In the absence of other information, it's reasonable to assume linear scaling. The curves are quite far apart. You can simply draw a horizontal line and move it up and down see that top x86 performance on Y-axis is exceeded by M1 for a large part of the curve that is drawn (and presumably exists).