The L40S is a high-performance datacenter GPU. Featuring 48GB of ultra-fast memory, it is engineered for the most demanding AI model training, large language models (LLMs), and complex scientific computing.
Recommended Scenarios
Deep Learning
Model Inference
Video Encoding
Architecture
Ada Lovelace
VRAM Capacity
48GB
Bandwidth
864 GB/s
CUDA Cores
18176
FP16 Perf.
183.2 TFLOPS
Power (TDP)
350W