The only side effect of changing the audio buffer size that I noticed is delay between keypress and sound. CPU load is the same. When I play music its sounds the same (in my opinion). So, what is negative side effect from decreasing buffer size? If no, why the setting is needed at all?
So, what is negative side effect from decreasing buffer size? If no, why the setting is needed at all?
No negative side in my opinion - when your system is able to run smaller buffer sizes without any glitches, you can get low latency. Most like low latency, but some don't care ;-). On slow PCs bigger buffer sizes sometimes help for crackle-free sound.
Decreasing the buffer size does increase CPU load. But only "somewhat". The effect tends to be stronger the smaller the buffer gets. However, it is difficult to measure since a part of that load might not be attributed to the software but is seen as operating system overhead.
The main problem in my opinion is not load, however. It is rather the timing jitter inherent to any multi-tasking operating system: you can't be really sure exactly when a program is going to be doing a specific task, as the operating system is always stopping and starting processes. The audio buffer's main purpose is to counteract this. So you usually would use the smallest value that produces clean audio without crackling noises. By increasing it further you will decrease CPU load, but only marginally.
However, when using the absolutely smallest possible buffer for clean audio there is still a (small) chance of getting an occasional crackle or pop noise once in a while, since we are talking about statistics here: the chance will never be exactly 'zero'. But it gets very small very quickly, so check which is the smallest buffer size that works and then increase it a little bit, and you should be fine.