Forecasting the future of AI + High Performance Computing

Graphcore engineers took one of the European Centre for Medium-Range Weather Forecasting's publicly available forecasting models – a Multi-Layer Perceptron (MLP) – and accelerated it on Graphcore IPU-POD systems with dramatic results. The IPU-POD system was shown to train the ECMWF’s predictive MLP model 5x faster than a leading GPU and a massive 50 times faster than ECMWF’s existing simulation methods running on a CPU.