Hector Martin

At a typical GOP interval of 30, you need 30 times the CPU power required to decode a video normally for smooth scrubbing to arbitrary frames. GOPs can be even larger than that.

the computer on the left is 500 times more powerful just in CPU than the computer on the right

And h.264 is probably over 20 times more complex to decode than the older codec. Times 30 you're at 600. And nobody is putting exact seek into browser players because it's not going to work for HD video which multiplies the CPU usage again.

i don't understand what the times 30 is here

h.264 frames can depend on an infinite number of frames before them. A more common setting is 30. That means decoding frame 30 requires decoding frames 1-29 too. So you need 30 times the CPU power to decode any *random* arbitrary frame in the time it normally takes to decode one.

Try encoding your video on the left with GOP size 1 and I bet it'll seek as smoothly as the video on the right.

(It will also be much larger, but still smaller than the video on the right since h.264 is still going to be a better codec even in intra only mode)