Esperanto Technologies Begins Trailing 1,093-Core ET-SoC-1 RISC-V Machine Learning Accelerator

Initial results from the company's early access program promising — and you can apply as a member now to receive a test chip of your own.

Esperanto Technologies has announced that its massively-parallel RISC-V artificial intelligence accelerator technology, which packs over 1,000 cores into a single chip, is now sampling for evaluation with unnamed "lead customers" — meaning general availability is not too far in the future.

"Harnessing the power of over 1,000 RISC-V processors is a major accomplishment, and we are very pleased with the results which validate our initial projections of performance and efficiency," says Art Swift, president and chief of Esperanto Technologies, of preliminary results from the company's early access program. "We look forward to extending access to a broader range of qualified companies, as we accelerate our RISC-V roadmap efforts with a growing number of strategic partners for applications spanning from Cloud to Edge."

Built atop the free and open-source RISC-V instruction set architecture, the ET-SoC-1 AI Inference Accelerator is a single-chip 7nm device packing 1,088 64-bit ET-Minion in-order cores with vector/tensor units as inference accelerators alongside four ET-Maxion out-of-order cores for running an operating system plus a single "service processor" core — making it an all-in-one system-on-chip, if so desired, as well as an accelerator.

The company expects the part to be the "highest-performance commercial RISC-V AI chip" when it launches, and initial feedback looks promising. "Our data science team was very impressed with the initial evaluation of Esperanto's AI acceleration solution," says Patrick Bangert, PhD, vice president of artificial intelligence at Samsung SDS and one of the early access customers trialing the part.

The company claims its technology allows for highly efficient acceleration, providing the workload is parallelizable. (📹: RISC-V International/Esperanto Technologies)

"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."

More information on the ET-SoC-1, pricing for which has yet to be announced, is available on the Esperanto Technologies website — along with an application form to apply for consideration as a recipient of the next early access program shipment, which requires payment to be credited against initial product orders at general availability.

Gareth Halfacree
Freelance journalist, technical author, hacker, tinkerer, erstwhile sysadmin. For hire: freelance@halfacree.co.uk.
Latest articles
Sponsored articles
Related articles
Latest articles
Read more
Related articles