Nvidia launches tesla t4, the fastest card for data centers

Nvidia has announced its new GPU for machine learning and inference in data centers. The new Tesla T4 card is based on the Turing architecture, and is the successor to the current Tesla P4 model, which covers virtually all major cloud computing providers.

Nvidia Tesla T4 wants to revolutionize data centers

Nvidia argues that the Tesla T4s are significantly faster than the P4s. For language inference, for example, T4 is 34 times faster than using a CPU and more than 3.5 times faster than P4. The maximum throughput for Tesla P4 is 260 TOPS for 4-bit integer operations and 65 TOPS for floating point operations. T4 is on a standard profile 75-watt PCI-e card.

We recommend reading our post on Details of Nvidia's Turing architecture appear

Most importantly, Nvidia designed these chips specifically for AI inference, which makes Tesla T4 such an efficient GPU for making inferences is Turing's new Tensor Core. Jensen Huang has already talked about the new Tensor Core, and what it can do for gaming, rendering, and AI operations. In total, the chip features 320 Tensor Core Turing and 2, 560 CUDA cores.

In addition to the new chip, Nvidia is also rolling out an update to its TensorRT software to optimize deep learning models. This new version also includes the TensorRT Inference Server, a fully containerized microservice for data center inference that seamlessly connects to an existing Kubernetes infrastructure.

Turing is one of the most innovative Nvidia architectures, even in the gaming market where it will allow for the first time to use ray tracing in real time to improve the graphic finish of the most cutting-edge games.

Techpowerup font