The NVIDIA T4 is a cutting-edge, single-slot, low-profile PCI Express Gen3 Universal Deep Learning Accelerator. It's built on the TU104 NVIDIA graphics processing unit (GPU) and boasts 16 GB of GDDR6 memory with a maximum power limit of 70 W. Designed to fit seamlessly into various systems, the T4 is offered as a passively cooled board, requiring system airflow to maintain optimal thermal conditions. It's compatible with both x8 and x16 PCI Express configurations. Renowned for its versatility, the NVIDIA T4 serves as a universal deep learning accelerator, making it ideal for distributed computing environments. Powered by NVIDIA Turing Tensor Cores, the T4 delivers groundbreaking multi-precision performance, accelerating deep learning and machine learning training and inference tasks, as well as video transcoding and virtual desktop operations. Part of the NVIDIA AI Platform, the T4 supports all AI frameworks and network types, ensuring exceptional performance and efficiency for large-scale deployments.
