
General purpose SDK to accelerate parallelized HPC workloads
Esperanto Technologies™, the leading developer of high performance, energy-efficient AI systems based on the RISC-V instruction set, have introduced a new software development kit (SDK) that extends support beyond AI inference to a broader range of general purpose, high-performance computing (HPC) and mixed HPC plus AI use cases. The new SDK complements Esperanto’s existing AI-centric software stack to enable direct programming of each of the 1000+ 64-bit RISC-V CPUs that comprise the ET-SoC-1 inference chip and their accompanying vector/tensor units for massively parallel computation.
The ET-SoC-1 inference chip is designed to be the world’s highest performance RISC-V commercial chip, with over a thousand RISC-V processors on a single TSMC 7nm chip delivering a massively parallel, flexible architecture that combines exceptional performance and ultra-low power consumption.
“One of Esperanto’s key strengths in the AI acceleration market segment is that we use standard 64-bit RISC-V cores as the basis for our compute fabric, rather than overly specialized hardware. Many customers are keenly interested in directly programming these cores in addition to using our ML software stack to accelerate machine learning workloads,” said Art Swift, president and CEO at Esperanto Technologies. “We are making this SDK available so that customers can bring their own parallel workloads to ET-SoC-1 and take advantage of its 1000+ RISC-V cores for massive parallelism.”
Esperanto’s General Purpose SDK enables the direct programming of all of the ET-SoC-1 64-bit RISC-V cores and their attached vector/tensor units to accelerate mathematical computation. This makes ET-SoC-1 ideal for accelerating digital signal processing and other highly parallel workloads. Customers can also combine machine learning and general purpose workloads on the same ET-SoC-1 chip, enabling pre- and post-processing of AI workload data on the same device.
The General Purpose SDK is incorporated into Esperanto’s on-prem and Cloud Access evaluation systems, along with Esperanto’s current machine learning (ML) SDK, and a suite of performance analysis tools and pre-integrated models for fast experimentation, benchmarking and development. Esperanto evaluation servers deliver high performance combined with high energy efficiency and low total cost of ownership (TCO). Available in a standard 2U-high form factor, each Esperanto evaluation server includes dual Xeon host processors and either 8 or 16 ET-SoC-1 PCIe cards. Each Esperanto PCIe card has over 1,000 64-bit RISC-V CPUs with attached vector/tensor units, delivering up to over 16,000 RISC-V CPUs per server. Esperanto’s evaluation servers enable customers to obtain performance and power data from running a variety of industry standard AI models, including Generative AI, as well as the ability to bring their own models and data.