NVIDIA · Data center · Accelerator card
NVIDIA L40 48GB
Exact L40 reference specifications and AI deployment guidance, separated from FusionPointAI commercial inventory.
Recent generation
Reference only
Verified technical specifications
| Manufacturer | NVIDIA |
|---|---|
| Exact SKU | L40 |
| Product family | NVIDIA L40 |
| Architecture | Ada Lovelace |
| Generation | Ada Data Center |
| Entity type | Accelerator card |
| Lifecycle | Recent generation |
| Launch date | The manufacturer has not published this information. |
| VRAM | 48 GB |
| Memory technology | GDDR6 ECC |
| Memory bandwidth | 864 GB/s |
| Power | 300 W |
| Interface | PCIe Gen4 x16 |
| Form factor | Dual slot · PCIe · Accelerator |
| Interconnect | PCIe |
| MIG / partitioning | Not supported |
| AI precision support | FP32, TF32, BF16, FP16, FP8, INT8 |
| Software stack | cuda, vllm, sglang, tensorrt-llm |
AI workload suitability
LLM inference
Best suited to smaller checkpoints, compressed models or distributed configurations sized by the recommendation engine.
Fine-tuning and training
Designed for sustained accelerator workloads; training suitability depends on interconnect, precision and cluster topology.
Image, video and embeddings
FP16 acceleration supports common generative and vector workloads within the listed VRAM envelope.
Power and cooling
300 W published power requires chassis, airflow and power delivery for this exact form factor.
Multi-GPU use
PCIe is the recorded interconnect path; topology must still be validated at node level.
Software ecosystem
cuda, vllm, sglang, tensorrt-llm
Product timeline
Media gallery
Official manufacturer source
https://www.nvidia.com/en-us/data-center/l40/
Last verified: 2026-08-24