NVIDIA launched its flagship GPU, Titan RTX, in an event yesterday. The high-end GPU comes with 72 Turing RT as compared to the previous value of 68 in RTX 2080Ti and features 4,608 CUDA cores.
NVIDIA Titan RTX has massive 24GB of GDDR6 VRAM which is twice the VRAM in RTX 2080Ti. The GPU has been priced at $2,500.
Owing to its high-end specifications, the latest GPU from NVIDIA could be used for AI research activities, playing graphics-rich games and in workstations. It also packs full-fat TU102 GPU similar to NVIDIA Quadro RTX 6000 which costs $6,300 and is aimed at workstation users.
The Titan RTX is also similar to RTX 6000 when the number of CUDA, RT Cores, Tensor, and VRAM are considered, and comes at much less price.
Users also have the option to join Quadro RTX 6000 and Titan RTX using Titan RTX NVLink Bridge which costs $80. The bridge allows you to join both the cards thus resulting in a card that has 100 GB/s bandwidth and 48 GB VRAM.
Titan RTX is perfect for AI researchers because of tensor cores that could be helpful in delivering real-time ray-traced graphics; its extra memory makes it ideal for running larger simulations than you could do with RTX 2080Ti that has 11 GB RAM.