[Edaily Reporter Kwon Oh-seok] Acryl, an AI transformation (AX) infrastructure company, announced on the 11th that it is launching a full-scale performance verification project called “K-Scale Evaluation” for GPUBASE across major global cloud environments, including Amazon Web Services (AWS). GPUBASE is a core solution of the AX infrastructure platform “Jonathan.”

This project is an extension of a preliminary evaluation conducted last March with a global top-tier cloud provider using hundreds of GPUs, and it will expand the scope of verification to an environment with a cumulative total of over 1,000 GPUs. Acryl plans to comprehensively assess the performance, scalability, stability, and compatibility of GPUBASE in major CSP environments, including AWS, through Megazone Cloud, a domestic cloud management service provider.
The evaluation will be conducted using a dual "horizontal" and "vertical" structure. The horizontal validation will verify GPUBASE’s portability and operational stability across various cloud environments, including AWS, while the vertical validation will focus on testing multi-path transmission, load handling capabilities, and large-scale scalability within a single RoCEv2-based cluster.
In particular, by leveraging AWS’s ultra-high-speed network technology, EFA (Elastic Fabric Adapter), to bundle multiple GPUs into a single system-like environment, the company plans to verify GPUBASE’s cluster operational performance and assess its applicability within global cloud infrastructure.
The validation will proceed in stages. First, the latest GPU-based performance will be evaluated across various CSP environments, including AWS EFA, followed by a comprehensive validation of core functions within a RoCEv2-based single-cluster environment. Concurrently, performance testing will be conducted through a domestic accredited testing laboratory to secure a KOLAS-accredited test report, thereby objectively demonstrating the technology’s reliability.
GPUBASE is GPU cluster optimization software developed in-house by Acryl. It supports the efficient operation of large-scale AI infrastructure through GPU resource monitoring, partitioning, and virtualization functions. This enables enterprises and institutions to flexibly utilize GPU resources throughout the entire process, from AI model development to training, inference, and deployment.
Based on these validation results, Acryl plans to accelerate the expansion of global cloud partnerships and its entry into overseas markets. The company’s strategy is to secure evaluation reports, accredited certification documents, and case studies of global CSP collaborations to establish GPUBASE as a competitive core platform in the global AX infrastructure market.