AI Edge Computing Solutions

Firefly's Hardware


  • CSA1-N8S1684X

    140T@INT8 peak computing power, 17.6TFLOPS@FP32 high-precision computing power, support for deep learning frameworks.

    Know More
  • CSA1-N8S1684

    256TOPS (INT8) peak computing power, 128TFLOPS (FP16/BF16) computing power, 16TFLOPS (FP32) high-precision computing power, support for deep learning frameworks.

    Know More

Advantages


  • Easy Development

    Open SDK; each computing unit has a hardware DEBUG port and supports bootloader-level debugging.

  • Powerful AI Performance

    It has a dedicated AI acceleration engine at the hardware level and supports multiple learning frameworks, such as PyTorch, MXNet, PaddlePaddle, ONNX, and Darknet, at the software level. It supports ultra-large-scale parameter models, such as Gemma-2B, Qwen1.5-1.8B, Llama2-7B, ChatGLM3-6B, and other private deployments of large language models.

  • Efficient Management

    The integrated aBMC management system can easily monitor array servers, simplify configuration, provide abnormal warnings, and perform remote operation and maintenance (such as software KVM).

  • Low Cost

    The server highly integrates computing units, storage, universal serial bus, network controller, power controller and sensor into one, reducing users' procurement, development and operation costs.

Application Scenario

Business Customization

Contact Info


Phone: +86 18688117175

Address: Room 2102, Hongyu Building, #57 Zhongshan 4Rd, East District, Zhongshan, Guangdong Province, China

Email: global@t-firefly.com

Follow us