VERA RUBIN NVL72 CRUSHES BLACKWELL ON AI INFERENCE
■ AI-SUMMARIZED FROM 1 SOURCE ▸ TIMELINE
AMD's Vera Rubin NVL72 accelerator delivers up to 7x better token throughput per megawatt than Nvidia's Blackwell on large language model inference, according to SemiAnalysis testing. The result exceeds Nvidia CEO Jensen Huang's stated 3x efficiency gains for models in the 1T-3T parameter range.
■ MORE FROM THE HARDWARE DESK
Volkswagen has unveiled an efficiency-focused electric vehicle that achieves nearly twice the efficiency of the Lucid Air, the current production benchmark. The automaker borrowed design concepts from Slate to achieve the breakthrough performance.
Valve has not committed to a release window for Steam Deck 2, with the company still evaluating how and when to deliver the next-generation handheld. The company has consistently stated that a sequel will arrive, but only when the right hardware components are available.
Apple has released its latest iPhone and wearable lineup this year and plans to introduce additional products through early 2027. The company's roadmap includes new devices across multiple categories.
Valve has released its Steam Frame virtual reality headset at a $1,059 price point, higher than the company originally targeted. The company attributes the cost increase to global semiconductor and storage market pressures.