AMD Chairwoman and CEO Lisa Su is in Taipei Monday to deliver the opening keynote speech of this year’s Computex tech trade show in Taipei Monday. CNA photo June 3, 2024
Taipei, June 3 (CNA) Advanced Micro Devices Inc. (AMD) boss Lisa Su on Monday described ties with Taiwan Semiconductor Manufacturing Co. (TSMC) as “very strong,” amid reports of a looming partnership with TSMC’s primary rival, Samsung Electronics.
Asked during a post-keynote press conference at Computex in Taipei if AMD planned to purchase chips made on Samsung’s 3 nanometer (3nm) gate-all-around (GAA) process, Su said the U.S.-based IC-design giant was committed to using “the most advanced technology.”
“[This means] we’re certainly going to use 3 nanometer, 2 nanometer and beyond,” AMD’s chairwoman and CEO said.
However, Su was quick to underline the strength of AMD’s current relationship with TSMC, noting that the company had not specified the use of “any particular vendor for 3 nanometer or gate all around.”
According to Su, AMD and TSMC have “talked about some of the products in 3 nanometer that we’re doing right now.”
The AMD CEO was also asked about the company’s reported plan of setting up a research and development center in Taiwan following Nvidia, which has set up its first artificial intelligence (AI) research and development center in Asia here in Taiwan in a five-year project starting in 2023.
Nvidia set up its AI R&D center in Taiwan as part of the Taiwan government’s “Supreme A+ Program” — initiated in 2020 — to attract global pioneers in innovation to invest in state-of-the-art technologies in semiconductors, communications and AI in Taiwan. It was reported last month that AMD was planning to follow suit.
However, Su said on Monday that while “Taiwan is a very important area for us to do research and development — we already have over 1,000 people here doing a significant amount of development for all of our product portfolio, including PCs, or data center, as well as our supply chain …, I don’t have any exact information on the R&D centers [in Taiwan] at this moment.”
The AMD CEO emphasized that Taiwan “is very, very important to the semiconductor ecosystem” when asked to comment on Taiwan’s role in the global supply chain.
“We do a lot of our manufacturing here, with key suppliers, like TSMC, and we also have a number of partners that help us build out the ecosystem here in Taiwan, like our ODM partners, as well as a number of contract manufacturers,” she said.
The press conference was held after Su delivered a one-and-half-hour-long opening keynote speech at Computex.
In the opening keynote speech, the tech business leader unveiled AMD’s latest AI processors aimed at ushering in a new era of AI experiences, the company said in its press release.
The new AMD Ryzen AI 300 Series processors are said to have the world’s most powerful neural processing unit (NPU) for next-generation AI PCs that pave the way for a future full of AI-infused computing directly on laptops, according to the AMD boss.
NPU is for AI-specific acceleration, which works alongside central processing unit (CPU) that is good for traditional workloads and graphic processing unit (GPU) that has been great for graphics, processing, and gaming, Su explained.
For desktop PCs, on the other hand, the new AMD Ryzen 9000 Series desktop processors “represent a significant step forward, offering users cutting-edge computing power and reliability and delivering an impressively better performance compared to the prior generation,” according to the company.
During the keynote, Su invited executives from Microsoft (via video), HP, Lenovo and Asus to unveil new PC experiences powered by AMD’s latest third-generation AMD Ryzen AI 300 Series processors and AMD Ryzen 9000 Series desktop processors.
Source: AMD
On enhancing the computing power of data centers, the AMD CEO detailed the company’s “multi-generational accelerator roadmap, showing how it plans to deliver performance and memory leadership on an annual cadence for generative AI,” the press release stated.
Su said that the AMD Instinct MI325X AI accelerators, with a planned launch in the fourth quarter of this year, will be followed by the MI350 Series in 2025 and MI400 series accelerators in 2026.
She noted during her speech that MI325X “offers twice the memory and 1.3 times faster memory bandwidth” compared to Nvidia’s AI chip H200.
Su also touted that the company’s MI350 series “will be built with advanced 3-nm process technology” and “will deliver the biggest generational leap in AI performance in our history.”
Comparing the MI350 series to Nvidia’s B200 — Nvidia’s AI chip that is also said to be available in the fourth quarter of this year — Su said AMD’s series supports up to 1.5 times more memory capacity and delivers 1.2 times more performance.
