This is similar to how Nvidia tapped the Volta-based V100 server GPU for the GV100 workstation ... versus GV100’s 870 GB/s, 6,912 CUDA cores versus GV100’s 5,120 CUDA cores and 400 GB/s ...
With the A100’s ability to perform inference and training workloads faster than Nvidia's T4 and V100 ... Cores, 6,912 FP32 CUDA Cores, 432 Tensor Cores, 108 streaming multiprocessors and 40 GB ...
New Vega series IP cores to set the new standard for low power GHz GPU performance, API completeness and silicon cost savings in mainstream SOCs over previous generation cores San Francisco, ...