Cloud INSIDER

World’s First Brain-Inspired AI Inferencing Chip Taped Out on the Azure Cloud

[fa icon="calendar"] Apr 7, 2021 1:15:48 PM / by Six Nines staff

An Ambitious Start-up Leans Heavy Into the Azure Cloud to Prototype a First-of-its-Kind Semiconductor

Startups with big, complex visions that require large amounts of computing muscle face significant challenges at their outset: “how do we build the environment needed to turn our vision into a reality?” Ambitious tasks, such as modeling a silicon-ready prototype for a new type of inferencing chip, are no small feat. Innovation like this requires HPC-level computing resources used by a team of people working in concert. With responsibilities and deadlines mapped out, everyone is racing against the clock to get the new technology built and out to the market as quickly as possible. Where computing is concerned, everything simply needs to work.

In the past, building an on-prem system was the only path forward, a difficult hurdle to clear for a startup. Many startups begin with visionary engineers who are experts in their professional fields, not building HPC (or any other IT) environments. It is a considerable hindrance to acquire the space, hardware, and power needed for an on-prem system – paperwork, hassles, permits, and setbacks became a big part of the process. Today, companies like Microsoft’s Azure provide cloud-based computing resources for entrepreneurs trying to build the next great chip, leaving behind the hassle of having to “roll-your-own” iron for the job.

Chip startup, d-Matrix, experienced first-hand the power of having the combination of the Azure compute cloud and an experienced partner in Six Nines to help them navigate their robust set of needs to operate and drive to results successfully. With critical needs of making a suite of Cadence design tools available across the company, ready to scale as the project progressively advanced without any downtime – and with controls for managing the costs, the team got to work.

Within three months of outlining their custom Cloud Adoption Framework, Six Nines turned over a full implementation to d-Matrix. This implementation included an entire IT infrastructure for building their first silicon, including Azure’s CycleCloud, an enterprise-friendly HPC cluster, and workload management tool for orchestrating and managing Azure HPC environments.

The results became a new milestone for both d-Matrix and Microsoft Azure. Within 14-months, d-Matrix was able to get their chip taped out - the world’s first brain-inspired AI Inferencing chip taped out in the Azure cloud.

“We have been executing a very aggressive schedule and it was clear very early that Six Nines was experienced and had the expertise to get us up and running very quickly. The project managers we worked with were thorough and diligent, providing us with a headlights view of the build and responding very quickly to any issues that cropped up. Our compute requirements far exceeded what we had planned, but with the help of Six Nines and Azure, we were able to get a rock solid, scalable environment up and running and taped out our first chip within 14 months of being founded.”

-Sudeep Bhoja, Co-founder and Technologist, d-Matrix

 

Learn more about this use HPC cloud use case:

 

Read the d-Matrix case study

 

Talk with one of our experts about moving your IT infrastructure to the cloud.

CONTACT US TODAY

Topics: HPC, EDA, Azure

Six Nines staff

Written by Six Nines staff

Subscribe to Email Updates

Recent Posts