The NVIDIA L20 GPU is a powerful universal GPU for the data center, delivering end-to-end acceleration for the next generation of AI-enabled applications?from generative AI and model training to 3D graphics and video. Built on the NVIDIA Ada Lovelace architecture, it features fourth-generation Tensor Cores and third-generation RT Cores to accelerate AI-enhanced graphics and large language model (LLM) inference. The L20 is designed for energy-efficient data center deployment with enterprise-ready security and full vGPU software support.
Specifications
| Feature | Details |
| FP32 Performance | 59.8 TFLOPS |
| TF32 Tensor Core | 59.8 TFLOPS |
| BFLOAT16 Tensor Core | 119.5 TFLOPS | 239 TFLOPS* |
| FP16 Tensor Core | 119.5 TFLOPS | 239 TFLOPS* |
| INT8 Tensor Core | 239 TFLOPS | 478 TOPS* |
| GPU Memory | 48 GB GDDR6 with ECC |
| Memory Bandwidth | 864 GB/s |
| RT Cores | 92 (Third-Generation) |
| CUDA Cores | 11,776 |
| L2 Cache | 96 MB |
System & Connectivity
| Feature | Details |
| System Interface | PCIe Gen4 x16: 64 GB/s |
| Max Power Consumption | 275W |
| Thermal Solution | Passive |
| Form Factor | 2-slot FHFL (Full-Height, Full-Length) |
| Display Connectors | 4x DisplayPort 1.4a |
| Power Connector | 1x 16-pin PCIe |
| Media Engine | 3x NVENC (+AV1), 3x NVDEC, 4x NVJPEG |
| vGPU Software Support | NVIDIA vPC/vApps, NVIDIA RTX Virtual Workstation (vWS) |
| Security | Secure and Measured Boot with Hardware Root of Trust |
