A few months after the launch of three Quadro RTX cards, Nvidia is now also serving the mid-market segment. The Nvidia Quadro RTX 4000 gets 8GB of GDDR6, 57 teraflops of deep learning performance and succeeds the Nvidia Quadro P4000.
At the annual Autodesk University Conference in Las Vegas, Nvidia showed the Quadro RTX 4000 to the general public. It’s the fourth graphics card in the Quadro series to use the Turing architecture released earlier in August. Then the following maps were launched:
- Nvidia Quadro RTX 8000: 10 gigarays/s, 4,608 CUDA cores, 576 Tensor cores, 48GB of GDDR6 memory
- Nvidia Quadro RTX 6000: 10 gigarays/s, 4,608 CUDA cores, 576 Tensor cores, 24GB of GDDR6 memory
- Nvidia Quadro RTX 5000: 8 gigarays/s, 3,702 CUDA cores, 384 Tensor cores, 16GB GDDR6 memory
Now Nvidia adds the Quadro RTX 4000 to the list that replaces the Quadro P4000. The card gets 8GB of GDDR6 memory on board and 288 Tensor cores for 57 teraflops of deep learning performance.
The Nvidia Quadro RTX 4000 also gets 36 RT (ray tracing) Cores on board, good for 6 gigarays/s. The graphics card will have 2,304 CUDA cores on board.
The graphics card uses Nvidia’s NVLink interface to connect multiple cards to each other. It also supports the new VirtualLink standard which allows VR headsets to be connected via a single USB-C cable.
The focus of the new Quadro RTX cards is on ray tracing, a complex technique for creating highly realistic lighting effects.
Ray tracing has so far been too demanding to render in real time, but not so much with the new Turing-gpus. They have RT Cores who are dedicated to the task of ray tracing with as little delay as possible. According to Nvidia, the Turing architecture can do this 25 times faster than the Pascal generation of GPUs (Quadro P-series).
The recommended retail price of the Nvidia Quadro RTX 4000 is $900. HP, Lenovo and Dell have already indicated that their professional workstations will soon be configurable with the new Quadro card.our launch article.