Matrix Multiplications on GPUs Run Faster When Given "Predictable" Data (2024)

- **Power limitscan improve LLM performance**  
  > "People have been noticing the effects of this in local LLM inference. Power limiting seems to improve overall performance!" — gdevenyi

- **Hardware constraints drive software redesign and efficiency**  
  > "Constraints require optimizations and rearchitectures... They will need to make do with what people have, a ps5/pro or similar in PC power." — gchamonlive  
  > "I actually think it is a good thing to introduce constraints to AI and the overall tech industry. Hopefully everyone will have to look at improving performance without having to add RAM or increase CPU/GPU performance." — aNoob7000

- **Modern systems are more complex than simple answers like branch prediction**  
  > "I went in expecting to find 'branch prediction' as the answer, but apparently things are even more complex nowadays." — nzach