reduce per-inference operational overhead

Technical specifications

ManufacturerLenovo
Product LineThinkSystem GPU Module
GPU ChipNVIDIA L40S (Ada Lovelace Architecture)
CUDA Cores18,176
Tensor Cores568 (4th Generation)
RT Cores142 (3rd Generation)
GPU Memory48GB GDDR6 with ECC
Memory Bandwidth864 GB/s
FP32 Performance91.6 TFLOPS
FP16 Performance (Tensor)362.05 TFLOPS (with sparsity: 724.1 TFLOPS)
INT8 Throughput733 TOPS (with sparsity: 1457 TOPS)
FP64 Performance1.45 TFLOPS
InterfacePCIe Gen 4 x16
Form FactorDual-slot, full-height full-length (FHFL) add-in card
Thermal Design Power (TDP)300W
CoolingActive (server-validated forced-air cooling)
Display Outputs4x DisplayPort 1.4a
Multi-GPU SupportNVLink (platform-dependent topology)
vGPU SupportNVIDIA vGPU software compatible (NVIDIA AI Enterprise license required)
Video Encode/DecodeAV1, H.265, H.264 hardware encode and decode engines
Validated PlatformsLenovo ThinkSystem Gen 3 and Gen 4 servers
Launch Generation2025-Q2
OS / Framework SupportLinux, Windows Server; CUDA, TensorRT, Triton Inference Server, PyTorch, TensorFlow

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

Technical Specifications

BrandLenovo
CategoryGPUs
SKULENO-4X67A84633
Part Number4X67A84633
ConditionNew
Product LineThinkSystem GPU Module
GPU ChipNVIDIA L40S (Ada Lovelace Architecture)
CUDA Cores18,176
Tensor Cores568 (4th Generation)
RT Cores142 (3rd Generation)
GPU Memory48GB GDDR6 with ECC
Memory Bandwidth864 GB/s
FP32 Performance91.6 TFLOPS
FP16 Performance (Tensor)362.05 TFLOPS (with sparsity: 724.1 TFLOPS)
INT8 Throughput733 TOPS (with sparsity: 1457 TOPS)
FP64 Performance1.45 TFLOPS
InterfacePCIe Gen 4 x16
Form FactorDual-slot, full-height full-length (FHFL) add-in card
Thermal Design Power (TDP)300W
CoolingActive (server-validated forced-air cooling)
Display Outputs4x DisplayPort 1.4a
Multi-GPU SupportNVLink (platform-dependent topology)
vGPU SupportNVIDIA vGPU software compatible (NVIDIA AI Enterprise license required)
Video Encode/DecodeAV1, H.265, H.264 hardware encode and decode engines
Validated PlatformsLenovo ThinkSystem Gen 3 and Gen 4 servers
Launch Generation2025-Q2
OS / Framework SupportLinux, Windows Server; CUDA, TensorRT, Triton Inference Server, PyTorch, TensorFlow

Frequently Asked Questions about Lenovo ThinkSystem GPU Module NVIDIA L40S 48GB PCIe

What does the Lenovo ThinkSystem GPU Module NVIDIA L40S 48GB PCIe do?

The Lenovo ThinkSystem GPU Module NVIDIA L40S 48GB PCIe accelerates AI/ML training, inference, scientific HPC, and virtualization (vGPU) workloads. Typical deployments include LLM training clusters, computer-vision pipelines, financial risk modeling, and rendering farms.

What are the headline specs of the Lenovo ThinkSystem GPU Module NVIDIA L40S 48GB PCIe?

Key specifications for the Lenovo ThinkSystem GPU Module NVIDIA L40S 48GB PCIe: new condition; manufacturer Lenovo; product line ThinkSystem GPU Module; gpu chip NVIDIA L40S (Ada Lovelace Architecture); cuda cores 18,176; tensor cores 568 (4th Generation); rt cores 142 (3rd Generation). Manufacturer part number 4X67A84633. For the full datasheet with electrical, environmental, and compliance details, contact our pre-sales engineering team.

Is the Lenovo ThinkSystem GPU Module NVIDIA L40S 48GB PCIe compatible with my infrastructure?

The Lenovo ThinkSystem GPU Module NVIDIA L40S 48GB PCIe requires a PCIe Gen4 or Gen5 x16 slot, server power adequate for the card's TDP, and CUDA/ROCm driver support in your hypervisor or bare-metal OS. Sales engineering will confirm chassis fit (1U/2U/4U), PCIe lane count, and PSU headroom before quoting.