Nvidia unveiled the Quadro RTX 4000 at the annual Autodesk University Conference today, making it the company's first midrange professional GPU powered by the Turing architecture. It’s the latest edition to the Quadro RTX family, which already consists of the Quadro RTX 8000, RTX 6000, and RTX 5000.
In addition to a single-slot design that can fit into a majority of chassis, the Quadro RTX 4000 boasts 40 percent more memory bandwidth than the previous Quadro P4000 generation and 36 RT cores for real-time ray tracing applications. It also has 2,304 CUDA cores, 288 Tensor cores for AI-enhanced environments, and special video encode and decode engines that can handle resolutions up to 8K.
The Quadro RTX 4000 boasts 7.1 TFLOPS of FP32 Performance and hardware support for VirturalLink, which is the new open industry standard for connecting VR headsets to PCs and other devices. For deep learning work, it can do around 57 TFLOPS of FP16.
The Quadro RTX 4000 uses the same TU106 core as the RTX 2070, and the same 8 GB GDDR6 chip and memory configuration, but Nvidia specs the card at 160W, which is lower than the RTX 2070’s 185W. The peak performance is also slightly lower thanks to reduced clockspeeds on the Quadro RTX 4000.
It’s also worth noting the current status of RTX availability, as the RTX 2070 is the most readily available and is generally between $500 and $600. RTX 2080 prices still have not hit their base MSRP of $699, and the RTX 2080 Ti is still in limited supply even at the higher Founders Edition price of $1,199.
Starting in December, the Quadro RTX 4000 will be available on Nvidia’s website and other major computer and hardware retailers at an estimated $900. If you happen to be in Las Vegas this week, you can try out the new GPU at the Autodesk conference.