OpenAI Currently Not Adopting Google TPU Chips; NVIDIA and AMD to Continue as Key Suppliers
4 day ago / Read about 0 minute
Author:小编   

OpenAI is currently conducting initial tests on Google's Tensor Processing Units (TPUs) but does not intend to deploy them on a large scale at this time. The AI firm, which is supported by Microsoft, is experimenting with TPUs from Alphabet. These TPUs are custom ASIC chips developed by Google exclusively for machine learning tasks, designed to speed up the training and inference of neural networks. Utilizing a dataflow-driven architecture, these chips efficiently minimize memory access latency through streamlined matrix multiplication pipelining. Despite these preliminary tests, OpenAI does not have any immediate plans for a large-scale deployment of TPUs.