<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:media="http://search.yahoo.com/mrss/"><channel><title>ITANU Magazine: Hardware</title><link>https://itanu.org/magazine/desk/hardware/</link><description>Technology news in nine languages, three editions a day. Every story names its sources and is checked against them.</description><language>en</language><lastBuildDate>Mon, 16 Nov 2026 17:19:41 GMT</lastBuildDate><atom:link href="https://itanu.org/magazine/desk/hardware/feed.xml" rel="self" type="application/rss+xml"/><item><title>NVIDIA says Vera Rubin systems deliver up to twice GB300 training speed per GPU</title><link>https://itanu.org/magazine/nvidia-vera-rubin-mlperf-training-debut-tfqpvtzr/</link><guid isPermaLink="true">https://itanu.org/magazine/nvidia-vera-rubin-mlperf-training-debut-tfqpvtzr/</guid><pubDate>Mon, 16 Nov 2026 17:19:41 GMT</pubDate><dc:creator>ITANU Magazine newsroom</dc:creator><category>Hardware</category><description>The Vera Rubin NVL72 system made its first appearance in MLPerf Training v6.1. MLCommons, which runs the benchmark, confirmed that this round was the first to include Rubin-generation GPUs.</description><content:encoded>&lt;figure&gt;&lt;img src=&quot;https://itanu.org/assets/magazine-img-a1828ad62e1dda00-960.jpg&quot; width=&quot;1166&quot; height=&quot;948&quot; alt=&quot;A white Nvidia sign with a green logo and street address stands among trees and shrubs on a corporate campus.&quot;&gt;&lt;figcaption&gt;File photo: Nvidia sign at its Santa Clara headquarters campus, 2020 · Will Buckner / Wikimedia Commons · CC BY 2.0&lt;/figcaption&gt;&lt;/figure&gt;&lt;h3&gt;Key points&lt;/h3&gt;&lt;ul&gt;&lt;li&gt;NVIDIA says Vera Rubin NVL72 reached up to 2x the per-GPU performance of GB300 NVL72 in MLPerf Training v6.1.&lt;/li&gt;&lt;li&gt;MLCommons confirmed that NVIDIA Rubin-generation GPUs and Google&amp;apos;s TPUv7 (Ironwood) appeared in the round for the first time.&lt;/li&gt;&lt;li&gt;Results came from 24 submitting organizations, which entered 153 systems and produced 250 test results.&lt;/li&gt;&lt;li&gt;NVIDIA&amp;apos;s provisional Vera Rubin results in MLPerf Endpoints are still awaiting MLCommons verification.&lt;/li&gt;&lt;/ul&gt;&lt;p&gt;NVIDIA&amp;apos;s Vera Rubin NVL72 system posted its first MLPerf Training results on 16 November 2026. The company says that, measured per GPU, the system reached up to 2x the performance of its GB300 NVL72 system. MLCommons, the group behind the benchmark, said the v6.1 round was the first to include NVIDIA Rubin-generation GPUs.&lt;/p&gt;&lt;p&gt;MLPerf Training measures how quickly hardware can bring various AI models to a chosen accuracy level. NVIDIA says it swept every benchmark in v6.1.&lt;/p&gt;&lt;p&gt;NVIDIA also reported results on scaling. In the DeepSeek-V3 benchmark, it says Vera Rubin NVL72 went from 256 to 1,024 GPUs at about 97% efficiency, and four times as many GPUs cut training time from 16 minutes to about four. NVIDIA ran the same benchmark on 10,240 GPUs in GB300 systems, the largest scale in the round, and Microsoft Azure matched it with 10,240 NVIDIA GPUs in GB200 systems.&lt;/p&gt;&lt;p&gt;According to MLCommons, 24 organizations submitted results, including AMD, Google, Dell, HPE, Oracle, CoreWeave and Crusoe, a first-time entrant. Together they entered 153 systems and produced 250 test results. Cloud-based systems made up 37% of the total, compared with 29% in v6.0, and a dozen entries ran on at least 1,000 accelerators. Hardware came from NVIDIA, AMD and Google, and Google&amp;apos;s TPUv7 (Ironwood) also appeared for the first time.&lt;/p&gt;&lt;p&gt;Version 6.1 adds two new benchmarks. One is DLRMv4-HSTU, a recommendation test that replaces the earlier DLRM-dcnv2 and trains on the Yambda dataset. The other is Qwen3.5-397B, a large-scale workload for post-training a large language model.&lt;/p&gt;&lt;p&gt;NVIDIA also entered MLPerf Endpoints for the first time. The figures for Kimi K3, a large open model designed for agent-style workflows, are provisional. NVIDIA says they are pending MLCommons Association verification and may change after peer review.&lt;/p&gt;&lt;h3&gt;Sources&lt;/h3&gt;&lt;ol&gt;&lt;li&gt;&lt;a href=&quot;https://blogs.nvidia.com/blog/mlperf-training-endpoints-vera-rubin&quot;&gt;NVIDIA Sweeps MLPerf Training and Endpoints as Vera Rubin Debuts With Leading Performance&lt;/a&gt;, NVIDIA Blog&lt;/li&gt;&lt;li&gt;&lt;a href=&quot;https://mlcommons.org/2026/11/mlperf-training-v6-1-results/&quot;&gt;MLPerf Training v6.1 Results: Record Participation&lt;/a&gt;, MLCommons&lt;/li&gt;&lt;/ol&gt;&lt;p&gt;ITANU’s newsroom system produced this article with AI and checked it against the sources listed above.&lt;/p&gt;</content:encoded><media:content url="https://itanu.org/assets/magazine-img-a1828ad62e1dda00-1166.jpg" medium="image" type="image/jpeg" width="1166" height="948"><media:credit>Will Buckner / Wikimedia Commons</media:credit><media:description>A white Nvidia sign with a green logo and street address stands among trees and shrubs on a corporate campus.</media:description></media:content></item></channel></rss>
