Nvidia has introduced a new top GPU model, the Titan "T-Rex" RTX. It is the most powerful graphics card so far and has a massive 24 GB DDR6 RAM - twice as much as older models of the Titan series. It uses a fully developed Turing graphics chip and is said to be up to 15% faster than the RTX 2080 Ti. The Titan RTX should have enough memory and be powerful enough to work with 8K video in real time.
The central graphics unit TU102-GPU uses 4608 CUDA cores, which run at 1350 MHz in basic mode and up to 1770 MHz in turbo mode. The total memory bandwidth of the 24 G B DDR6 memory is 672 GB/s. Nvidia states the thermal design power of the Titan RTX at 280 watts. With its specialized 576 tensor cores, Nvidia particularly targets complex deep learning applications. Data analysis is accelerated by the use of the open source RAPIDS libraries, which can be integrated into the most common data science workflows to accelerate machine learning.
Three DisplayPorts 1.4, one HDMI 2.0b and one USB Type-C port are provided for interfaces - up to four displays can be recorded simultaneously. Two Titan RTX GPUs can be interconnected via 100 GB/s NVIDIA NVLink for even more memory and performance.
However, the price for the top performance is also enormous: the Titan RTX costs a whopping 2,700 Euros and is thus more than twice as expensive as the RTX 2080 Ti, which is currently available from around 1,200 Euros. The Deep Learning units are mainly responsible for the high price.
The acquisition of the Titan RX for video editing will rarely be worthwhile - only if applications can use the full GPU power and it really is on the additional 15% performance - in all other cases an RTX 2080 Ti or other, much cheaper and somewhat weaker model will suffice. But of course we are looking forward to first practice benchmarks in the video area.