Sapeon unveils new AI chip for data centre
The new AI inference neural processing unit (NPU), dubbed X330, is four times faster and two times more energy efficient than the X220, the first AI chip the company developed for data centers in 2020, according to the company.
Representative image
SEOUL: Sapeon, a South Korean artificial intelligence (AI) processor fabless company, has unveiled its new AI semiconductor for data centre operations.
The new AI inference neural processing unit (NPU), dubbed X330, is four times faster and two times more energy efficient than the X220, the first AI chip the company developed for data centers in 2020, according to the company.
Sapeon said the X330 is also two times faster and 1.3 times more energy efficient compared with the latest AI inference chip launched by its competitor. The company did not elaborate on the details of the competitor.
The X330 can be used across various industries and supports large language models (LLMs), including the transformer-based LLM, which is the foundational technology of ChatGPT, said the company.
The new AI chip will start mass production through the seven nanometer process of Taiwan's TSMC, the world's biggest contract chip maker, in the first half of next year following prototype tests and credibility verification.
The company said it also plans to introduce intellectual property for self-driving cars and an AI NPU for edge devices, such as surveillance cameras, in the future.
Sapeon, originally a joint venture created by SK Telecom Co., South Korea's leading mobile carrier; SK hynix Inc., the country's No. 2 chipmaker; and SK Square, an investment arm of SKT, spun off from SK Telecom last year. It is headquartered in the U.S.' Silicon Valley.