PowerColor has introduced a pioneering idea at Computex for a consumer Radeon model that integrates an AI processor. The company’s distinct strategy involves embedding a Neural Processing Unit (NPU) into the graphics card to optimize energy consumption.

Unlike Nvidia’s AI PC scheme with GeForce RTX that promotes systems using GPUs that surpass integrated NPUs, PowerColor’s NPU chip (Edge AI) aims to regulate GPU power output in real-time. By utilizing the NPU chip, PowerColor’s graphics cards from the Hellhound and Red Devil series could potentially achieve a reduction of up to 22% in energy consumption while maintaining high performance, as indicated by @harukaze5719.

Image credit: Quasar Zone

The NPU introduces two modes: Eco and Boost. In Eco mode, the graphics card actively monitors changes in GPU temperature and adjusts fan speeds dynamically to reduce energy consumption. Conversely, Boost mode adjusts GPU power usage and heat dissipation dynamically to ensure peak performance and stability.

During the demonstration by PowerColor, real-world testing images shared by ITMedia (via VideoCardz) displayed a significant decrease in energy consumption with the NPU activated (from 338W to 261W). Nevertheless, this led to an increase in temperatures from 51ºC to 61ºC due to reduced fan speed (from 1918 RPM to 1037 RPM). Additionally, this seemed to come at the cost of decreased performance, dropping from 118 FPS to 107 FPS.

Support our work by becoming a Patron!