Leading the RISC-V revolution for AI and HPC
with the industry’s most efficient products.
with the industry’s most efficient products.
Our AI/HPC compute servers offer the highest level of RISC-V parallelism and core density in the industry, enabling best-in-class TCO. Our ET-SoC-1 inference chip is designed to be the world’s most efficient RISC-V commercial chip, delivering a massively parallel, flexible architecture that combines exceptional performance with ultra-low power consumption
THE ESPERANTO GENERATIVE AI APPLIANCE
Say hello to the industry’s first Generative AI Appliance
based on RISC-V technology
based on RISC-V technology
Deploy fine-tuned business applications based on Generative AI
quickly, securely and with low TCO.
quickly, securely and with low TCO.
The Esperanto Generative AI Appliance is a full solution including hardware, LLM software stack and the latest foundation models, ready for fine-tuning to the task at hand.
Features:
- Currently running the latest LLMs and diffusion models such as LLaMA 2, Vicuna, StarCoder, OpenJourney and Stable Diffusion
- Includes 4 ET-SoC-1 cards running the Esperanto Generative AI software stack
- Lowest total cost of ownership (TCO) for Generative AI inferencing – much lower than GPU-based systems
- Provides much stronger data privacy than cloud-based alternatives
- Complete system ready to be fine-tuned to a specific task in a single day
THE ESPERANTO AI / HPC COMPUTE SERVER
Meet the world’s most efficient solution for AI inference,
HPC, and mixed mode workloads.
HPC, and mixed mode workloads.
Access up to 16,000 RISC-V processors for exceptional compute efficiency.
The Esperanto AI /HPC compute server features up to 16,000 RISC-V processors.
Features:
- 8 or 16 ET-SoC-1 PCIe cards, each with more than 1,000 RISC-V compute cores
- 2 Intel Xeon® Gold 6326 16-core or Xeon Platinum 8358P 32-core host processors
- Up to 1 Terabyte of DDR4-3200 main memory
- Esperanto Machine Learning and General Purpose Software Development Kits
- Dozens of Generative AI, Transformer, Computer Vision and Recommendation System AI Models
- Standard 2U 19” rack-mounted chassis for the datacenter environment
THE ESPERANTO ET-SoC-1 PCIe CARD
Developed jointly with our strategic partner, Penguin Solutions,
for production quality, interoperability and international certification.
for production quality, interoperability and international certification.
Leading the industry in performance efficiency.
The ET-SoC-1 PCIe card delivers the compute performance and efficiency of one ET-SoC-1 chip in a compact, low-profile PCIe Gen 4 form factor.
Features:
- Single ET-SoC-1 chip on compact, low-profile PCIe Gen 4 card
- 32 Gigabytes of LPDDR4x DRAM
- Full set of monitoring sensors & physical management ports
- Driven by Esperanto Machine Learning and General Purpose Software Development Kits
- Up to 16 cards per system
- Jointly developed, tested and certified by Penguin Solutions
- Safety and emissions certifications for North America, European Union, United Kingdom, South Korea and Japan
THE ET-SoC-1 AI / HPC ACCELERATOR CHIP
Designed to be the world’s most efficient RISC-V commercial chip.
Industry-leading energy efficiency across a range
of AI, HPC and mixed mode workloads.
of AI, HPC and mixed mode workloads.
The ET-SoC-1 chip features over one thousand RISC-V processors on a single 7nm chip.
Features:
- 1088 energy-efficient ET-Minion 64-bit RISC-V in-order cores, each with a custom vector/tensor unit optimized for ML applications
- 4 high-performance ET-Maxion 64-bit RISC-V out of-order cores for running an OS in self-hosted mode
- Over 160 million bytes of on-chip SRAM
- Interfaces for large external memory with low-power LPDDR4x DRAM and eMMC FLASH
- PCIe Gen4 x8 and other common I/O interfaces
Request Access to
Esperanto Systems
Esperanto Systems