While NVIDIA is phasing out its Quadro line of GPU’s for new products, we thought we would look at one Turing-based GPU with enormous potential that we have not looked at yet. In this review, we are going to show you around the NVIDIA Quadro RTX 4000, which is a low power single slot GPU. We are then going to subject the Quadro RTX 4000 to our battery of compute-focused GPU tasks. Let us take a look at what the Quadro RTX 4000 has to offer.
NVIDIA Quadro RTX 4000 Overview
NVIDIA Quadro RTX 4000 is a single-slot GPU with a length of 9.5″ long. The outer case is very stylish with a shiny chrome covering and a bright green band across the middle. It is an impressive looking graphics card, be warned, it will attract fingerprints on the chrome surfaces.
While several of the Quadro RTX GPUs include a backplate, which we like, the Quadro RTX 4000 does not. This would add some structural rigidity and help protect the components mount on the backside of this $900 GPU.
At the end of the Quadro RTX 4000, we find the video outputs, 3x DisplayPort 1.4, and 1x VirtualLink port for VR head-mounted displays.
The power connection is located at the end of the Quadro RTX 4000, which should be more comfortable for server mountings, 1x 8-Pin connection.
The top of the single-slot cooler card has stereo and sync ports for professional workstations, but there are no NVLink/ SLI connectors.
We wanted to take a moment and focus on why this single-slot configuration is going to be interesting to our readers. Aside from the typical professional workstations where a single-slot cooler can mean a smaller workstation, there is also an opportunity to maximize density. An example uses a Gigabyte C621 AORUS Xtreme motherboard designed for Xeon CPUs. Here, a total of seven Quadro RTX 4000 GPUs can be installed. For AI workloads that can scale to multiple GPUs better than scaling up in a single GPU, this is a great example of what can be accomplished using an off-the-shelf motherboard.
NVIDIA offers much larger GPUs but the combination of the single-slot form factor and the blower-style cooler makes scaling to multiple GPUs more feasible.