Skip to main content

Tiny chip company takes on Intel, Arm with 'swiss-army knife' of AI processors

Image of SMIs Mozart chip on the Accelerando card
(Image credit: SMI)

To address the particular requirements of the rapidly advancing Artificial Intelligence (AI) software, chip-startup SimpleMachine, Inc (SMI) has designed a chip specifically for accelerating a wide variety of AI and machine learning apps.

The chip, christened Mozart, isn’t the first chip specially designed to handle AI workloads. However, SMI claims that unlike others that are designed for specific workloads, Mozart can adapt to a wide variety of AI workloads on the fly. 

Mozart is the culmination of 10 years of research at the University of Wisconsin-Madison led by the university’s Computer Science professor and SMI’s Founder, CEO and CTO, Karu Sankaralingam.

Bring on the orchestra

Karu reasons that Mozart is the result of his belief that chip design needed to evolve to meet the growing demand of AI across a wide spectrum of industries. 

It’s exactly because of this that SMI enables you to program Mozart to suit your AI use-case. The chip’s software interface includes direct TensorFlow support as well as API’s for C/C++ and Python. 

According to SMI, Mozart can accelerate a wide variety of AI applications from image detection and classification to graph processing and gene sequence alignment.

"The chip’s design can support very large models today and is capable of running up to 64 different models simultaneously," said Greg Wright, SMI’s Chief Architect. He adds that the next iteration of the chip that’s expected towards the end of 2021 will be 20 times faster than the current one.

One of the good things about Mozart that will help accelerate its adoption is that it can slot it in a standard PCIe slot. Additionally, you’ll also be able to leverage Mozart via SMI’s Symphony Cloud Service. However SMI hasn’t yet shared any pricing information about the chip or the cloud service.