Less than 8 hours
to fine-tune LMM
The industry's most powerful universal GPU
Unlock the next level of multi-workload performance with the NVIDIA L40S GPU on our advanced AI servers. The L40S GPU is designed to handle the next generation of data center workloads, from generative AI and large language model (LLM) inference and training to 3D graphics, rendering, and video. It combines powerful AI computing with best-in-class graphics and media acceleration.
Breakthrough AI Performance By the Numbers
Less than 4 days
to train GPT3
Over 82 images/min
for image generative AI
Accelerating End-to-End Generative AI
L40S is the best GPU for accelerating end-to-end generative AI pipelines, powering improved auto-speech recognition, large language models, text-to-speech, text-to-image, audio-to-face, and rendering.
L40S Features: Cutting-Edge Capabilities to Transform AI
Ada Lovelace Architecture
features latest generation Tensor cores, RT cores, and CUDA cores to empower workloads from 3D rendering to generative AI
Transformer Engine
mixes FP8 and FP16 precisions to accelerate performance, improve memory utilization for inference and training
Unparalleled Performance
1.5 PFLOPS Tensor performance, 212 TFLOPS RT Core performance, and 91.6 TFLOPS SP performance
L40S Performance Comparison
Aivres AI Servers Supporting L40S
KR4268V2 (AMD)
Up to 8 L40S GPUs
Supporting AMD EPYC™ 9004 processors
Learn More
KR4268V2 (Intel)
Up to 8 L40S GPUs
Supporting 4th Gen Intel® Xeon® Scalable processors
Learn More
