Another test with cubic versus lapsrn, both upscaled 8x from the same image:
Is it to be expected that often the models aren’t as good as cubic?
