{"blog_title":"Vengineer\u306e\u5984\u60f3","title":"\u30e2\u30b8\u30e5\u30fc\u30eb\u5316\u3092\u5fb9\u5e95\u7684\u306b\u3057\u305f NVIDIA Vera + Rubin Compute Tray ","provider_name":"Hatena Blog","author_name":"Vengineer","author_url":"https://blog.hatena.ne.jp/Vengineer/","width":"100%","categories":[],"published":"2026-01-14 08:00:00","type":"rich","height":"190","description":"\u306f\u3058\u3081\u306b CES 2026 \u306b\u3066\u3001\u767a\u8868\u306b\u306a\u3063\u305f\u3001NVIDIA Vera + Rubin Compute Tray \u304c\u51c4\u304b\u3063\u305f\u306e\u3067\u3001\u8a18\u9332\u306b\u6b8b\u3057\u307e\u3059\u3002 \u8a73\u7d30\u306f\u3001\u4e0b\u8a18\u306e\u52d5\u753b\u3067\u3002 www.youtube.com \u3053\u3061\u3089\u3067\u3082\u3002 NVIDIA unveils Vera Rubin : up to 10\u00d7 lower cost per million inference tokens and 4\u00d7 fewer GPUs for MoE training vs Blackwell, plus ~5\u00d7 better network power efficiency via Spectrum\u2011X Ethernet\u2026","blog_url":"https://vengineer.hatenablog.com/","version":"1.0","provider_url":"https://hatena.blog","html":"<iframe src=\"https://hatenablog-parts.com/embed?url=https%3A%2F%2Fvengineer.hatenablog.com%2Fentry%2F2026%2F01%2F14%2F080000\" title=\"\u30e2\u30b8\u30e5\u30fc\u30eb\u5316\u3092\u5fb9\u5e95\u7684\u306b\u3057\u305f NVIDIA Vera + Rubin Compute Tray  - Vengineer\u306e\u5984\u60f3\" class=\"embed-card embed-blogcard\" scrolling=\"no\" frameborder=\"0\" style=\"display: block; width: 100%; height: 190px; max-width: 500px; margin: 10px 0px;\"></iframe>","url":"https://vengineer.hatenablog.com/entry/2026/01/14/080000","image_url":"https://cdn-ak.f.st-hatena.com/images/fotolife/V/Vengineer/20260107/20260107112702.png"}