Why *= 2 is slower than = col * 2 in Pandas

This problem looks trivial, but I got curious about what different Pandas update strategies actually cost. I separated the benchmark into two layers: Warm in-memory execution — how fast each Pandas u