Nvidia becomes 'three-chip' company with new ARM-based CPU
Nvidia, the world’s leading graphic computing firm, launched a data center central processing unit (CPU) based on ARM architecture, making it a direct competitor of chip giant Intel.
Nvidia previously announced a US$40 billion acquisition of ARM, but the deal has not yet been approved by regulators.
The company's Grace CPU, scheduled to be available in 2023, will perform 10 times better for artificial intelligence training (AI) and energy efficiency than current chips, Nvidia officials said during its online GPU Technology Conference.
The Grace CPU will be able to handle “the world’s most advanced applications," such as natural language processing, content recommending systems and AI supercomputing. It will be used in gene sequencing, drug, physics and autonomous driving research and tests requiring huge amounts of data calculations, Jensen Huang, Nvidia's founder and chief executive officer, said during an online speech.
Currently, most data center chips are manufactured by Intel and AMD.
The move is Nvidia’s latest expansion after grabbing market share in the graphic chip and data processing unit (DPU) sectors.
Nvidia is now a “three-chip” company, Huang said, referring to GPUs, CPUs and DPUs.
During the GTC conference, Nvidia also launched its new DPU used for autonomous car firms.
Startup car firms, including many from China, showcased during the GTC conference.
Nvidia is still facing challenges in its ARM acquisition, such as approval from the Chinese government. The China-United States tech dispute may also bring uncertainty to the deal, industry insiders said.
In 2018, Qualcomm failed to acquire NXP in an acquisition deal valued at US$44 billion. Chinese regulators did not approve the deal.
