It was at GTC is where Nvidia unveiled and announced their newest graphic card, the Quadra GV100 powered by Volta graphics processor, the newest addition to the Quadra lineup.
Not your run of the mill consumer base graphics card as its GV100 GPU is made solely for workstation level applications. Consumers can still go with current Titan V which can still handle the same resource intense workload. But the Quadra drivers gains a big advantage in performance over the Titan V.
The Volta powered graphics card according NVIDIA is suppose to provide a special professional driver code path for applications that includes CATIA, Solidedge, and much more.
One word describes the Quadro GV100 overall specs, beast, the better word would be astounding. It has 32GB of HBM2 ( high bandwidth memory) on board, which is scalable up to 64GB addressable frame buffer using a two-way NVLink interconnect technology ( we'll closely examine NVLink technology later on in this post), basically what you have here is two GPUs combined onto one graphics card. When it comes to performance, it falls nothing short of being impressive, the Quadro GV100 produces 7.4 teraflops of precision performance or 14.8 teraflops of a single performance. Deep learning task can enable Teraflops to balloon up to a scary 118.5. There are a total of 10,240 CUDA cores ( yes you read right, 10,000) and 236 TFLOPS Tensor cores.
The standout technology within this graphic card is NVLink. What is NVLink you ask? Its a new wired based semiconductor technology created by NVIDIA, its main purpose is to code transfer and data control a processor system between the CPU and GPU. NVLink puts together point-to-point connection with measured data rates of about 20 and 25 Gbit/s. NVLink primary goal is to focus on high performance,data transfer between lanes will be more efficient and faster than the PCIe that's currently in your computer.
Earlier I mentioned that the Quadra GV100 is not for the average PC builder who's looking to just game or partake in mundane computer related task, the $9,000 price point most certainly dictate it's market target. It's mainly for users looking for powerful workstation that can handle complex 3D rendering and science applications. The standout feature for me is the NVIDIA RTX Technology, developers can now put Quadro GV100 to full use with real time ray tracing. The technology (Ray Tracing) will take 3D applications to a whole new level.
Owning a graphics card like this of its powerful magnitude brings about many benefits. You can implement deep learning development with ease, a supercharge more fluid rendering performance, a much faster double precision combine with memory scaling up to 64GB and we certainly can't leave out its VR capability which features a massive on-board memory capacity.
The Quadro vDWS is now readily available for over 120 systems from 33 vendors. The NVIDIA Quadro GV100 discrete graphics card can now be purchased immediately on NVIDIA's website.