AMD Challenges NVIDIA With the Launch of MI300X AI Chip

  • The CEO of AMD, Lisa Su, announced the launch of a new AI GPU chip. 
  • The chip will be vital to the company’s strategy toward AI computing offerings.
  • The MI300X, made of multiple chiplets, can handle large language models of up to 80 billion parameters, making it a competitor to market leader NVIDIA.

Lisa Su, the CEO of Advanced Micro Devices (AMD), has revealed an advanced GPU for AI operations. The chip, MI300X, is expected to be the leading challenger to NVIDIA GPUs, which dominate the AI computing market.

GPU chips such as these are essential for companies looking to develop products such as ChatGPT. If AMD succeeds at marketing the MI300X to customers, it will allow the company to tap into a largely unexplored market, as AMD has primarily been known for conventional computer processors.

The new AMD MI300X, which uses a CDNA architecture, can use up to 192 GB of memory, 153 billion transistors and can run large language models with 80 billion parameters. The company will also likely benefit from its partnership with Microsoft, with custom offerings for the tech giant.

See More: European Union Drafts Law To Regulate AI

AMD Releases AI Computer Lineup

Su also unveiled a lineup of AI ‘AMD Instinct Platform’ computers. The platform will combine eight MI300X chips and use 1.5 terabytes of HMB3 memory. Servers made with these platforms will conform to the Open Compute Platform industry standard.

AMD’s MI300A is already competing with NVIDIA’s Grace CPU Hopper GPU chip combinations. The MI300A combines Zen4 CPU chiplets, unlike the CDNA 3 chiplets in the MI300X.

NVIDIA recently made waves in the AI market with offerings such as the AI-powered supercomputer platform DGX GH200 and a new ethernet switch capable of moving large volumes of AI data.

The AMD announcement is important at a time when businesses are heavily investing in AI and LLMs. The global market for AI chips has already reached a value of $30 billion. This is expected to increase to more than $150 billion by 2028.

What do you think will come next in AI hardware? Let us know your thoughts on LinkedInOpens a new window , TwitterOpens a new window , or FacebookOpens a new window .

Image source: Shutterstock