NVIDIA L40S
- Type: Desktop
- Process Node: 5 nm
- Memory Types: GDDR6
- Released Date: Oct 2022
Full Specifications And Prices
LAUNCH
| Announced | Oct 2022 |
| Available | Yes |
Board Design
| Type | Desktop |
| Foundry | TSMC |
| TDP | 300W |
| Output Port | 1x HDMI 2.1 3x DisplayPort 1.4a |
| Suggested PSU | 700W |
| Power Connectors | 1x 16-pin |
| Interface | PCIe 4.0 x16 |
| Generation | Tesla Ada |
GPU Features
| CUDA | 8.9 |
| DirectX | 12 Ultimate (12_2) |
| OpenCL version | 3.0 |
| OpenGL | 4.6 |
| Shader Model | 6.7 |
| Vulkan version | 1.3 |
Graphics Processor
| Architecture | Ada Lovelace |
| Fabrication process | 5 nm |
| Die Size | 609mm² |
| GPU name | AD102 |
| GPU Variant | AD102-???-A1 |
| Transistor count | 76,300 million |
Memory Specs
| Memory type | GDDR6 |
| Max size | 48GB |
| Max bandwidth | 864.0 GB/s |
| Memory Bus | 384bit |
| ECC Memory Support |
Clock Speeds
| Base Clock | 1110MHz |
| Boost Clock | 2520MHz |
| Memory Clock | 2250MHz |
Render Config
| L1 cache | 128 KB (per SM) |
| L2 cache | 48MB |
| Shading units | 18176 |
| TMUs | 568 |
| RT Cores | 142 |
| ROPs | 192 |
| GPU Major Cores | 142 |
| Tensor Cores | 568 |
Theoretical Performance
| FP16 (half) | 91.61 TFLOPS |
| FP32 (float) | 91.61 TFLOPS |
| FP64 (double) | 1431 GFLOPS |
| Texture Rate | 1431 GTexel/s |
| Pixel Rate | 483.8 GPixel/s |
Reviews And Comments
Disclaimer Note
Disclaimer. We can not guarantee that the information on this page is 100% correct.
