Advanced Micro Devices (AMD) has commenced mass production of its latest artificial intelligence (AI) and high-performance computing (HPC) accelerator, the Instinct MI300X. The first batch of MI300X units is set to be delivered to longstanding partner LaminiAI, a company specializing in AI solutions. LaminiAI has a history of collaboration with AMD! earning it priority access to the cutting-edge MI300X accelerators.
Sharon Zhou, CEO of LaminiAI, shared the milestone on social media! stating that the initial batch of MI300X accelerators has been put into production. LaminiAI plans to deploy these accelerators to run large language models (LLM) aimed at meeting the demanding requirements of enterprise users. Zhou further revealed that the company has received several machines, each equipped with eight Instinct MI300X accelerators! and hinted at the imminent arrival of additional LLM clusters.
The MI300X is an advanced iteration within the AMD Instinct MI300 series, strategically designed for AI and HPC computing. It distinguishes itself by abandoning the x86 CPU core and incorporating an increased number of CDNA 3 computing units! totaling 304—significantly higher than the MI300A’s 228 units. Moreover, the MI300X boasts 192GB of HBM3 high-speed memory, with a remarkable peak bandwidth of up to 5.3 TB/s.
AMD’s official performance tests indicate that the MI300X outperforms the currently available Nvidia H100 80GB and may even pose a competitive challenge to the upcoming H200 141GB. Major industry players such as Meta and Microsoft have already made substantial purchases of the AMD Instinct MI300 series! with LaminiAI proudly standing as the first company to publicly embrace and utilize the MI300X.
The AMD Instinct MI300X represents a significant leap forward in AI and HPC capabilities! and its adoption by LaminiAI underscores the company’s commitment to staying at the forefront of technological innovation. As AI applications continue to evolve, partnerships like this one between AMD and LaminiAI pave the way for enhanced performance and efficiency in meeting the growing demands of enterprise users.