NVIDIA L40S 48GB Specifications for AI Enthusiasts
The NVIDIA L40S GPU was introduced in August 2023 (part number 900-2g133-0080-000). It is the successor to the NVIDIA L40 (part number 900-2G133-0010-000) and like the L40 is based on the AD102 chip from the Ada Lovelace architecture. Compared with the L40 it has nearly double the TF32 Tensor Core TFLOPS and FP16 Tensor Core performance, and increases the maximum power draw from 300W to 350W. It is targeted for generative AI and graphics and video applications. It lacks FP64 precision, but has great FP32, FP16 and mixed-precision performance. It uses the PCIe Gen4 x16 interconnect and Unlike the NVIDIA L4, it allows connecting monitors with 4x DisplayPort connectors.
- GPU Architecture: ada
- Hardware-Accelerated GEMM Operations:FP16 FP32 BF16 FP8 INT8 INT4 TF32 FP64 INT1
- CUDA Compute Capability : 8.9
Specifications for NVIDIA L40S
| Raw Performance | |
|---|---|
| Tensor Core Count: 568 | |
| FP32 TFLOPs: 183 | |
| FP16 TFLOPs: 362 | |
| Int8 TOPs: 1466 | |
| Memory Capacity (GB): 48 | |
| Memory Bandwidth (GB/s): 864 |
Real-time NVIDIA L40S GPU Prices
We're tracking 21 of the NVIDIA L40S GPUs currently available for sale. The lowest price is $7,500
Buy NowCompare Price/Performance to other GPUs
We track real-time prices of other GPUs too so that you can compare the price/performance of the NVIDIA L40S GPU to other GPUs.
Compare GPU Price/PerformanceReferences
- https://www.nvidia.com/en-us/data-center/l40s/
- https://www.exxactcorp.com/blog/components/NVIDIA-L40S-GPU-Compared-to-A100-and-H100-Tensor-Core-GPU
- https://resources.nvidia.com/en-us-l40s/nvidia-l40s-product
- https://nvidianews.nvidia.com/news/nvidia-global-data-center-system-manufacturers-to-supercharge-generative-ai-and-industrial-digitalization
- https://flyytech.com/2023/08/08/nvidia-completes-proviz-ada-lovelace-lineup-with-three-new-graphics-cards/
- https://www.nvidia.com/en-us/technologies/ada-architecture/