Nvidia’s tailored-for-China H20 AI chip now available for pre-orders, set for competition with Huawei: report

Nvidia’s tailored-for-China H20 AI chip now available for pre-orders, set for competition with Huawei: report

Prior to the United States curbs, Nvidia held over 90% of the AI chip market in China. Credit: Nvidia

Nvidia is making a tactical go back to the Chinese market by presenting its China-specific H20 AI chip, a relocation that follows United States policies impacted the business’s sales in China. Nvidia has actually started taking pre-orders from suppliers for its H20 AI chips, priced nearly on par with Huawei’s Ascend 910B, according to Reuters

Why it matters: To resolve difficulties presented by the United States’s restrictions on the sale of particular GPUs (Graphics Processing Units) to China, Nvidia has actually introduced cut-down variations certified with American export policies. In the middle of issues about possible constraints on accessing Nvidia’s items, Huawei’s chip is commonly acknowledged as the leading alternative AI offering in China.

Information: Prior to the United States curbs, Nvidia held over 90% of the AI chip market in China, according to the Reuters report. It is now dealing with growing competitors from regional rivals such as Huawei. The H20 AI chip is priced at $12,000 to $15,000, placing itself as a rival to Huawei’s Ascend 910B.

  • In October 2023the United States federal government enforced brand-new limitations on the export of innovative AI chips, leading Nvidia to instantly stop deliveries of high-performance AI chips consisting of the A100, A800, H100, H800, and L40S items. Consequently, Nvidia started the advancement of brand-new AI chips particularly created for the Chinese market, consisting of the H20, L20, and L2. All 3 chips are customized variations of Nvidia’s H100 AI chip
  • While the H20 is anticipated to offer less computing power than Nvidia’s flagship H100 AI chip, specs recommend its efficiency is likewise inferior to Huawei’s Ascend 910B in particular crucial elements. Especially, the H20 might drag the 910B in FP32 efficiency, an important metric that determines processing speed, and is ranked at less than half of its rival’s ability, the source behind the Reuters report informed the news company.
  • The H20 is most likely to have an edge over the 910B in terms of interconnectivity speed, with the H20 being competitive in applications that need the connection of a big number of chips to work as a system, the report discussed.
  • In regards to efficiency, the H20 chip’s AI computing power is somewhat less than 15% of the H100, according to United States media outlet WccftechThe H20 AI chip includes 96GB memory capability operating at approximately 4.0 Tb/s, 296 TFLOPs calculating power, and an efficiency density of 2.9 TFLOPs/die, compared to the H100’s 19.4 TFLOPs/die.
  • Distributors have actually supposedly notified customers that they can begin shipments of H20 items in little batches throughout the very first quarter of 2024, with bigger amounts offered from the 2nd quarter. Last month, it was reported that Nvidia plans to standardize the H20 in the 2nd quarter of this year.

Context: On Jan. 20, Nvidia CEO Jensen Huang checked out the business’s workplaces in Shenzhen, Shanghai, and Beijing for the yearly celebrations kept in event of China’s Lunar New Year vacation, with the business clarifying that the go to did not include organization operations.

Jessie Wu is a tech press reporter based in Shanghai. She covers customer electronic devices, semiconductor, and the video gaming market for TechNode. Get in touch with her by means of email: jessie.wu@technode.com.


Learn more

Leave a Reply

Your email address will not be published. Required fields are marked *