Next-Gen Tesla T4 GPU Announced at GTC 2019
Peter Donnell / 4 years ago
Next-Gen Tesla T4
While this tech is far from the consumer market right now, Nvidia just dropped a hardware bomb at GTC 2019. They teased what is basically the 2nd generation of Tensor Core based GPU hardware. Albeit, all for server customers at this time. It features 64GB of GDDR6 ultra-fast VRAM memory while using 4x T4 cards in a deployable server mount. These are built for computational workloads for distributed computing. Bringing “hyperscale” computing down to a rack mount.
“The NVIDIA® T4 GPU accelerates diverse cloud workloads, including high-performance computing, deep learning training and inference, machine learning, data analytics, and graphics. Based on the new NVIDIA Turing™ architecture and packaged in an energy-efficient 70-watt, small PCIe form factor, T4 is optimised for mainstream computing environments and features multi-precision Turing Tensor Cores and new RT Cores. Combined with accelerated containerized software stacks from NGC, T4 delivers revolutionary performance at scale.” – Nvidia
The card clocks in as a single-slot low-profile design. It’s just 6.6″ long, and uses PCIe Gen3. Using the new TU104 Nvidia GPU, 16 GB GDDRT6 and just 70W max power limit meaning it can be air cooled with ease. It will support 8x and 16x PCIe.
Instead of a busy network of data centre hardware, the next-generation of Nvidia data centres will blend the network and the computer into one living ecosystem. This is no surprise, with Mellanox Technologies being the latest purchase by Nvidia. Basically, the next-gen of their ethernet, making it more integral with the servers/nodes. Replacing the weakest link in a chain of hyperscale computing, basically.
Between the new cards, being able to run huge processing tasks and scalability, as well as the new interconnect system, Nvidia hope to dominate the next-generation of data centres and computation.