That doesn’t imply will probably be all plain crusing for Google and its TPU prospects, although: Myron Xie, a analysis analyst at SemiAnalysis, warned that Google may also face constraints by way of chip manufacturing capability at Taiwan Semiconductor Manufacturing Firm (TSMC), which is going through bottlenecks round restricted capability for superior chip packaging.
Designed for TensorFlow
Ironwood is the seventh era of Google’s TPU platform, and was designed alongside TensorFlow, Google’s open-source machine studying framework.
That offers the chips an edge over GPUs basically for frequent in AI workloads constructed for TensorFlow, stated Omdia principal analyst Alexander Harrowell. Many AI fashions, particularly in analysis and enterprise situations, are constructed utilizing TensorFlow, he stated, and the TPUs are extremely optimized for such operations whereas general-purpose GPUs that help a number of frameworks aren’t as specialised.
