LIVE · SAT, MAY 23, 2026 --:--:-- ET

Issue Nº 32 COST TOTAL $11535.05 ARTICLES TODAY 1 TOKENS TOTAL 6.75B

Running the wire

Breaking SpaceX Starship successfully deploys mock satellites in test flight Chips Micron Virginia Fab Begins Producing America's Most Advanced DRAM; Output to Quadruple Chips Huawei ships 122TB SSD using proprietary NAND packaging to sidestep US 3D chip sanctions Chips Enthusiast runs 1-trillion-parameter LLM locally on single GPU with 768GB Intel Optane Policy South Korea deputy PM: AI wealth must benefit public, amid Samsung labor tensions Policy IEEE Launches Medical Mobile App Verification Program for Developers Breaking GeForce NOW Gets James Bond '007 First Light' Game With Ultimate Bundle Market Global Factory Activity Sags on War-Driven Inflation as Central Banks Weigh Rate Hikes Breaking Google Cloud Rolls Out Cross-Engine Iceberg Support in BigQuery Research Uber Improves Restaurant Recommendations Using Real-Time Signals and Listwise Ranking Breaking SpaceX Starship Launches Successfully; Second Attempt Succeeds Chips U.S. quantum strategy shifts focus to hardware utility over ambition Market Qualcomm stock surges on investor rush into AI chip demand Market Reddit stock drops 6% as Meta launches competing forums app Funding This Week's Biggest Funding Rounds: Medical Devices, AI Gadgets, Frontier Labs Lead Policy WA researchers call out state education office for toothless K–12 AI policy Breaking Hard Drive Component Price-Fixing Lawsuit Alleges 13-Year Scheme Chips Commercial space payloads adopt agile screening for high-reliability AI hardware Chips Chinese DRAM maker CXMT enters mainstream with Corsair Vengeance DDR5 Breaking Meta data center allegedly fouls Georgia town water; EPA launches investigation Breaking SpaceX Starship successfully deploys mock satellites in test flight Chips Micron Virginia Fab Begins Producing America's Most Advanced DRAM; Output to Quadruple Chips Huawei ships 122TB SSD using proprietary NAND packaging to sidestep US 3D chip sanctions Chips Enthusiast runs 1-trillion-parameter LLM locally on single GPU with 768GB Intel Optane Policy South Korea deputy PM: AI wealth must benefit public, amid Samsung labor tensions Policy IEEE Launches Medical Mobile App Verification Program for Developers Breaking GeForce NOW Gets James Bond '007 First Light' Game With Ultimate Bundle Market Global Factory Activity Sags on War-Driven Inflation as Central Banks Weigh Rate Hikes Breaking Google Cloud Rolls Out Cross-Engine Iceberg Support in BigQuery Research Uber Improves Restaurant Recommendations Using Real-Time Signals and Listwise Ranking Breaking SpaceX Starship Launches Successfully; Second Attempt Succeeds Chips U.S. quantum strategy shifts focus to hardware utility over ambition Market Qualcomm stock surges on investor rush into AI chip demand Market Reddit stock drops 6% as Meta launches competing forums app Funding This Week's Biggest Funding Rounds: Medical Devices, AI Gadgets, Frontier Labs Lead Policy WA researchers call out state education office for toothless K–12 AI policy Breaking Hard Drive Component Price-Fixing Lawsuit Alleges 13-Year Scheme Chips Commercial space payloads adopt agile screening for high-reliability AI hardware Chips Chinese DRAM maker CXMT enters mainstream with Corsair Vengeance DDR5 Breaking Meta data center allegedly fouls Georgia town water; EPA launches investigation

Chips Saturday, May 23, 2026 at 11:31 AM

Enthusiast runs 1-trillion-parameter LLM locally on single GPU with 768GB Intel Optane

A developer achieved a technical milestone by running Kimi K2.5, a 1-trillion-parameter language model, on a single system using 768GB of Intel Optane DIMM memory paired with one GPU.

The setup achieved roughly 4 tokens per second, demonstrating feasibility of large-model inference on commodity hardware with extended memory.

Read at source →