Latest1O1
Latest1O1
  • Home
  • Blog
Innosilicon Unveils Fenghua No.3 GPU, Aiming at AI, Gaming, and Beyond
Wookawooka |
24 Sep 2025

Innosilicon Unveils Fenghua No.3 GPU, Aiming at AI, Gaming, and Beyond

technology
Join us on Telegram

Innosilicon Technology may not rank among the top GPU makers, but the company continues to push forward with its Fenghua (Fantasy) graphics card lineup. Recently, it unveiled the Fenghua No.3, its latest flagship, marking a major step beyond the earlier Fenghua No.1 and No.2, which relied on Imagination Technologies’ PowerVR IP. Unlike its predecessors, the new model is built on the open-source RISC-V architecture, reportedly drawing inspiration from OpenCore Institute’s Nanhu V3 project.

Although details remain scarce, Innosilicon emphasized that Fenghua No.3 was designed in-house from the ground up. It’s also said to be compatible with Nvidia’s CUDA ecosystem, which—if true—could dramatically broaden its applications. Described as an “all-function GPU,” it targets multiple fields, from AI and scientific computing to CAD, medical imaging, and gaming, suggesting multiple versions may be in development.

For gaming, the card supports DirectX 12, Vulkan 1.2, OpenGL 4.6, and ray tracing. On the AI side, the card reportedly carries over 112GB of HBM memory, handling models up to 72B parameters on a single unit and scaling to 685B across eight cards. It also supports frameworks like DeepSeek V3 and Qwen 3.

Additional highlights include support for YUV444 color for professional-grade visual fidelity, native DICOM compatibility for medical imaging, and the ability to drive six 8K monitors at once. While China’s GPU industry still lags behind the U.S., innovations like the Fenghua No.3 reflect steady progress toward technological self-reliance.

#AI #hardware #latest101 #technology

Related Posts

Is Nvidia Really Stopping VRAM Supply to GPU Partners?

Is Nvidia Really Stopping VRAM Supply to GPU Partners?

Wookawooka
2025 Nov 27

A new rumor claims Nvidia may stop providing VRAM with its GPUs due to the global memory shortage, forcing AIB partners to source their …

OpenAI to Retire GPT-4o and Multiple Legacy ChatGPT Models in February 2026

OpenAI to Retire GPT-4o and Multiple Legacy ChatGPT Models in February 2026

Wookawooka
2026 Jan 30

OpenAI has announced it will discontinue several older ChatGPT models, including GPT-4o, as usage shifts heavily toward GPT-5.2 and newer systems.

Nvidia Invests $5 Billion in Intel to Power Next-Gen AI and PCs

Nvidia Invests $5 Billion in Intel to Power Next-Gen AI and PCs

Wookawooka
2025 Sep 19

Nvidia is investing $5 billion in Intel through a landmark partnership to develop custom chips for AI data centers and PCs — a move …

China’s Huawei Challenges Nvidia With AI Superclusters

China’s Huawei Challenges Nvidia With AI Superclusters

Wookawooka
2025 Sep 18

At its annual Shanghai event, Huawei revealed new AI infrastructure built on Ascend chips, with claims of beating Nvidia’s top systems by 2027.

Nvidia RTX 50 Cards Vanish as “Super” Refresh Rumors Heat Up

Nvidia RTX 50 Cards Vanish as “Super” Refresh Rumors Heat Up

Wookawooka
2025 Sep 13

Nvidia’s RTX 5000 Founders Edition is vanishing, hinting at a possible Super upgrade on the way.

Grokpedia: Elon Musk’s AI Challenge to Human Truth

Grokpedia: Elon Musk’s AI Challenge to Human Truth

Wookawooka
2025 Oct 28

Elon Musk’s xAI has launched Grokpedia, an AI-powered encyclopedia that generates and updates content in real time — aiming to challenge Wikipedia with a …

We'd Love to Hear from You!

All rights reserved ©Latest1O1|Privacy & Policy
created by Libya Digital Wizard |