NVIDIA 900-2G133-0010-000 L40 48GB GDDR6 384bit Graphics card
The NVIDIA L40 (part number 900-2G133-0010-000) is a datacenter GPU built on NVIDIA’s Ada Lovelace architecture family, positioned for AI inference, edge AI, and professional visualization workloads. With 48GB of GDDR6 memory on a 384-bit interface and a low ~70W power profile (guidance only, verify board-specific draw), it targets integrators and OEMs deploying GPU servers in energy-constrained or space-limited environments. The passive thermal design makes it suited for rack servers with proper chassis airflow rather than desktop workstations.
Key specifications (table)
| Parameter | Specification |
| Part number | 900-2G133-0010-000 |
| GPU family | NVIDIA L40 (Ada Lovelace architecture generation, datacenter/professional line) |
| Memory capacity | 48GB GDDR6 |
| Memory interface | 384-bit |
| TDP | ~70W (guidance only; actual system power requirements depend on host platform and configuration) |
| Cooling | Passive heatsink (requires server chassis with directed airflow) |
| Form factor focus | Compact, space-saving design for tight enclosures |
| Target workloads | Edge AI, AI inference, energy-constrained enterprise deployments |
| Framework support | TensorFlow, PyTorch, ONNX and other major AI frameworks |
| Positioning | Enterprise-ready build with long-term availability and software support |
Good to know
- Because the card uses passive cooling, confirm your server chassis provides sufficient front-to-back airflow before deployment; it is intended for rack servers, not standard desktop cases.
- Within the L-series family, the L40 sits alongside higher-power sibling SKUs; this listing is positioned toward low-power, inference-oriented deployments rather than maximum-throughput training builds.
- When buying, verify driver and framework compatibility (CUDA versions for TensorFlow/PyTorch/ONNX) with your existing software stack, and confirm power and slot clearances on your target motherboard.
For tray and reel pricing, current stock by part number and date code, and volume discounts on bulk orders, request a bulk quotation.
The NVIDIA L40 (part number 900-2G133-0010-000) is a datacenter GPU built on NVIDIA’s Ada Lovelace architecture family, positioned for AI inference, edge AI, and professional visualization workloads. With 48GB of GDDR6 memory on a 384-bit interface and a low ~70W power profile (guidance only, verify board-specific draw), it targets integrators and OEMs deploying GPU servers in energy-constrained or space-limited environments. The passive thermal design makes it suited for rack servers with proper chassis airflow rather than desktop workstations.
Key specifications (table)
| Parameter | Specification |
| Part number | 900-2G133-0010-000 |
| GPU family | NVIDIA L40 (Ada Lovelace architecture generation, datacenter/professional line) |
| Memory capacity | 48GB GDDR6 |
| Memory interface | 384-bit |
| TDP | ~70W (guidance only; actual system power requirements depend on host platform and configuration) |
| Cooling | Passive heatsink (requires server chassis with directed airflow) |
| Form factor focus | Compact, space-saving design for tight enclosures |
| Target workloads | Edge AI, AI inference, energy-constrained enterprise deployments |
| Framework support | TensorFlow, PyTorch, ONNX and other major AI frameworks |
| Positioning | Enterprise-ready build with long-term availability and software support |
Good to know
- Because the card uses passive cooling, confirm your server chassis provides sufficient front-to-back airflow before deployment; it is intended for rack servers, not standard desktop cases.
- Within the L-series family, the L40 sits alongside higher-power sibling SKUs; this listing is positioned toward low-power, inference-oriented deployments rather than maximum-throughput training builds.
- When buying, verify driver and framework compatibility (CUDA versions for TensorFlow/PyTorch/ONNX) with your existing software stack, and confirm power and slot clearances on your target motherboard.
For tray and reel pricing, current stock by part number and date code, and volume discounts on bulk orders, request a bulk quotation.
