Esperanto is leading the RISC-V revolution for AI, and helping create an exciting new category for AI solutions

For designers who need the most energy-efficient, high-performance computing solutions for Artificial Intelligence inference applications, Esperanto Technologies delivers technology based on the open RISC-V instruction set architecture (ISA).

Industry-Leading Performance Per Watt For Key AI Workloads

The ET-SoC-1 ML inference chip is designed to be the world’s highest performance RISC-V commercial chip

The historical limitations on many-core designs have always been complexity and power consumption, but Esperanto’s new generation of AI inference processing solutions are designed to deliver unmatched energy efficiency (performance/watt), scalability, and flexibility for key AI workloads.

Esperanto’s no-compromise solution breaks through previous barriers in delivering a massively parallel, flexible processor that combines high performance and ultra-low power consumption. Our high-performance ET-Maxion™ core is designed to deliver the best single-thread RISC-V performance. Our ET-Minion™ core compute array is designed for energy efficiency while delivering hundreds of TeraOps of computing performance.

The ET-SoC-1 has over a thousand RISC-V processors on a single 7nm chip. Additional details include:

  • 1088 energy-efficient ET-Minion 64-bit RISC-V in-order cores, each with a vector/tensor unit
  • 4 high-performance ET-Maxion 64-bit RISC-V out-of-order cores
  • Over 160 million bytes of on-chip SRAM
  • Interfaces for large external memory with low-power LPDDR4x DRAM and eMMC FLASH
  • PCIe x8 Gen4 and other common I/O interfaces
  • An innovative low-power architecture and circuit techniques supporting peak compute rates of 100 to 200 TOPS and operation at under 20 watts for ML recommendation workloads
Chip image with DDRs

Incorporating RISC-V, today’s most advanced and open instruction set architecture, together with a comprehensive low-power design approach, Esperanto is delivering a breakthrough value proposition for AI inference applications.

Industry-Leading Energy Efficiency

Energy efficiency for a range of datacenter inference workloads

Power consumption continues to be the most critical problem facing high-performance computing today.  Our approach employs an array of over a thousand of our low-power ET-Minion cores, delivering energy efficiency with high integer and floating-point throughput, including tensor and vector acceleration optimized for ML / DL workloads. Our distributed memory architecture puts on-chip memory where it’s needed, improving processing utilization and relieving memory bandwidth bottlenecks.

PCIe card with 6 ET-SOC-1 Chips

Esperanto’s ET-Minion™ on-chip RISC-V cores

The ET-Minion core, based on the open RISC-V ISA, adds proprietary extensions optimized for machine learning. This general-purpose 64-bit microprocessor executes instructions in order, for maximum efficiency, while extensions support vector and tensor operations on up to 256 bits of floating-point data (using 16-bit or 32-bit operands) or 512 bits of integer data (using 8-bit operands) per clock period. Multiply-accumulate operations always work with 32-bit accumulators to avoid loss of precision for large arrays, a real problem for competing products that accumulate to smaller representations. The longest tensor operations execute for 512 cycles without consuming memory bandwidth or power to fetch and decode instructions. These operations are perfectly matched to the execution units; both run at full speed for the full duration of each operation.

Esperanto’s ET-Maxion™ on-chip RISC-V cores

The ET-Maxion, our custom 64-bit single-thread RISC-V microprocessor core, implements advanced features such as quad-issue out-of-order execution, branch prediction, and sophisticated prefetching algorithms to deliver high single-thread performance. It runs the Linux operating system (and others) along with complex applications. Four ET-Maxion cores are on each chip, along with fully coherent cache memory. ET-Maxion cores are ideal for scheduling tasks across the array of ET-Minion cores, managing data movement and hosting operating systems. In an accelerator configuration, this offloads host processing tasks, delivering better overall system performance per dollar.

High-performance on-chip interconnect

Esperanto’s mesh-style network on chip (NoC) was specified to match the requirements of advanced workloads. It distributes operands and results between cores, caches, memory, and I/O. It includes sophisticated debug features so software developers—ours, and yours—can get a high-level view of data movement and simultaneously drill down to uncover and relieve bottlenecks at any point on the network.

PCIe card with 6 ET-SOC-1 Chips

On-chip management and security

Even in a hosted configuration, it is useful to have an on-die management processor. Ours, based on an ET-Minion core, connects to an on-die hardware “root of trust” subsystem to prevent unauthorized or corrupted firmware updates and ensure you can always track activity inside the chip. These elements communicate through the PCI Express interface or secure local interfaces in both hosted and self-hosted configurations.

Scalable, flexible architecture

The Esperanto RISC-V architecture is highly scalable to address different applications, different levels of performance, different power profiles, and varying system form factors. This flexibility delivers optimal energy efficiency for each target application with a family of architecturally consistent products. With over a thousand high-performance, energy-efficient 64-bit RISC-V cores on one chip, Esperanto can deliver TeraFlops of scalable compute performance.

Easy software development

Programming sophisticated AI chips is historically difficult and time-consuming. Fortunately, Esperanto has already done the heavy lifting for the major ML frameworks. Our host-based software development tools are designed to convert your trained model into an optimized package of code and data that will run efficiently on our architecture.

Join The Esperanto Early Access Program

Esperanto’s Early Access Program for qualified customers will begin later in 2021.

For information, contact: [email protected]

RISC-V Foundation

RISC-V (pronounced “risk-five”) is a free and open ISA enabling a new era of processor innovation through open standard collaboration. Founded in 2015, the RISC-V Foundation comprises more than 100 member organizations building the first open, collaborative community of software and hardware innovators powering innovation at the edge forward. Born in academia and research, the RISC-V ISA delivers a new level of free, extensible software and hardware freedom on architecture, paving the way for the next 50 years of computing design and innovation. Visit https://riscv.org/

Latest Esperanto News