L40 GPU

The NVIDIA L40 is a professional graphics and compute GPU based on the Ada Lovelace architecture, positioned as a versatile workstation and data center solution. Built on a 4nm manufacturing process, the L40 bridges the gap between consumer gaming GPUs and dedicated data center accelerators, offering 48GB of GDDR6 memory with ECC for professional workloads that require substantial memory capacity and reliability. The L40 features 18,176 CUDA cores and 568 fourth-generation Tensor Cores, delivering 90.5 TFLOPS of FP32 performance and 181.05 TFLOPS of FP16 performance. With 864 GB/s of memory bandwidth across a 384-bit memory interface, the GPU provides sufficient throughput for memory-intensive applications. The Ada Lovelace architecture introduces third-generation RT Cores for hardware-accelerated ray tracing and supports 8K AV1 encoding and decoding capabilities. In cloud deployments, the L40 serves workloads requiring large memory footprints combined with professional-grade reliability, including AI model training with medium to large datasets, 3D rendering pipelines, virtualized desktop infrastructure, and multi-tenant workstation environments where the 48GB memory capacity can be shared across multiple users or applications.