NVIDIA 900-2G133-0080-000 L20 48GB GDDR6 384bit Graphics card
The NVIDIA L20 48GB (part number 900-2G133-0080-000) is a datacenter GPU aimed at system integrators and enterprises building AI inference, rendering, and virtualized workstation deployments. Its low power draw and passive, fanless thermal design make it a strong fit for thermally constrained enclosures, edge servers, and remote installations where maintenance access is limited. Note that passive cooling requires adequate chassis airflow — verify your server platform supports passively cooled GPU modules before ordering.
Key specifications (table)
| Parameter | Specification |
|---|---|
| Part number | 900-2G133-0080-000 |
| GPU family | NVIDIA L20 |
| Memory capacity | 48GB |
| Memory type | GDDR6 |
| Memory interface | 384-bit |
| TDP | ~70W (guidance only; ideal for thermally constrained or remote deployments) |
| Cooling | Passive (fanless, requires compatible chassis airflow) |
| Form factor | Compact — suits small enclosures, edge servers, and embedded platforms |
| AI framework support | TensorFlow, PyTorch, ONNX |
Good to know
- Compatibility: as a passively cooled datacenter card, it is intended for rack servers and embedded platforms with proper forced airflow; it is not designed for desktop workstation use. Confirm PCIe power and slot clearance in your chassis.
- Family positioning: within NVIDIA’s datacenter lineup, the L20 targets efficiency-focused deployments, prioritizing low power draw and compact integration over the higher compute density of flagship training accelerators — a common choice for inference-heavy workloads.
- Buying tip: part numbers in this family have multiple variants (e.g. -00A0-000 revisions); always match the exact part number to your platform qualification list before purchasing to avoid firmware or compatibility mismatches.
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 L20 48GB (part number 900-2G133-0080-000) is a datacenter GPU aimed at system integrators and enterprises building AI inference, rendering, and virtualized workstation deployments. Its low power draw and passive, fanless thermal design make it a strong fit for thermally constrained enclosures, edge servers, and remote installations where maintenance access is limited. Note that passive cooling requires adequate chassis airflow — verify your server platform supports passively cooled GPU modules before ordering.
Key specifications (table)
| Parameter | Specification |
|---|---|
| Part number | 900-2G133-0080-000 |
| GPU family | NVIDIA L20 |
| Memory capacity | 48GB |
| Memory type | GDDR6 |
| Memory interface | 384-bit |
| TDP | ~70W (guidance only; ideal for thermally constrained or remote deployments) |
| Cooling | Passive (fanless, requires compatible chassis airflow) |
| Form factor | Compact — suits small enclosures, edge servers, and embedded platforms |
| AI framework support | TensorFlow, PyTorch, ONNX |
Good to know
- Compatibility: as a passively cooled datacenter card, it is intended for rack servers and embedded platforms with proper forced airflow; it is not designed for desktop workstation use. Confirm PCIe power and slot clearance in your chassis.
- Family positioning: within NVIDIA’s datacenter lineup, the L20 targets efficiency-focused deployments, prioritizing low power draw and compact integration over the higher compute density of flagship training accelerators — a common choice for inference-heavy workloads.
- Buying tip: part numbers in this family have multiple variants (e.g. -00A0-000 revisions); always match the exact part number to your platform qualification list before purchasing to avoid firmware or compatibility mismatches.
For tray and reel pricing, current stock by part number and date code, and volume discounts on bulk orders, request a bulk quotation.
