Sapion's 'x220' enables high-quality AI services in NHN Cloud.

Sapion announced on the 29th that it has completed technology verification to provide 'Virtual Tryon', a fashion-specific AI service, at the NHN data center where the AI ​​semiconductor 'X220' has been introduced.

'Virtual Try-On' is a technology that allows you to virtually try on clothes you already have on your own photos to see if the clothes you've chosen look good on you. You can try out not only individual clothes but also other clothes that would look good on you.

NHN Cloud announced that in its "Virtual TryOn" service segment, the Sapion Semiconductor X220 GPU was 5.1 times faster than the NVIDIA T4 GPU in terms of processing speed. This is expected to enable the provision of differentiated, high-quality services at low costs.

Sapion's 'X220' is a domestically produced AI semiconductor that was developed using the company's own internal technology for the HW and SW of the NPU (Neural Processing Unit), the core technology of AI semiconductors. In the MLPerf test, it was evaluated to have superior performance in areas such as data center inference compared to third-party GPUs.

In November last year, Sapion CEO Ryu Su-jeong said in an e4ds news interview that they were working with NHN on a server verification project.

Since 2020, Sapion and NHN Cloud have been verifying improved technology compared to GPUs, expanding the infrastructure for commercial service operation, and reviewing the application of NHN Cloud's AI services.

NHN Cloud has built a 5.22 Peta Ops (Peta OPS) Sapion AI infrastructure in its NHN data center, and this year, it expanded and built a 9.22 Peta Ops Sapion AI infrastructure. OPS (Operations Per Second) is a unit of measurement for the number of operations per second. Tera Ops means one trillion operations per second, and Peta Ops means one quadrillion operations per second.

This expansion brings the total capacity of domestically produced AI semiconductor-based cloud infrastructure to 14.44 petaops, equivalent to 14 quadrillion calculations per second.

Meanwhile, Sapion is targeting the data center inference service semiconductor and autonomous driving semiconductor markets. CEO Ryu anticipated the launch of the "x330" this year and signaled his intention to expand commercial services to secure a leading position in the domestic AI market.