Like most computers, the Axe-Fx has a limited amount of CPU or processing power. CPU is used when blocks are added to presets. Different blocks use different amounts of CPU, and within the same block different “Types” or settings can also use different amounts.

There are actually two TigerSHARC processors in the Axe-Fx II: one is dedicated to the Amp blocks, and the other handles all other blocks. A completely blank preset uses about 9% CPU – this accounts for the Display and a few other background processes. Adding a single Amp block raises the CPU only to about 12%. Compare this to a Reverb block which can raise the CPU to about 20%.

The Axe-Fx II functions normally till about 90-92%. At that point, the CPU is being overloaded, having to process too many things at once. It’s usually best to try to not go above 90%. The Axe-Fx is programmed to prioritize audio at that point, so the display may seem to lag, preset changes may become slow, and there may also be crackle’s and pops in the audio signal. A warning will flash on the screen displaying “Excess CPU Usage! Reduce Load!” This means that somehow you must reduce the CPU usage in your preset.
There are many ways to do this, but one of the easiest is to change Cab blocks from Stereo or Ultra Res to Normal Res. For many, going from Ultra Res to Normal Res has very little to no affect on the sound when playing live – people have said in studio recording situations, they can hear the difference. Of course going from a Stereo Cab to a Mono cab will cause a noticeable difference if you were using extreme panning or 2 different cabinet sounds.

Another easy method is to simply delete (replace with Shunts) any blocks that you aren’t really using in the preset. Even if a block is Bypassed, it still utilizes the same amount of CPU so that when it is Engaged; it is ready to go and the CPU is allocated for it so there are no surprises when it is Engaged. If you aren’t using a Reverb block that’s in a preset or something similar, replace it with a Shunt (the blank block with a line through it) and you’ll reduce the CPU, allowing for something you might actually use.
More methods are posted in the Axe-Fx II Wiki here:

Using the USB functionality on the Axe-Fx II also uses a bit of CPU – about 2-4% more – so keep this in mind if you are connecting to a computer to use Axe-Edit or record via USB.

Many ask why there are limits to the number of types of blocks you can use in the Axe-Fx II, for example why are there only 2 Reverb blocks, but 4 Volume blocks. Why can’t we just add more and more of the same type of block until we run out of CPU? Forum member “Zer0th” took a stab at the answer, and it was confirmed by Cliff of Fractal Audio:


I suspect that in order to minimise switching time and [many] other overheads… it isn’t case of ‘oh, this newly selected preset uses a flanger, I’ll create an instance of a flanger object’ but rather that he instantiates all of the allowed instances of all the allowed block types in memory at power on and simply* switches the internal virtual wiring per preset.
Well, that’s my guess anyway.
[*] relatively simple… no dynamic destruction/creation of blocks required or memory management on a change of preset