This also depends on the device. Reading from or writing to the "current" address is faster than a random access. But continued read/write is typically limited by the actual device (EEPROMs tend to have "pages" e.g of 64 bytes).
Full random write access:
START | 7BIT_DEV_ADR | 0(Write) | ACK(Slave) | 8BIT_ADR_HI | ACK(Slave) | 8BIT_ADR_LOW | ACK(Slave) [ 8BIT_DATA | ACK(Slave) ]xN | STOP
Current address read:
START | 7BIT_DEV_ADR | 1(Read) | ACK(Slave) | [ 8BIT_DATA | ACK(Master) ]xN | 8BIT_DATA | NACK(Master) | STOP
If you count START, STOP and ACK/NACK as one clock, you can calculate with the above information how much clocks you need to transfer N bytes of data.