Nvidia reveals their Quadro GV100 Volta GPU
Volta, but not for gamers…
Published: 27th March 2018 | Source: Nvidia – USTREAM | Author: Mark Campbell
Nvidia reveals their Quadro GV100 Dual Volta GPU
Nvidia has made an incredibly confusing announcement at GTC 2018, unveiling their Quadro GV100 GPU. During this reveal, CES Jensen Huang spoke about this graphics card like it was a single unit, one with over 10,000 CUDA cores, 64GB of HBM2 memory and 236 TFLOPS of Tensor Core performance.
The reality of the situation is that Jensen was talking about two Quadro GV100 GPUs, linked together using NVLink2 technology, allowing each GPU to utilise the memory of a second Quadro GV100 GPU to create a 64GB frame buffer, the largest in the industry.
The Quadro GV100 will use the same GV100 GPU core as the Volta V100 and the Titan V, though this new GPU will make use of four 8GB HBM2 memory stacks to create a huge 32GB frame buffer per GPU, allowing this GPU to handle large datasets.
Below are the full specifications of the Quadro V100, which has a 250W TDP, 14.8TFLOPS of single-precision performance, 7.4TFLOPS of double precision performance and 118.5TFLOPs of Tensor performance. An insane level of GPU performance for a single graphics card.
This is the first Quadro GPU to make use of RTX technology, allowing the graphics card to accelerate raytracing workloads on the Nvidia OptiX API, using DirectX 12 DXR Raytracing or when using the Vulkan API.