PowerColor Radeon Cards Featuring Integrated AI NPU To Consume Up To 20% Less Power

At Computex, PowerColor didn’t unveil any fresh graphics card designs. However, the company did introduce its latest concept—a consumer Radeon model reportedly featuring an AI processor.

The NPU, short for Neural Processing Unit, is a term every PC enthusiast should now know. It’s set to become a standard component in most next-gen CPUs, with a few exceptions in the enthusiast segment. Mobile platforms, in particular, will lead in NPU adoption, especially as Microsoft rolls out its Copilot+PC systems this summer.


In essence, graphics cards serve as expansive NPUs capable of executing calculations much more rapidly. This led NVIDIA to develop their GeForce RTX AI PC program, endorsing systems featuring GeForce GPUs that significantly outpace integrated NPUs.

Conversely, PowerColor has a distinct approach and objective for incorporating NPUs into GPUs. They’ve embedded a bespoke NPU chip directly onto the graphics card as a separate component. This chip appears capable of regulating the GPU’s power consumption and fan profiles, leading to reduced power usage.


PowerColor employs an NPU chip from Kneron, which has affirmed its role in reducing GPU output. Essentially, the chip seems to dynamically adjust the power limit, leading to decreased consumption. The “breakthrough” PowerColor highlights resembles the benefits gamers can attain through software technologies like Radeon Chill and Radeon Boost, which aim to curb power usage during demanding gameplay sessions.

Nevertheless, PowerColor’s approach functions on a hardware level, taking into account additional factors like GPU temperatures and fan speed regulation. What becomes evident is that the Kneron AI chip appears to offer two modes, namely ECO or Boost mode, presumably manageable through software controls.


During the PowerColor demonstration, real-world testing revealed a notable decrease in power consumption. The PowerColor/Kneron AI chip seems to execute immediate adjustments to GPU power configurations and fan velocities. While this approach could potentially prolong the GPU’s lifespan, it does impact performance and leads to elevated temperatures. It appears that comparable functionality could be incorporated into software without necessitating an NPU.

Source & Images: Kneron, ITMedia