How is AHB faster than APB? The transfer will be a normal single transfer.

If you ignore all the benefits of AHB over APB, AHB is not faster than APB, both will take a minimum of two cycles to complete a transfer.

But that's like comparing the speeds of a speedboat and a rowing boat if you take away the benefits of the speedboat's powerful engine, they are both then just slow boats needing manual propulsion.

So yes, if you are only interested in a single transfer, both protocols can complete a transfer in two clock cycles, but if you look at a sequence of transfers the pipelining of the AHB protocol means that each additional transfer might only add one clock cycle (compared to an additional two for APB), and the HBURST information on AHB might then allow a slave to process transfers more efficiently if it knows in advance how many transfers will be requested (prefetching read data or buffering write data), whereas APB slaves just process each transfer in a burst as a single unrelated transfer (so no opportunities for efficiencies).

But APB is simpler to design with, compared to AHB, so if performance is not critical and simplicity is, that's an advantage for APB (where you might locate a lot of low bandwidth peripheral slaves such as timers, UARTs, serial devices, keyboard interfaces, and so on).

If you ignore all the benefits of AHB over APB, AHB is not faster than APB, both will take a minimum of two cycles to complete a transfer.

But that's like comparing the speeds of a speedboat and a rowing boat if you take away the benefits of the speedboat's powerful engine, they are both then just slow boats needing manual propulsion.

So yes, if you are only interested in a single transfer, both protocols can complete a transfer in two clock cycles, but if you look at a sequence of transfers the pipelining of the AHB protocol means that each additional transfer might only add one clock cycle (compared to an additional two for APB), and the HBURST information on AHB might then allow a slave to process transfers more efficiently if it knows in advance how many transfers will be requested (prefetching read data or buffering write data), whereas APB slaves just process each transfer in a burst as a single unrelated transfer (so no opportunities for efficiencies).

But APB is simpler to design with, compared to AHB, so if performance is not critical and simplicity is, that's an advantage for APB (where you might locate a lot of low bandwidth peripheral slaves such as timers, UARTs, serial devices, keyboard interfaces, and so on).