There's not really a simple answer. If you can take advantage of it using a RenderTexture to cache rendered objects can be a big win. It lets you draw quite complex things into the RenderTexture, which then can be drawn as a simple Quad/Image.
But there is no benefit if you have to redraw the RenderTexture every frame; it's just extra overhead then. If you don't have to redraw all of it – maybe only have to update a few areas of it – then you could still see a performance gain, but then you are at the point of doing some testing/careful measuring.
Apart from this there can be other benefits to using a RenderTexture. You can better use BlendModes in one, to achieve effects that would be more expensive to do other ways. You can combine more things to be drawn into your RenderTexture, to reduce the number of Texture/draw calls in your app.