DeepX Proves Energy Efficiency of AI Semiconductors, Secures Green Technology Certification
TECHWORLD ·
✦ AI Summary
DeepX said its low-power AI inference technology has received the government's green technology certification.
Four DX-M1- and DX-H1-based AI acceleration modules also obtained green technology product verification, and the products include DX-M1 M.2 LPDDR5x2, DX-M1M M.2 (2GB), DX-H1 V-NPU, and DX-H1 Quattro.
KTL-certified test results showed that DeepX's product line reduced power consumption and lowered the operating temperature of the high-performance product line by more than 20 degrees, while the certification and verification are valid through September 9, 2030.
DeepX said on the 14th that the energy efficiency of its self-developed low-power AI inference technology has been officially recognized by the government. The company said it has obtained the government's "green technology certification."
Along with that, four AI acceleration modules based on DX-M1 and DX-H1 that use the technology also separately obtained "green technology product verification." The verified products include two DX-M1-based M.2 acceleration modules for small edge AI devices and two DX-H1-based PCIe acceleration modules for high-performance AI inference.
The certified technology is "AI inference technology for low-power edge computing acceleration." The four products are DX-M1 M.2 LPDDR5x2, DX-M1M M.2 (2GB), DX-H1 V-NPU, and DX-H1 Quattro.
The certification and verification are valid through September 9, 2030.
DeepX said its technology was recognized for its ability to reduce power consumption in AI computing while maintaining processing performance. The company said this was confirmed through KTL-certified test results.
According to KTL-certified test results, DeepX's PCIe high-performance product line reduced power consumption by 21.84% to 26.25% compared with a reference system. DeepX's M.2 small-form-factor product line also showed power savings of 17.26% to 23.35% versus comparable products.
The company also posted differentiated results in terms of heat generation. The high-performance product line was found to operate at temperatures more than 20 degrees lower than comparable products, and DeepX said this demonstrated its technical competitiveness in improving power efficiency and managing heat.
As a core technology for realizing low-power performance, DeepX pointed to its self-developed high-efficiency neural processing unit (NPU) architecture. The company said that during AI computing, large volumes of data are repeatedly moved between memory and processing units, and that data movement and memory access are among the main causes of increased power consumption in AI systems.
By applying its proprietary NPU design, DeepX reduced unnecessary data movement and memory access while maintaining processing performance for AI inference, and also implemented a structure that lowers power consumption.
The Green Certification Deliberation Committee evaluated the technical characteristics of the NPU architecture and judged its power and heat reduction effects to be technically superior.
The green technology certification is a system based on the Framework Act on Carbon Neutrality and Green Growth for Coping with the Climate Crisis. Under this system, the government officially verifies technologies for new industries and climate change response.
Companies and products that receive green certification can get bonus points and preferential treatment in public procurement and commercialization processes, and they may also be eligible for policy funding, guarantees, and financial support. They may also receive various benefits, including support for entering overseas markets.
DeepX expects the certification to have a positive effect on entering the public sector market and expanding its business at home and abroad. A DeepX official said that as AI spreads beyond data centers and into industrial sites such as robots, factories, and mobility, the importance of high performance and low power consumption will grow, adding that the certification signifies objective recognition of the energy efficiency of DeepX's low-power AI semiconductor technology.
The speaker said the company plans to expand the application of its low-power AI technology across a wide range of industrial sites, expressing its determination to help ease the energy burden that comes with the spread of AI.
Source: TECHWORLD · Park Kyu-chan
Original: https://www.epnc.co.kr/news/articleView.html?idxno=406885
References
This article was produced with the help of an automated content generation algorithm.
Source: TECHWORLD
View originalThis article was summarized and organized by BizCrush based on the original article from TECHWORLD. For exact quotations and full details, please refer to the original article.