I was thinking about yield variance and the efficiencies of the various cores. This came to mind as I was remembering the days of yore when you could buy processors marketed as being over clockable ( and one hack that springs to mind was drawing a line with a graphite pencil between two exposed pads on one particular processor which enabled overclocking )The SoC is the same. The 7 core are rejected 8 core with a failure in one of the GPU. It's called binning. The rejected core is disabled after test with a fusible link.
Going back to the yield variance, the Geekbench and Cinebench charts for the M1 processor do show a range of scores with some overlap. I did the thermal pad mod on my base 8/7/256 Air and was consistently getting benchmark scores that were the equivalent to the 8/8 scores on the M1 Pro.
I wondered if, for example, one of the M1 graphics cores could test below specification, resulting in binning and allocation to the 8/7 models, while the remaining processor cores were actually of the highest quality.