Intel Announces AI Chip to Challenge Nvidia, AMD – BusinessKorea
At the Computex Conference in Taipei, Taiwan, which has become the world’s largest event in the AI sector, Intel announced a new Artificial Intelligence (AI) chip aimed at challenging Nvidia and AMD.
According to international reports, Intel’s CEO Pat Gelsinger unveiled the 6th generation “Xeon6” server CPU AI chip at the conference, targeting AMD, which has been encroaching on Intel’s market share in the data center sector. Alongside this, he introduced the “Gaudi 3 Accelerator Kit,” an AI training chipset to compete with Nvidia’s H100 at less than half the price.
The announcement came in the wake of Nvidia and AMD each launching new AI chips just two and one day(s) prior, respectively. Intel had released its 5th generation Xeon processor for data center workloads six months ago, and the Gaudi 3 processor for AI model training and deployment was announced two months ago.
CEO Gelsinger emphasized that the 6th generation Xeon chip consumes less power while offering significantly stronger performance. The Xeon6, a 6th generation processor, shows a performance improvement of up to 4.2 times compared to its predecessors. This chip is crucial for Intel as it continues to lag behind AMD in data center market share.
According to data cited by Reuters from Mercury Research, Intel’s market share of x86 chips in the data center sector has decreased by 5.6 percentage points over the past year to 76.4%, with AMD currently holding 23.6%.
On this occasion, CEO Gelsinger also outlined a strategy to counter Nvidia, the leading company in AI chips. He presented a plan to position Intel as the sole serious contender to Nvidia by offering superior cost effectiveness. “Intel’s AI accelerator, ‘Gaudi2,’ is priced at one-third the cost of competing AI GPUs, and ‘Gaudi3’ at two-thirds,” Gelsinger stated. He added, “Gaudi3 provides up to 40% faster training times compared to Nvidia’s H100 GPU of the same class, and when running massive language models it offers inference speeds up to twice as fast as the H100.”
Gelsinger highlighted the attractive pricing of the Gaudi3 accelerator set, which features eight Gaudi3 chips in a server system priced at US$125,000, significantly lower than Nvidia’s equivalent server systems, which cost over US$300,000. The regular price of Intel’s previous generation, Gaudi 2, was US$65,000.
Intel also announced a next-generation processor for laptops named “Lunar Lake,” which is said to use 40% less electricity and boasts significantly enhanced performance, set to be released in the third quarter. Gelsinger remarked that Intel is dedicated to AI PCs and “will drive the largest number of next-generation AI PCs in the industry.”
The products will be manufactured using TSMC’s 3-nm process. Lunar Lake is a System on Chip (SoC) that reduces power consumption by up to 40% compared to its predecessors and triples AI computing performance. Following the release of Lunar Lake and “Arrow Lake” this year, Gelsinger announced that “Panther Lake” would be launched next year.
Regarding the growing foundry business, Gelsinger emphasized, “Intel is already number two.” Since the first quarter of this year, Intel has separated accounting for its design and foundry operations and also counts its own volume in foundry performance. Gelsinger was scheduled to attend the “AI Summit” in South Korea on June 5 but canceled his visit. He mentioned, “I will visit South Korea again at the end of the year.”
Gelsinger also noted collaboration with Samsung. He explained, “We are collaborating with Samsung Medison on an AI-powered ultrasound solution, making it easier and quicker for doctors to capture ultrasound images.”
Intel confirmed that plans to commence mass production using the 18A (1.8-nm) process within the year are proceeding as scheduled.
Never miss another article! Enter your email address below to receive our free daily newsletter:
