Advertisement

NVIDIA unveils the Quadro RTX 4000, its mainstream workstation GPU

You pay a premium over the RTX 2070 equivalent, but less than before.

NVIDIA has unveiled the Quadro RTX 4000, a workstation and deep learning version of its GeForce RTX 2070 gaming GPU. Like that model, it packs 2,304 NVIDIA CUDA cores, 288 Turing Tensor Cores for AI and 8GB of GDDR6 graphics memory, but has fewer ray-tracing (RT) cores with 36 rather than 42. It also uses slightly less power (150 watts) compared to the RTX 2070's 185 watts, likely because of reduced clocks speeds. It's equipped with 8K video decoding and encoding capability for multiple professional formats, and connects to VR headsets by VirtuaLink.

At $900, it's a lot cheaper than its Quadro RTX 5000, 6000 and 8000 stablemates, which range in price from $2,300 to $10,000. Still, the price is nearly double that of a $500 GeForce RTX 2070, so what do you get for that? Mostly, these cards are designed to be glitch-free, especially for OpenGL, when viewing or rendering mission critical graphics in apps from Autodesk, Adobe and others. With gaming cards, a minor display error when rendering a scene usually isn't a serious problem.

The Quadro RTX 4000 will be available from NVIDIA in December, and at other sites and retailers for an estimated $900. NVIDIA is showing it off for the first time this weekend in Las Vegas at the Autodesk conference.