Samsung has officially announced its artificial intelligence chip, dubbed 'Mach-1,' set to be released in early 2025. An AI chip that could act as Samsung's bid to compete against NVIDIA's newly-announced 'B200' chip, to be released later this year.
As per Tech Radar, the new chip will go into production by the end of this year and launch in early 2025. It is expected to target edge applications with low power consumption needs.
The announcement was made at the 55th regular shareholders' meeting of the corporation. The CEO of Samsung Semiconductor, Kye Hyun Kyung, stated that the SoC was nearing completion and that the chip design had successfully undergone technological validation on FPGAs.
(Photo : Chung Sung-Jun/Getty Images)SEOUL, SOUTH KOREA - AUGUST 25: The Samsung logo is displayed at the Samsung office on August 25, 2017 in Seoul, South Korea.
According to reports, the Mach-1 accelerator will be able to solve the bottleneck problems that current AI accelerators have while transporting data between the GPU and RAM. It is intended to handle AI inference workloads. Reduced power efficiency and slower data transfer rates are frequently the outcomes of this.
With the reported establishment of a research center devoted to creating a completely new kind of semiconductor required for artificial general intelligence long-standing goal in AI development unveiled their new AI chip.
Initially, the lab will concentrate on creating chips for large language models, hosting and supporting AI models with an emphasis on inference. According to Kyung Kye-Hyun, the head of the company's semiconductor business, it plans to deliver new chip designs. This iterative process will offer higher performance and support for growing larger models at a fraction of the power and cost.
Read Also: Apple's M4 Chips are Coming in 2025, Why Is There a Delay?
South Korea's AI Chip Bid
Other South Korean Startups are also looking to catch up to NVIDIA, also with backing from Samsung. Rebellions, a 2020 startup, is collaborating with Samsung's foundry division to create a state-of-the-art artificial intelligence chip.
This chip, which might launch on the market as early as the second half of this year, is anticipated to transform AI computing. It will integrate Samsung's top-tier HBM3E memory.
Another notable startup is Sapeon, which was spun off from SK Telecom, the largest telecom provider in South Korea. With plans to supply Taiwan Semiconductor Manufacturing Co. with its AI chip, Sapeon hopes to provide its cutting-edge solution to clients by June.
NVIDIA's Flagship AI Chip
Coinciding with the release of all of the new AI chips, the major player in the market, NVIDIA, recently revealed its potent Blackwell architecture, GB200 GPU, and CUDA software upgrades.
According to Jensen Huang, CEO of NVIDIA, the Blackwell chip is a breakthrough device that offers more productivity, efficiency, and processing capacity. It is also anticipated to fuel the next wave of technical advancements that will drive this emerging industry forward.
Huang claims that accelerated computing is a game-changer that surpasses general-purpose computing and is necessary to sustain affordability, sustainability, growth and scalability.
Related Article: Intel, Biden Announce $8.5 Billion in Direct Funding to Advance US Chip Manufacturing
(Photo : Tech Times)
