Santa Clara, California, Sept. 12, 2022 (GLOBE NEWSWIRE) -- SAPEON (CEO Soojung Ryu, www.sapeon.com), a global AI semiconductor company, announced that its first commercialized artificial intelligence semiconductor chip, X220, was acknowledged for its groundbreaking AI processing speed and unrivaled efficiency in the latest MLPerf benchmark test.
SAPEON X220 has demonstrated industry-leading performance, outperforming A2[1], NVIDIA's latest GPU, which dominates the market. In MLPerf’s Inference: Datacenter benchmark test, which evaluates the performance of AI cloud services for data centers, X220-Compact outperformed NVIDIA A2 by 2.3 times, while X220-Enterprise achieved 4.6 times faster performance.
The benchmark test results also showed that SAPEON is superior in terms of power efficiency in addition to absolute performance. In terms of performance-per-watt capabilities (based on maximum power consumption), the X220-Compact was 2.2 times more efficient than the NVIDIA A2, and the X220-Enterprise was 2.0 times more efficient.
One surprising revelation from the MLPerf benchmark results is that even though SAPEON X220 is a product released in 2020 that – in order to increase price competitiveness – uses a 28nm process technology rather than an ultra-fine semiconductor process, it is more efficient than competitors that mostly use a 7nm process.
The Inference: Datacenter benchmark measures the performance of large-scale artificial intelligence services in precisely the type of data centers that SAPEON is targeting. In addition, SAPEON’s product commercialization stage was verified as “Available,” with the company having demonstrated the maturity of both its hardware and software offerings.
SAPEON began as a project started by SK Telecom to improve SK Group’s artificial intelligence computing infrastructure. SAPEON has been used in artificial intelligence speaker NUGU, intelligent video security solution T-view, AI-based media quality upscale solution SUPERNOVA, and others. To establish itself in the global AI semiconductor industry, SAPEON have founded an independent corporation this year based on the technology and expertise gained through these projects. Expanding upon its previous achievements, SAPEON has now demonstrated the superiority of its product performance through this MLPerf result and plans to expand its business.
Soojung Ryu, CEO of SAPEON said, “SAPEON is one of the world’s leading AI semiconductor companies, and we are pleased that X220 – the competitiveness of which has already been verified through internal commercialization – has been independently recognized for its performance excellence in MLPerf benchmark testing and has received great attention in the global market. We plan to expand X220’s various application fields in the future, and in the second half of next year, we are confident that we will create a gap with competing products with the next-generation chip X330, which represents a further improvement in performance.”
About MLPerf
The MLPerf benchmark evaluates how quickly AI semiconductors output accurate results.
MLPerf is an artificial intelligence-related performance test contest conducted by a consortium of AI leaders from academia, research institutes and industries, including Google, Microsoft, Facebook, Stanford, and Harvard, with the mission to build a fair and useful benchmark. MLPerf is divided into Training and Inference testing, and MLPerf training consists of eight workloads covering use cases such as computer vision, reinforcement learning, and language.
About SAPEON Inc.
SAPEON is an independent corporation targeting the global market with SK Telecom’s self-developed AI semiconductor SAPEON as the first result of cooperation among the three companies of SK Telecom’s SK ICT Alliance with SK Square and SK Hynix. For global business, SAPEON's headquarters is US corporation SAPEON Inc. in Santa Clara, California, Silicon Valley. SAPEON Korea, a Korean company, oversees businesses in Korea and Asia. The SAPEON chip is the first Korean non-memory semiconductor for data centers that executes large-scale calculations necessary for realizing AI services at high speed and low power. For more information about SAPEON and its products, visit https://www.sapeon.com/, LinkedIn (sapeon), Facebook (@SAPEON.Korea), Instagram (sapeonkorea), and YouTube (sapeon).
[1]A2: NVIDIA’s latest GPU for data center service, comparable to SAPEON X220 (PCIe Low-Profile Form factor, 60W power consumption).
Attachments
