Courtesy of SCMP

Huawei Challenges Nvidia with New AI Supernode Boasting 300 Petaflops

To highlight Huawei's advancements in AI infrastructure and its potential to rival Nvidia's offerings.

15 Apr 2025, 20.30 WIB

101 dibaca

Ikhtisar 15 Detik

Huawei meluncurkan CloudMatrix 384 Supernode untuk bersaing dengan Nvidia.
CloudMatrix 384 Supernode menawarkan lebih banyak petaflops dibandingkan NVL72.
Inisiatif Huawei menunjukkan upaya untuk mencapai kemandirian dalam teknologi komputasi di tengah sanksi.

Wuhu, Anhui, China - Huawei, a company under US sanctions, has made significant strides in achieving self-sufficiency in computing power, particularly in the field of artificial intelligence (AI). The company recently launched the CloudMatrix 384 Supernode, which is being compared to Nvidia’s NVL72 system. This new supernode is described as a 'nuclear-level product' and is aimed at addressing computing bottlenecks in AI data centers.

Nvidia’s NVL72, introduced last year, features a 72-graphics processing unit (GPU) NVLink domain that functions as a single, powerful GPU. This allows for real-time inference for trillion-parameter large language models (LLMs) at speeds 30 times faster than previous generations. NVLink is a high-speed interconnect technology developed by Nvidia to enable efficient communication and data sharing between multiple GPUs.

Huawei’s CloudMatrix 384 Supernode, currently deployed in the company’s data centers in Wuhu, Anhui province, has achieved 300 petaflops of computing power. This is significantly higher than the 180 petaflops offered by Nvidia’s NVL72. Huawei has not yet responded to requests for further comments on this development.

Sumber: https://www.scmp.com/tech/tech-war/article/3306644/tech-war-huawei-launches-new-ai-architecture-said-rival-nvidias-products?module=top_story&pgtype=section