In an effort to make their Azure cloud computing platform more attractive to AI developers, Microsoft has installed new Graphcore IPUs (Intelligence Processing Units).
The processors are similar to GPUs but optimised for AI development rather than game playing, offering 3 times the performance than other comparable solutions.
This would allow AI developers to train models on large pools of data much faster and cheaper.
Graphcore, in which Microsoft holds an investment, says their IPU offered 1.6 petaflops of machine intelligence processing using eight Graphcore C2 IPU PCIe cards, each with two IPU processors and connected via high-speed Graphcore’s IPU-Link technology running on a standard 4U chassis.
The stack currently supports Tensorflow and will fully support PyTorch by the end of 2020.
“We are now extremely pleased that we are making Graphcore technology commercially available to a wider group of customers,” said Co-founder and CEO Nigel Toon. “We are looking forward to supporting innovators achieve the next great breakthroughs in machine intelligence on IPUs.”
In the latest round of funding, Graphcore raised $200m from investors including Microsoft and BMW, valuing the company at $1.5bn. The round was led by investment funds like Merian Global Investors, Amadeus Capital Partners, Atomico, Robert Bosch Venture Capital, Atomico and Sofina. The company also has received investments from some big names in the AI industry including Demis Hassabis, co-founder of Google’s DeepMind, and Zoubin Ghahramani, chief scientist at Uber.