DGX-1 is based on outdated Tesla P100, V100 is beast in Deep Learning. They've made tensor units especially for that part in V100 and wasted 30% of space in the SM )), I can't remember if they launched another DGX-1 made of V100. That'd be really confusing, could've named it DGX-2 or something
PS dammit https://www.nvidia.com/content/dam/en-zz/Solutions/Data-Center/dgx-1/NVIDIA-DGX-1-Volta-AI-Supercomputer-Datasheet.pdf they did exactly that! So which one it would be, P100 or V100? Sorry for G damn Quadros