L40S

The NVIDIA L40S is based on the Ada Lovelace architecture with 48 GB of GDDR6 VRAM. It provides datacenter-grade scalability in a versatile form factor, balancing compute performance with power efficiency for AI inference and mixed visual-compute workloads.

With 48 GB of VRAM, the L40S handles Large Language Models (LLM) and fine-tuning of medium-to-large models efficiently. It is well-suited for inference serving, video processing pipelines, and multi-modal AI applications where memory capacity and throughput need to scale across datacenter deployments.