Current thinking about the key hardware specs which could be on offer with the RTX Titan are; a full Turing TU102 GPU (with 4,608 CUDA cores, 288 TMUs, 96 ROPs, plus 576 Tensor cores and 72 RT cores), with 12GB of GDDR6 memory. If you are a gamer, all you need to know that this is a slightly faster RTX 2080 Ti with a shit-ton of VRAM which you probably won't need.
Whether your workloads consist of AI, real-time ray-traced graphics, next-gen virtual reality and high performance computing, Nvidia is positioning the TITAN RTX as the flawless solution. Huang added, more specifically, that "The introduction of T-Rex puts Turing within reach of millions of the most demanding PC users - developers, scientists and content creators". Statistically, this allows the card to deliver up to 130 teraflops of deep learning performance, and up to 11 GigaRays of real-time ray-tracing performance per second.
24GB of high-speed GDDR6 memory with 672GB/s of bandwidth, twice the memory of previous-generation TITAN GPUs, to fit larger models and datasets. Of particular interest for data scientists will be the fact that the Titan RTX supports RAPIDS, the open-source libraries that integrate with popular data science workflows to boost machine learning use cases.
NVIDIA will be launching the TITAN RTX - which company CEO Jensen Huang refers to as "T-Rex" in today's announcement - later this month in both the U.S. and Europe.
Incredible performance and memory bandwidth for real-time 8K video editing. All of the Titan RTX's professional attributes also commands a professional pricetag: it is to be made available in Europe and the USA for US$2,499 later in December.