The calculated composite scores (summary scores at the top) aren't linear so the results aren't intuitively comparable. If A scores 8.0 and B scores 10.0 in GPUBoss, the expectation is that B is 20% faster than A, but thats not the case.
This can be proven by computing the performance ratio for the composite score and the average performance ratio for the games used to calculate the composite score (listed below the composite score). The two numbers should match up, but they don't.
The site is set up for cross architecture comparison of irrelevant specs which is very misleading for anyone who isn't sufficiently experienced.
The site drops irrelevant quote bombs.