본문 바로가기
bar_progress

Text Size

Close

CU Box Introduces Latest NVIDIA Servers... "Advancing AI Technology"

CUBox announced on the 7th that it has introduced and officially started operating a large-scale NVIDIA H100 GPU server for use in artificial intelligence (AI) algorithm research and generative AI service development. The H100 GPU server that CUBox has started operating this time consists of a total of 16 nodes.


NVIDIA's latest GPU, the H100, enables AI training up to 9 times faster compared to the previous generation GPU. Additionally, inference performance can be improved by 30 times, and HPC (High Performance Computing) application performance can be enhanced up to 7 times, making it essential for training large language models (LLM) or ultra-large image generation models. By including the existing 10-node A100 servers, CUBox now owns a total of 26 GPU server nodes, equipping it with the largest scale of GPU resources among specialized AI video recognition companies in Korea to utilize AI video recognition technology more effectively.


CUBox proved its status as a world-class company with CNN-based video recognition technology through the 2021 NIST FRVT (National Institute of Standards and Technology Face Recognition Vendor Test) in the United States. Not resting on this achievement, the company focused its capabilities on developing video recognition algorithms based on the ViT (Vision Transformer) model, which other AI companies found burdensome. The ViT model guarantees high algorithm performance but requires heavy datasets, making high-performance computing power essential for application and development.


CUBox developed algorithms based on the ViT model using the GPU servers it previously owned. To strengthen this capability, it decided to introduce large-scale latest GPU servers. A CUBox representative stated, "We will actively utilize the latest GPU servers for AI technology development that must secure a competitive edge in the global market, such as the ongoing Incheon Airport Smart Pass algorithm supply project," adding, "This GPU server will also be actively used in next-generation technology projects such as order-picking robots, diagnostic medical AI, and 3D X-Ray development."


Along with this, CUBox is building high-performance generative AI services that can demonstrate the capabilities of specialized companies through the introduction of large-scale GPU servers and plans to showcase them in the market in the near future. A CUBox representative emphasized, "We will continue aggressive infrastructure investments to become a world-class video recognition company, and through this, we will continuously maintain technological superiority," adding, "By applying actual technology to our products such as Incheon Airport Smart Pass and FaceOn, we will enable the general public to experience the best technological capabilities that CUBox possesses."


© The Asia Business Daily(www.asiae.co.kr). All rights reserved.

Special Coverage


Join us on social!

Top