Method of operation

Action Replays that were capable of saving the system's state to tape or disk all operated in broadly the same way. By attaching to the computer's memory bus (via the Zorro expansion slot on the Amiga), all memory access by the processor could be monitored. By keeping track of all writes to hardware registers (for example, to the video or sound hardware) the Action Replay could keep a complete copy of the state of all those registers in its own internal memory. This state could then be saved, along with the contents of the computer's RAM for later re-loading. By pressing a button on the Action Replay, an interrupt loaded a special monitor program from the Action Replay's ROM which could then be used to load, save and modify the computer's memory. It was even possible to alter CPU registers of the paused code, and later revisions included a complete disassembler. Some developers used this feature when writing their own games. Cheating was possible by altering values stored in RAM.

A knob on the Amiga version of the Action Replay MKIII allowed games to be slowed down, to make them easier. This worked by intercepting vertical blank interrupts and running a wait loop until at least the next vertical blank. This did, however, have the limitation of only being able to reduce game speed by 50% or more.

The Action Replay was a formidable opponent for anyone trying to prevent their game code or graphics being ripped, or their game saved for later re-loading. A few weaknesses were discovered in the Amiga version. It was possible to detect when the interrupt (hardware level 7, unmaskable) which the Action Replay used had been triggered, but only after the Action Replay monitor program had exited. At best the programmer could then choose to crash his program deliberately, making saved copies crash as well. However, it was possible to patch code which did this using the monitor program, so even that was not a total defence. Another technique involves using the CIA time-of-day clock alarm to detect when more than a certain amount of time has passed without it triggering an interrupt, implying that the Action Replay monitor program was running. Later revisions of the Action Replay defended against this by resetting the CIA time-of-day clock. Perhaps the only undefeatable method of foiling the AR is to use a one-time co-processor program to alter hardware registers. The AR is unable to monitor writes by the co-processor, but it can read the co-processor program to determine what writes it makes. By using a one-time program to set up hardware registers, then changing the program to one that ignores them it becomes impossible for the AR to know the true state of those registers (which are write only). Using this technique the programmer could blank or corrupt the screen on leaving the AR monitor. Yet another method of confusing the AR graphics ripper was to set a LACE bit on a low resolution image, thus confusing the AR ripping code which would believe to be dealing with high resolution graphics, and would consequently incorrectly "rip" the image.