At an event in San Francisco, AMD CEO Lisa Su said that its latest semiconductor MI300X chip will have 192 gigabytes of memory, more than any current NVIDIA chip. He stressed that the chip will be a key performance measure of how well it will handle large-scale AI systems needed for services similar to Chat GPT.CEO Su said,
“There is no doubt that AI will be a key driver of (semiconductor) silicon consumption in the near future,” adding, “Customers will receive sample chips in the third quarter, and production will increase at the end of the year.”He also boasted a system that combines eight MI300X chips into one computer. The system is competing with Nvidia’s similar products.Nvidia’s stock price has soared 170% this year. Nvidia occupies 80-95% of the AI computing market. Although startups such as Intel and Cerebras have competitive products, there are virtually no large production competitors against Nvidia because it is a small amount of production. Nvidia’s leadership in this sector was possible because it has been providing software tools to AI researchers as well as chips for more than a decade and learning how to predict what is needed on chips that take years to design, Reuters explained. Experts were also divided over AMD’s ambitious plan. “Even if AMD’s software solutions are competitive in terms of hardware performance, we are still not convinced that they are competitive with Nvidia’s software solutions,” Moore Insight & Strategy said. However, Sumis Chintara, vice president of Meta, said, “We have worked closely with AMD to make it easier for AI developers to switch from one platform to another using free tools,” adding, “It doesn’t really take much work to move from one platform to another.”
TED PARK
US ASIA JOURNAL
