In this fast paced technological age we live in it does feel like there is a new latest and greatest piece of tech being released every month. But the NVIDIA Quadro GV100 is really one that professionals should stop and take note of. NVIDIA have finally brought their latest Volta GPU architecture to the Quadro Workstation graphics range and the improvements over the previous Pascal generation are substantial. Now with capabilities for real-time ray tracing, advanced Artificial Intelligence, Simulation and VR capabilities, industries such as 3D Content Designers, Architects and Scientific Researchers are in for a treat.
Volta GPU Architecture
Continuing the trend of smaller is better, the new Volta architecture is based on the new 12nm FFN (FinFET NVIDIA) high-performance manufacturing process that has been specially customised for NVIDIA. What this has enabled them to do is produce a GPU that packs an impressive 5120 CUDA cores onto a single chip which is 1536 more than the previous generation NVIDIA Quadro GP100. This has enabled the NVIDIA Quadro GV100 to take the mantle as the most powerful computing platform in the world today.
With regards to deep learning there are two main categories of software, those performing training tasks and those performing inference. These processes historically have required different GPU hardware to get the most out of your software, but with the introduction of NVIDIA Quadro GV100 including 640 state-of-the-art Tensor cores, this is being brought to a single GPU. The Tensor cores are new mixed precision technology which is purposely designed and built for Deep Learning and these produce up to 8x performance gains on previous generations in regards to neural network training. This is not surprising, as each Tensor Core performs 64 floating point fused multiply-add (FMA) operations per clock, and each Streaming Multiprocessor (SM) performs a total of 1024 individual floating point operations per clock.
High Bandwidth Memory (HBM2)
Currently unique to the NVIDIA Quadro GV100 is the impressive 32GB HBM2 memory on board as standard. HBM2 is cutting edge High Bandwidth Memory that is highly optimised to deliver incredible speeds, making it the industry’s fastest graphics memory (870 GB/s peak bandwidth). Also, with 32GB it boasts huge capacity, doubling that of its data centre sever based counterpart, the Tesla V100. This makes the NVIDIA Quadro GV100 ideal for those working with low latency sensitive applications and handling large data sets. On top of all this it supports Error Correcting Code (ECC) without compromising on performance, making sure your GPU based calculations are extremely accurate.
NVDIA Quadro GV100 – Who Will Benefit?
With the levels of performance and flexibility this GPU provides many professionals from industries across the board will see impressive performance gains. Those working with big data crunching workflows, such as Artificial Intelligence / Deep Learning and Scientific Simulation researchers will fly through processing times. Architects will be able to design ever bigger and more intricate designs that were previously only possible in their imaginations. Whilst artists will be able to ray trace render even more complex photorealistic scenes in seconds, when it once took hours.