Here are a few points to use in order to increase performance, sorted by priority:
BUFFER SIZE: This doesn't compromise any aspect of audio quality. Thus it is the first place the user should try to optimize system performance.
A buffer of 512 is optimal. Lower buffer sizes will increase the peak CPU consumption. Higher Buffer sizes may reduce the average CPU consumption, but the peaks will be similar.
Beware that not all hosts have friendly default buffer settings. In fact, ASIO host users use small buffers for lower latency, as do people who use virtual instruments.
CPU MODE SWITCH
The CPU Mode switch at the top left allows switching to low CPU mode. This can cut peak consumption by a third. This mode gives excellent results, and the quality difference will be negligible for most applications. It is not the purist path, but it is a fine way to convolve the original sample with good accuracy.
COMPONENTS
Efficient Stereo consumes the same amount of resources as the Mono to Stereo, and about half of what the Full Stereo consumes. While it presents a compromise on the purist natural reverb, the overall sound quality isn't compromised in any way, and it is often a more practical choice. You may like Efficient Stereo as much as the Full Stereo in most cases.
CONVOLUTION LENGTH
You can set the convolution control to a shorter time. While the compromise is obvious (shortening the Reverb Tail decay time or gating it at a certain level where it would naturally continue to decay), please note that the noise floor in
The original convolution length shows in the IR properties display. Sometimes shortening by a bit presents negligible compromise on sound and meaningful savings in CPU, but the savings is in steps.
In other words, while not every millisecond will free some CPU, every 100ms should. You can get great results by using a shortened convolution length and using the envelope to smooth the slope so that it doesn't sound gated. On continuous material it
Waves