Your maths is incorrect, dividing by the number of devices doesn't make sense in this case, you should be multiplying them. E.g.

- 466 minutes of fourth-generation TPU time.

- 1,600 minutes of third-generation TPU time.

Using this logic, fourth generation TPUs are 3.4x better. But, comparing different cluster sizes is pointless. These things don't scale linearly.

you are most likely correct! I read ´Training took 1.82 minutes with 256 fourth-gen TPUs’ to mean Using all 256 TPU it took 1.82 wall clock time to do the training not 466 minutes.

This mean each TPU contributed to (1/256) assuming it scale linearly.

Even if it doesn’t the 256 TPU ran in parallel not sequentially right ?

I think your math is backwards. The training workload W is the same, and the time to complete it is:

W / (256 * speed_v4) = 1.82

W / (4096 * speed_v3) = 0.39

speed_v4 / speed_v3 = (4096 * 0.39) / (256 * 1.82) = 3.43

Note that this assumes that training speed is perfectly linear with the # of accelerators which is not true as you get to very large #s (like 4096!). So the true number should be smaller than the 3.43x above, and the reported 2.7x makes sense.

its only slightly (74 times) slower :)