Four ultra-low-power AI acceleration modules also receive green technology product designation

Power consumption cut by up to 26%, operating temperatures lowered by more than 20 degrees Celsius

Certifications expected to open doors to public procurement preferences and overseas market support

DeepX's low-power AI inference technology has earned the government's green technology certification after tests confirmed it reduces power consumption while maintaining processing performance. [Provided by DeepX]
DeepX's low-power AI inference technology has earned the government's green technology certification after tests confirmed it reduces power consumption while maintaining processing performance. [Provided by DeepX]

DeepX announced Monday that its low-power AI inference technology has received the government's green technology certification, after demonstrating that it cuts power consumption during AI computation while maintaining processing performance. The company said it expects the certification to help it enter the public-sector market and expand its business at home and abroad.

Four AI acceleration modules incorporating the technology also received green technology product designation: the DX-M1 M.2 LPDDR5x2 and DX-M1M M.2 (2GB), both DX-M1-based modules designed for compact edge AI devices, and the DX-H1 V-NPU and DX-H1 Quattro, DX-H1-based modules built for high-performance AI inference.

The certification and product designations are valid through Sept. 9, 2030. Companies and products holding green certification receive preferential treatment or bonus points in public procurement and commercialization processes, and can access a range of support through policy financing, guarantees, financial programs and overseas market entry initiatives.

In authorized testing by Korea Testing Laboratory (KTL), DeepX's PCIe high-performance product line reduced power consumption by 21.84 to 26.25 percent compared with reference products, while its M.2 compact line achieved reductions of 17.26 to 23.35 percent.

The high-performance line also ran more than 20 degrees Celsius cooler than the reference products during operation, demonstrating the company's capabilities in power efficiency and thermal management.

DeepX designed its neural processing unit to reduce unnecessary data movement and memory access during AI computation, lowering power draw without sacrificing throughput. The Green Certification Review Committee acknowledged the excellence of this architecture and its power and heat reduction effects.

"As AI spreads beyond data centers into industrial settings such as robots, factories and mobility, delivering high performance with as little power as possible will only become more critical," a DeepX official said. "This certification is meaningful because it provides objective recognition of the energy efficiency of DeepX's low-power AI semiconductor technology."


joze@heraldcorp.com