California-based Esperanto Technologies announced on April 20 that initial evaluations for its ET-SoC-1 AI inference accelerator are underway with leading customers. The 1,000 core RISC-V processor is being eyed by industry giants like Samsung SDS, before its release.

What is an AI inference accelerator?

An AI accelerator is a special piece of hardware that helps accelerate artificial intelligence and machine learning applications. They help improve performance and speed up processes. AI accelerators are preferred by companies as they reduce the time taken to come up with a solution.

In February 2022, Esperanto had announced that they are entering into a strategic partnership with Intel to advance RISC-V-based AI acceleration solutions. Through the partnership, the inference accelerator producer will focus on delivering AI acceleration silicon solutions for everything from cloud to edge.

What is RISC-V used for?

The RISC-V processor is extremely helpful for smaller device manufacturers as it is based on reduced instruction set computer (RISC) principles. As it is provided under an open-source license, unlike most ISA designs, it is free to use and end users need not worry about paying royalties.

Esperanto Technologies is a leader in the RISC-V space and has been providing computing solutions for artificial learning and machine learning based on the RISC-V set architecture. The company recognizes that power consumption in high-performance computing devices is a major problem and aims to build energy-efficient solutions. The Mountain View-headquartered firm was founded by Dave Ditzel, who currently serves as the CTO and a board member.

RISC-V based devices are optimized for applications in various fields, including wearables, embedded systems, artificial intelligence, machine learning, virtual reality, and augmented reality.

Inference Accelerator

Esperanto Technologies’ new 1,000-core RISC-V processor has impressed customers.

As of today, the company is known for its 64-bit RISC-V-based tensor compute cores that are “delivered in the form of a single chip with 1088 ET-Minion compute cores and a shared high performance memory architecture.”

The ET-SoC-1 AI inference accelerator is made of 1,088 energy-efficient, 64-bit bit processor cores that utilize the RISC-V instruction set architecture. With its low-energy requirements, the ET-SoC-1 is well on its way to becoming an industry favorite. Art Swift, President and CEO of Esperanto Technologies, revealed that the company has plans to increase its RISC-V presence in the future and will partner with other companies for a variety of applications. The new chip contains four high-performance RISC-V cores and has interfaces for flash memory and external DRAM.

Esperanto is confident that the chip works well with machine learning recommendation models and mentions that it excels, especially, while working with hyperscalers, as it has been specifically designed to meet the needs of customers with large-scale datacenters.

Feedback and Future Plans

Dr. Patrick Bangert, vice president of Artificial Intelligence at Samsung SDS gave the new 1,000 core RISC-V processor his approval and stated that the company’s data science team was “very impressed with the initial evaluation.” He went on to add that, “it was fast, performant and overall easy to use. In addition, the SoC demonstrated near-linear performance scaling across different configurations of AI compute clusters. This is a capability that is quite unique, and one we have yet to see consistently delivered by established companies offering alternative solutions to Esperanto.”

Karl Freund, founder and principal analyst at Cambrian-AI Research too voiced his approval for the new system, saying that it is delightful that the firm has been able to provide “customers evaluation access to their RISC-V hardware and software running off-the-shelf AI models with strong performance and efficiency. This really shows the company’s confidence in their first multi-core solution.”