Designed specifically for developers, designers, and artists from various industries, the Nvidia Titan RTX is based on the same Turing GPU and thanks to specific Turing Tensor cores, it provides significant performance from FP32, FP16, INT8, and INT4, making it suitable for training and inference of neural networks, while 24GB of GDDR6 memory, as well as NVLink, allows researches to use larger neural networks and data sets.
Nvidia CEO, Jensen Huang, has once again restated that "Turing is Nvidia's biggest advance in a decade – fusing shaders, ray tracing, and deep learning to reinvent the GPU". Calling it the T-Rex of Turing, Huang added that "the introduction of T-Rex puts Turing within reach of millions of the most demanding PC users — developers, scientists and content creators”.
Specification-wise, the Nvidia Titan RTX uses a fully-enabled 12nm 18.6 billion transistor TU102 Turing GPU with 6 Graphics Processing Clusters (GPCs), 36 Texture Processing Clusters (TPCs), 72 Streaming Multiprocessors for 4608 CUDA cores, 576 Tensor Cores, and 72 dedicated RT Cores. It has a base clock of 1350MHz, Boost clock of 1770MHz, and packs 24GB of 7,000MHz clocked GDDR6 memory on a 384-bit memory interface, pumping out 672GB/s of total memory bandwidth.
As it is a premium product, the Titan RTX features "a forged and machine-finished diecast aluminum cover" with two 13-blade fans and a hefty vapor-chamber. It features three DisplayPort, single HDMI, and one USB Type-C port, and needs two 8-pin PCIe power connectors to satisfy its 280W TDP.
According to Nvidia, the new Titan RTX will be available to professionals worldwide later this month with a price of $2,499.