Samsung Aims To Launch Mach-1 AI Accelerator By 2025, Competing Against SK hynix & Rivaling NVIDIA

Mar 22, 2024 at 07:26am EDT
Samsung Aims To Launch Mach-1 AI Accelerator By 2025, Competing Against SK hynix & Rivaling NVIDIA 1

Samsung is all set to unveil its Mach-1 AI chip to combat NVIDIA and its arch-rival in the HBM segment, SK hynix.

Samsung's Mach-1 Chip To Tackle NVIDIA In AI & SK hynix in HBM Segment, Aiming 2025 Release

The Korean giant has been pretty active in the AI industry recently, as the firm recently revealed plans to develop an AGI Computing Lab to accelerate R&D efforts in artificial general intelligence. Similarly, Kyung Kye-hyun, head of Samsung's semiconductor business, has revealed that the company plans to launch its very own AI chip soon, called the Mach-1 with the sole aim of standing out in the markets and giving tough competition to the likes of SK Hynix & Micron while disrupting NVIDIA's AI dominance.

Related Story Japan’s Semiconductor Gas Production Collapses To Zero After China Cuts Off Tungsten, Leaving TSMC, SK Hynix And Samsung Exposed

The current AI systems have performance degradation and power issues due to memory bottlenecks. We will address this problem by establishing an AGI computing lab and achieving AI architecture innovation. Mach-1 is at the center of our R&D efforts.

- Kyung Kye-hyun

While details regarding Mach-1 haven't been revealed yet, as far as what the Samsung official has told us, it is an AI accelerator that comes with significantly improved performance by reducing the "bottlenecks" present within the GPU and memory tile onboard. Samsung sees its Mach-1 chip being employed for high-power inferencing tasks and expects massive adoption in the markets upon its release.

It is a product specified to fit the transformer model. By using several algorithms, we aim to reduce the bottleneck phenomenon that occurs between memory and GPU chips to one-eighth of what we’re witnessing today and improve the power efficiency by eight times.

It (Mach-1) will enable large language model (LLM) inference even with low-power memory instead of power-hungry HBM.

“We will make prototype (Mach-1) chips by the end of this year. Then, we will see AI systems using our latest AI chips early next year,”

- Kyung Kye-hyun (via KED Global)

Samsung has reportedly secured a deal from Korean AI firm, Naver, who are slowly reducing their reliance on NVIDIA chips. The deal includes $752 million worth of Mach-1 chips with additional quantities expected. In terms of when you expect the debut of Mach-1, it is reported that the chip has already reached early prototyping stages, which means that we can see it in the markets by early 2025.

Samsung is aggressively expanding its portfolio of AI products, and judging by the trends we are witnessing; it won't take long before the Korean giant sees itself among the top contenders.

News Source: KED Global

About the author: Muhammad Zuhair is a hardware and technology reporter for Wccftech, specializing in the semiconductor industry and the complex interplay between technology, manufacturing, and geopolitics. His coverage focuses on the corporate strategies and technological roadmaps of industry giants like TSMC, NVIDIA, Samsung, and Intel. Zuhair's expertise lies in deconstructing complex topics such as fabrication nodes (e.g., 2nm process), the economic impact of policies like the CHIPS Act, and the strategic development of AI infrastructure from NVIDIA, AMD and Intel.

Follow Wccftech on Google to get more of our news coverage in your feeds.

Deal of the Day