Investors were: Daekyo Investment, SL Investment, Smilegate Investment, GNTech Venture Capital, Time Works Investment, Yuanta Investment, and Quantum Ventures Korea.
Panmesia says it has a way of running recommendation models 5x faster by feeding data from external memory pools to GPUs via CXL caching rather than host CPU-controlled memory transfers.
The company’s TrainingCXL technology was developed by computing researchers at the Korea Advanced Institute of Science & Technology (KAIST).
CXL is gaining traction as an interface technology designed to connect diverse system devices, including memory expansion units, accelerators, processors, and switches.
Its promise lies in the ease it offers in memory capacity augmentation and in streamlining data processing. One feature of CXL, memory pooling, presents avenues for cost savings in memory-related operations.
Meta, Microsoft, and Google are demonstrating heightened interest in CXL-centric solutions and Intel, AMD, Samsung Electronics, and SK Hynix have laid out their strategic blueprints to introduce CXL-compliant devices.
Panmnesia ckaims to have presented the world’s first CXL 2.0-based full- system memory pooling framework at the USENIX ATC conference in Boston last year. The company unveiled a solution this May that accelerates data centre-oriented AI applications, using CXL 3.0 technology.
At the Flash Memory Summit 2023 in Santa Clara, California, the company disclosed a full system, from devices to operating systems, connected through a multi-level switch architecture supporting CXL 3.0.
.”CXL 3.0 takes a fundamentally different viewpoint from CXL 2.0 in its approach to various processors and memory semiconductor devices to support high scalability,” says Myoungsoo Jung, CEO of Panmnesia, “regrettably, most companies are still only discussing certain features of CXL 1.1 or 2.0.”
Jung says that Panmnesia has already completed the development of hardware and software to support CXL 3.0, which will be of significant assistance to a diverse range of companies utilising CXL.
Panmesia’s CXL product line covers everything, including CXL IP, hardware, system solutions to enable these, and software essential for accelerating AI and data processing at the server and data centre levels. The products are planned to be available to a broader customer base following the completion of state-of-the-art silicon tape-out later this month.
Stay up to date with the latest in industry offers by subscribing us. Our newsletter is your key to receiving expert tips.
ASML, the Dutch maker of advanced chip-making machines that are critical to global supply chains, reaffirmed its long-term revenue outlook as it bets on an artificial intelligence-driven boom in semic
Nvidia is gearing up to debut its next-gen "Rubin" GPU in 2025, featuring eight HBM4 chips, with a Rubin Ultra version to follow, packing twelve. In a strategic bid to reclaim HBM market share, Samsun
The Trump administration first imposed export restrictions on semiconductor equipment to China in 2019, targeting ASML's EUV lithography machines. The Biden administration has since ramped up cont