I’ve been talking to dozens of Blender users who accidentally turned on square samples, or did it consciously but didn’t know what it’s doing exactly.
So for those who don’t know exactly what the “Square Samples” setting does in Cycles, here’s a brief explanation 
Short screencast:
Not sure I really see the point of the ‘Square Samples’ setting.
I don’t think I have used more than 1000 samples very often (especially now we have de-noiser) - and for most of the time i’m probably somewhere around the 100-500 mark.
If we consider what square samples actually does
2 samples = 4 squared samples
10 samples = 100 squared samples
30 samples = 900 squared samples
50 samples = 2500 squared samples.
To me, it seems odd to lose the granularity in terms of sampling (and possibly massively affect your render times inadvertently), to save typing perhaps 1 or 2 digits into a box.
What real world advantage does square samples actually offer?
To use that option properly you need to keep your samples at very low values, like 1 to 10. The benefit is finding a optimal number of samples to your scene. If you’re rendering at 1000 samples and double that, doesn’t means that you will get a ‘double’ quality, there is a decay of quality gain over sample amount, and it’s inverse exponential. Other render engines like Arnold (sampling) and Renderman (rendering efficiently) also works like that. That feature is meant for big animantions like movies with thousands frames. Finding that optimal number is part of a feature film pipeline, normally done in the Lightning stage.
I guess this is just closer to one way of mentally representing progress. I’m more of a linear dude myself, tbh
Thanks for the input. Yeah I’m recommending a range of 4-50 if you do turn it on.