SimpleMachines, Inc. (SMI) has created a programmable, high-performance chip called Mozart. The TSMC 16-nanometer design utilizes HBM2 memory and is sampling as a standard PCIe card.
According to the company, initial results have shown significant speedups across a broad set of AI applications unlike other specialized AI chips. Examples include recommendation engines, speech and language processing, and image detection, all of which can run simultaneously.
Mozart is the result of 10 years of research at the University of Wisconsin-Madison, led by SMI’s Founder, CEO and CTO Karu Sankaralingam, also a computer science professor at UW.
The working chip demonstrates that a clean-slate design can deliver the performance of custom-built single purpose processors while maintaining flexibility and programmability.
The chip’s software interface includes direct TensorFlow support as well as API’s for C/C++ and Python. Future generations of the chip will leverage its architecture to scale up and down the power spectrum from enterprise-class high-performance systems to 5-watt IoT devices.
“Mozart is an extremely complex chip, one of the few using HBM2 (high-bandwidth memory) for this type of application,” Sankaralingam said. “It took our silicon team under a week from having the chip in-house to running applications on the device, putting us on the fast track to take our first silicon (A0) to production.”
“The chip’s design can support very large models today and is capable of running up to 64 different models simultaneously,” added Greg Wright, SMI’s Chief Architect. “Our next-generation 7-nanometer design is expected to be ready to sample by the end of 2021 and will be 20x faster on a diverse set of workloads than current chips.”
Mozart will be available via a PCIe card called Accelerando, or via the Symphony Cloud Service (SMI’s hosted cloud service with easy access to public clouds like Azure, Google Cloud Platform, and AWS).