Solutions

As a leading domestic discrete-GPU supplier, Glenfly leverages its self-developed, iteratively evolving HMUA architecture and HUCA software stack to achieve a full-stack technology loop spanning graphics rendering, general-purpose computing and AI computing, empowering the digital transformation of thousands of industries.

Rendering · Video

A stable, reliable graphics and video processing platform
Empowering office and industry terminals

For PC office, smart office, industrial design and industry terminals, Glenfly uses an IMR rendering pipeline plus fully in-house VPU/VPP/DPU/NoC to deliver stable, reliable multi-screen collaboration together with multi-stream video codec and post-processing capabilities.

IMR rendering and video-processing pipeline diagram

Requirements coverage

  • PC office & smart office
  • industrial design、Graphics rendering
  • Industry terminals such as finance, education and healthcare
  • Multi-screen collaboration, multi-stream video codec and post-processing

Technology base

  • IMR rendering pipeline
  • Glenfly HUCA software stack & Glenfly HMUA architecture
  • Fully in-house closed-loop core IPs (VPU/VPP/DPU/NoC)
  • Multi-generation mature mass-production iterations

Solution advantages

  • Full-stack in-house closed-loop development for autonomous, controllable core technology
  • WHQL-certified with broad adaptation to the domestic IT ecosystem and industry terminals
  • Continuous PPA optimization focused on efficient design rather than mere stacking
  • Ten-million-scale shipment validation with mature, stable and reliable quality

Rendering · Video · Inference

An efficiently integrated intelligent processing platform, empowering multimodal
Multi-model, agent and other multi-scenario intelligent applications

Integrating core capabilities in AI inference, multimodal generative AI, graphics rendering, and video codec, it delivers efficient, collaborative heterogeneous computing for AI PCs, smart office, and industry applications—empowering comprehensive advances in intelligent interaction, content generation, visual presentation, and multimedia processing.

Integrated rendering, video and AI-inference pipeline

Requirements coverage

  • AI graphics workstations、 Industrial design
  • Cloud desktop、 Cloud office、Cloud video、 Cloud gaming
  • Digital twins, GIS and similar scenarios
  • Multimodal generative AI

Technology base

  • Fully closed-loop and customizable processing cores
  • Matrix + Mem Fabric
  • Glenfly HUCA Software Stack
  • GPU virtualization

Solution advantages

  • Fully in-house multimodal AI processing cores
  • in-house Task Scheduler that significantly improves Agentic AI execution efficiency and performance
  • Proprietary ISA design
  • Hardware-software co-optimization with flexible adaptation to diverse mainstream ecosystems

Inference · Training

A highly scalable AI training and inference compute platform
Empowering multimodal and Agentic AI scenarios

For large-model training, multimodal inference and Agentic AI, it integrates Matrix + Mem Fabric with HMUA + HUCA hardware-software co-design to provide full-stack customized computing power spanning from the ISA to operator libraries and from die-to-die to chip-to-chip interconnects.

AI training and inference pipeline

Requirements coverage

  • Integrated AI training&inference appliances
  • Local multimodal multi-model deployment
  • Agentic AI appliances

Technology base

  • Matrix + Mem Fabric & GPGPU
  • Glenfly HMUA & Glenfly HUCA integrated hardware-software co-design
  • Ecosystem customization from ISA to operator libraries
  • Die-to-die and chip-to-chip interconnects

Solution advantages

  • Dedicated optimization for multimodal generative AI that improves generation efficiency and quality
  • Deep optimization for Agentic AI scenarios enabling higher task collaboration and execution capability
  • Continuous PPA refinement and reconstruction, replacing mere performance stacking with an efficient architecture

SoC Solution

A high-performance, low-power heterogeneous-computing SoC platform
Empowering future intelligent terminals

Integrating in-house, high-efficiency IP such as high-performance GPU, heterogeneous AI acceleration, streaming-media codec and on-chip interconnect, and supporting the WOA ecosystem with multi-task collaborative scheduling, it provides terminal devices with efficient computing, smooth interaction and a rich application experience.

SoC heterogeneous computing platform diagram

Requirements coverage

  • SoC Chips
  • NVR, NAS and AI edge-computing boxes
  • AIoT and industrial-control terminals
  • Highly integrated scenarios such as embedded platforms

Technology base

  • Powerful GPU core
  • High-efficiency Matrix core with a Matrix & NPU AI dual-engine design
  • Support for the WOA (Windows on ARM) ecosystem
  • An efficient interconnect and priority-scheduling mechanism built for graphics, video, display, multimodal AI inference and Agentic AI

Solution advantages

  • Low-power optimization balancing performance and energy efficiency
  • Support for the WOA (Windows on ARM) ecosystem
  • Highly integrated design achieving deep fusion of computing and communication capabilities
  • In-house, controllable underlying ISA for efficient ecosystem adaptation