{"id":27541,"date":"2025-08-27T11:52:40","date_gmt":"2025-08-27T11:52:40","guid":{"rendered":"https:\/\/metaverseplanet.net\/blog\/?p=27541"},"modified":"2026-01-05T09:11:07","modified_gmt":"2026-01-05T09:11:07","slug":"google-details-its-most-powerful-ai-processor-ironwood","status":"publish","type":"post","link":"https:\/\/metaverseplanet.net\/blog\/google-details-its-most-powerful-ai-processor-ironwood\/","title":{"rendered":"Google Details Its Most Powerful AI Processor, Ironwood"},"content":{"rendered":"\n

Tech giant Google<\/strong> detailed its Ironwood TPU Superpod<\/strong> solution at Hot Chips 2025<\/strong>. The new system delivers a performance increase of over 16 times compared to the previous generation.<\/p>\n\n\n\n

At the Hot Chips 2025<\/strong> event, Google<\/strong> shared detailed information about its next-generation artificial intelligence accelerator<\/strong>, “Ironwood<\/strong>” TPU Superpod<\/strong>, which was introduced in recent months. This system, built on the company’s 7th generation TPU architecture<\/strong>, offers a massive performance boost compared to previous generations.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

The Ironwood TPU<\/strong> comes with 9216 chips<\/strong> in each pod, 192 GB of HBM memory<\/strong> per chip, and a processing power of 4614 TFLOPs<\/strong>. These figures correspond to more than 16 times<\/strong> the performance offered by the TPU v4<\/strong> introduced in 2022. This giant structure can achieve a total computing power of 42.5 Exaflops<\/strong>.<\/p>\n\n\n\n

The company claims that Ironwood<\/strong> is 24 times faster than similar system segments of El Capitan<\/strong>, known as the world’s most powerful supercomputer. Google<\/a><\/em><\/strong> makes this comparison based on FP8, but El Capitan<\/strong>, which offers 1.7 Exaflops, does not support this.<\/p>\n\n\n\n

Ironwood Solution Detailed<\/strong><\/h2>\n\n\n\n
\"\"<\/figure>\n\n\n\n

The TPU v4<\/strong> reached 275 TFLOPs with 4096 chips and 32 GB of HBM memory per chip, while the TPU v5p<\/strong>, which arrived a year later, offered 8960 chips, 95 GB of memory, and a capacity of 459 TFLOPs. Ironwood<\/strong> raises this bar much higher.<\/p>\n\n\n\n

At the heart of Ironwood<\/strong> is the new generation Ironwood SoC<\/strong>. Four chips are placed on each motherboard, and 16 motherboards are combined to form a rack<\/strong>. This structure is designed as a modular block of 64 chips. Google<\/strong> connects these blocks using a network architecture called 3D Torus<\/strong>. This 4x4x4 structure provides low latency and high flexibility in data communication.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

The system also uses Google\u2019s InterChip Interconnect (ICI)<\/strong> technology. This enables scalability up to 43 blocks, achieving a network capacity of 1.8 petabytes. An Ironwood Superpod<\/strong> contains a total of 144 racks<\/strong>. Additionally, optical switch cases and liquid cooling<\/strong> units are integral parts of this structure. Google<\/strong> supports the entire system with hybrid connection technologies. This means copper connections on the motherboard are combined with fiber optic lines to provide both flexibility and high bandwidth.<\/p>\n\n\n\n

On the energy front, each rack<\/strong> operates with a custom power distribution that converts a 416-volt AC input to a DC output. The entire system is liquid-cooled<\/strong>, and a single rack can consume over 100 kW<\/strong> of power.<\/p>\n\n\n\n

You Might Also Like;<\/h3>\n\n\n
    <\/ul>","protected":false},"excerpt":{"rendered":"

    Tech giant Google detailed its Ironwood TPU Superpod solution at Hot Chips 2025. The new system delivers a performance increase of over 16 times compared to the previous generation. At the Hot Chips 2025 event, Google shared detailed information about its next-generation artificial intelligence accelerator, “Ironwood” TPU Superpod, which was introduced in recent months. This …<\/p>\n","protected":false},"author":1,"featured_media":27545,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"googlesitekit_rrm_CAown96uCw:productID":"","footnotes":""},"categories":[332],"tags":[335,156],"class_list":["post-27541","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai-information","tag-ai-news","tag-chip-technology"],"amp_enabled":true,"_links":{"self":[{"href":"https:\/\/metaverseplanet.net\/blog\/wp-json\/wp\/v2\/posts\/27541","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/metaverseplanet.net\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/metaverseplanet.net\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/metaverseplanet.net\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/metaverseplanet.net\/blog\/wp-json\/wp\/v2\/comments?post=27541"}],"version-history":[{"count":0,"href":"https:\/\/metaverseplanet.net\/blog\/wp-json\/wp\/v2\/posts\/27541\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/metaverseplanet.net\/blog\/wp-json\/wp\/v2\/media\/27545"}],"wp:attachment":[{"href":"https:\/\/metaverseplanet.net\/blog\/wp-json\/wp\/v2\/media?parent=27541"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/metaverseplanet.net\/blog\/wp-json\/wp\/v2\/categories?post=27541"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/metaverseplanet.net\/blog\/wp-json\/wp\/v2\/tags?post=27541"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}