With the advent of specialized computation accelerators Google (TPU), NVIDIA has to contrast its products not only with Intel profile products , and the developer's guide to the GPU perceives the increase in the number of competitors calmly . At the quarterly report conference, which ended last Friday, the founder of NVIDIA once again had to oppose the TPU and the GPU.
The accelerator of tensor calculations (TPU), in the opinion of the general director of NVIDIA, is an integrated circuit for a special purpose, ASIC. Graphical same processors NVIDIA are able to perform not only specialized calculations, but also adapt flexibly to a wide range of tasks. Pascal and Volta can significantly speed up the work in the environment of artificial intelligence systems, and they cope with this much better than any ASIC. The cloud system, according to the head of NVIDIA. In the cloud system, a huge number of applications are used, which intensively process data. NVIDIA's GPUs are good not only as TPUs, but also excel at all the tasks the data centers deal with.
Related Products :
|