もっと詳しく

The new chip, made of two die stuck together, will speed performance by 40% for machine learning tasks such as EfficientNet.