Data center · accelerator
NVIDIA L40S
A passively cooled Ada Lovelace server GPU with 48 GB of GDDR6 ECC. It combines Tensor cores for AI, RT cores for rendering and hardware video engines on one PCIe card.
Technical specifications
| Architecture | NVIDIA Ada Lovelace |
|---|---|
| GPU memory | 48 GB GDDR6 ECC |
| Memory bandwidth | 864 GB/s |
| Host interface | PCIe Gen4 x16 |
| Maximum board power | 350 W |
| Cooling / size | Passive; 4.4 x 10.5 inches, dual-slot |
| Power connector | 16-pin |
| NVENC / NVDEC | 3 / 3; AV1 encode and decode |
| NVLink / MIG | Not supported |
Applications
- Training and fine-tuning neural networks.
- LLM and other neural-network inference.
- Ray-traced rendering, 3D visualization and simulation.
- Video encoding, decoding and AI video processing.
Compatibility
- Verify the exact server model, slot clearance, power cabling and cooling with the OEM. A PCIe slot alone does not establish compatibility. Passive cooling requires forced airflow from the server.
- L40S has no NVLink or MIG. Multiple cards do not provide a transparent shared memory pool, and vGPU is not the same feature as MIG.
- Check driver, CUDA/framework and operating-system support for the intended workload. Virtualization and enterprise software may require separate licenses and a supported host configuration.
Manufacturer sources
Specifications checked against the listed NVIDIA sources: .
Request a quote
Each quote is prepared individually for the selected variant, quantity and destination. The RFQ covers condition, configuration, source, warranty, availability and delivery timing.