💻 Technology Oct 31, 2026 · Rahim Amir

View Channel →

0 0 Save Read Full Article Source ↗ 👁 2 💬 0

Nvidia has quietly cut Rubin's listed HBM4 bandwidth from 22 TB/s to 19.2 TB/s per GPU and NVLink 6 from 3.6 TB/s to 3 TB/sThe eight-GPU HGX Rubin NVL8 fares worst, losing 26% of its memory bandwidth and dropping from 2.3 TB to 2 TB of HBM4 capacityOn paper, AMD's MI455X now leads on per-GPU memory bandwidth and Helios on scale-up bandwidth, though neither side's figures are benchmarkedNvidia has quietly lowered the memory and interconnect bandwidth it publishes for Vera Rubin, cutting per-GPU H

Sign in to join the discussion

📰

Mirendil, founded by former Anthropic researchers to build self-improving AI, raised a $1B Series A led by Kleiner Perkins and Clearlake Capital (Sarah Klearman/Wall Street Journal)

Techmeme · 1d ago

📰

Sources: social media companies' payouts in teen addiction suits mostly ranged from $2M-$2.5M; Meta agreed to pay two teens a combined $7M in the past two weeks (Olivia Carville/Bloomberg)

Techmeme · 1d ago

Five thoughts from Cisco’s first quarter of fiscal 2027

SiliconANGLE · 1d ago

Singapore says US authorities, and Supermicro, hardly helped investigation of Jensen Huang’s Chinese friend

www.theregister.com - Articles · 1d ago

📰

Sources: US officials disagree on whether Alice Huang spied on Jensen Huang for China; some at ODNI and CIA say intelligence collected so far doesn't prove it (CNN)

Techmeme · 1d ago

📰

Letter: ByteDance plans to buy back up to $6B worth of shares from existing investors at a ~$400B valuation, a 20% increase in valuation from its 2025 buyback (The Information)

Techmeme · 1d ago