As large generative AI models transition from training to large-scale inference deployment, the exponential growth in computing power demands is pushing chip power consumption to unprecedented levels. NVIDIA’s next-generation Rubin GPU already consumes 2,850 W of power, while the Rubin Ultra v...
더보기























