{"id":140419,"date":"2025-05-29T03:52:11","date_gmt":"2025-05-29T03:52:11","guid":{"rendered":"https:\/\/www.europesays.com\/uk\/140419\/"},"modified":"2025-05-29T03:52:11","modified_gmt":"2025-05-29T03:52:11","slug":"how-huawei-turns-sanctions-into-threat-to-nvidia","status":"publish","type":"post","link":"https:\/\/www.europesays.com\/uk\/140419\/","title":{"rendered":"How Huawei turns sanctions into threat to Nvidia"},"content":{"rendered":"<ul>\n<li><strong>Huawei\u2019s Supernode 384 computing architecture rivals Nvidia and circumvents US restrictions.<\/strong><\/li>\n<li><strong>Peer-to-peer architecture delivers up to 2.5x faster performance compared to legacy clusters.<\/strong><\/li>\n<\/ul>\n<p>Huawei\u2019s ambitious push into high-performance AI computing has taken a significant leap forward with its Supernode 384 architecture, positioning the Chinese tech giant as <a href=\"https:\/\/techwireasia.com\/2025\/05\/nvidia-ai-competition-huang-warns-china-not-behind-as-huawei-fills-chip-void\/\" target=\"_blank\" rel=\"noopener\">a formidable challenger<\/a> to Nvidia\u2019s market-leading position despite ongoing US sanctions.<\/p>\n<p>The Shenzhen-based company unveiled details of its computing framework at last Friday\u2019s Kunpeng Ascend Developer Conference, where executives outlined how the Supernode 384 system addresses critical bottlenecks that have long plagued large-scale AI training operations.<\/p>\n<p>\u201cAs the scale of parallel processing grows, cross-machine bandwidth in traditional server architectures has become a critical bottleneck for training,\u201d <a href=\"https:\/\/www.scmp.com\/tech\/big-tech\/article\/3311873\/huawei-pushes-ascend-ai-processor-based-supernode-computing-architecture-developers\" target=\"_blank\" rel=\"noopener\">said<\/a> Zhang Dixuan, president of Huawei\u2019s Ascend computing business, during his keynote address.<\/p>\n<p>Breaking from traditional computing<\/p>\n<p>The Supernode 384 represents a departure from conventional computing approaches. Unlike traditional Von Neumann architectures that rely on separate processing units, memory, and data buses, Huawei\u2019s system adopts a peer-to-peer computing model specifically optimised for next-generation AI workloads.<\/p>\n<p>The architectural shift proves particularly advantageous for Mixture-of-Experts (MoE) AI models \u2013 machine-learning systems that deploy multiple specialised sub-networks to tackle complex problems. Such models have become increasingly important as AI applications grow more sophisticated and demanding.<\/p>\n<p>The technical specifications are impressive. Huawei\u2019s CloudMatrix 384 system, built on the Supernode 384 foundation, comprises 384 Ascend AI processors distributed in 12 <a href=\"https:\/\/techwireasia.com\/2025\/05\/how-huawei-cloud-plans-to-scale-ai-in-asia-pacific\/\" target=\"_blank\" rel=\"noopener\">computing cabinets<\/a> and four bus cabinets.<\/p>\n<p>The configuration delivers 300 petaflops of computing power \u2013 equivalent to 300 quadrillion calculations per second \u2013 alongside 48 terabytes of high-bandwidth memory.<\/p>\n<p>Performance benchmarks tell a compelling story<\/p>\n<p>Benchmark results presented at the developer conference reveal the system\u2019s competitive edge. On dense AI models like Meta\u2019s LLaMA 3, the Supernode 384 achieved 132 tokens per second per card \u2013 2.5 times faster than legacy cluster systems.<\/p>\n<p>For communications-intensive applications, including Alibaba\u2019s Qwen and DeepSeek models, performance reached 600 to 750 tokens per second per card. The improvements stem partly from Huawei\u2019s decision to replace traditional Ethernet interconnects with high-speed bus connections, boosting communications bandwidth by 15 times.<\/p>\n<p>The company also reduced single-hop communications latency from 2 microseconds to 200 nanoseconds \u2013 a tenfold improvement that enables the CloudMatrix 384 cluster to function as a unified computing entity.<\/p>\n<p>Strategic response to geopolitical pressures<\/p>\n<p>The timing and positioning of Huawei\u2019s Supernode 384 cannot be separated from broader geopolitical tensions. US tech restrictions have significantly constrained Huawei\u2019s access to advanced semiconductor technologies, forcing the company to innovate inside existing constraints.<\/p>\n<p>According to SemiAnalysis, the CloudMatrix 384 likely uses Huawei\u2019s latest Ascend 910C AI processor. While individual chip performance may lag behind cutting-edge alternatives, the report suggests Huawei compensates through superior architecture.<\/p>\n<p>\u201cHuawei is a generation behind in chips, but its scale-up solution is <a href=\"https:\/\/techwireasia.com\/2025\/04\/huaweis-ai-chip-breakthrough-signals-a-major-challenge-to-nvidias-dominance-in-china\/\" target=\"_blank\" rel=\"noopener\">arguably a generation ahead<\/a> of Nvidia and AMD\u2019s current products in the market,\u201d the SemiAnalysis <a href=\"https:\/\/semianalysis.com\/2025\/04\/16\/huawei-ai-cloudmatrix-384-chinas-answer-to-nvidia-gb200-nvl72\/\" target=\"_blank\" rel=\"noopener\">report<\/a> noted.<\/p>\n<p>Implications for the Global AI Landscape<\/p>\n<p>Huawei\u2019s architectural innovation carries implications for the <a href=\"https:\/\/techwireasia.com\/2025\/04\/with-nvidia-chips-out-of-reach-huawei-steps-in-with-new-ai-processor\/\" target=\"_blank\" rel=\"noopener\">global AI computing market<\/a>. The company has already deployed CloudMatrix 384 systems in data centres in the Anhui province, Inner Mongolia, and Guizhou province, demonstrating practical implementation is a reality.<\/p>\n<p>The Supernode 384\u2019s scalability potential \u2013 capable of linking tens of thousands of processors \u2013 positions it as a viable platform for training increasingly sophisticated AI models. The capability becomes important as organisations in industries seek to implement AI solutions at unprecedented scales.<\/p>\n<p>For the broader technology ecosystem, Huawei\u2019s advancement represents both opportunity and challenge. While it provides an alternative to Nvidia\u2019s dominant position, it also highlights the fragmenting nature of global technology infrastructure amid ongoing geopolitical tensions.<\/p>\n","protected":false},"excerpt":{"rendered":"Huawei\u2019s Supernode 384 computing architecture rivals Nvidia and circumvents US restrictions. Peer-to-peer architecture delivers up to 2.5x faster&hellip;\n","protected":false},"author":2,"featured_media":140420,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[3164],"tags":[323,1395,60933,3284,60934,53,16,15],"class_list":{"0":"post-140419","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-computing","8":"tag-ai","9":"tag-china","10":"tag-cluster","11":"tag-computing","12":"tag-learning-models","13":"tag-technology","14":"tag-uk","15":"tag-united-kingdom"},"share_on_mastodon":{"url":"https:\/\/pubeurope.com\/@uk\/114589096342744016","error":""},"_links":{"self":[{"href":"https:\/\/www.europesays.com\/uk\/wp-json\/wp\/v2\/posts\/140419","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/uk\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/uk\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/uk\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/uk\/wp-json\/wp\/v2\/comments?post=140419"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/uk\/wp-json\/wp\/v2\/posts\/140419\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/uk\/wp-json\/wp\/v2\/media\/140420"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/uk\/wp-json\/wp\/v2\/media?parent=140419"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/uk\/wp-json\/wp\/v2\/categories?post=140419"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/uk\/wp-json\/wp\/v2\/tags?post=140419"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}