Have you repeated the test, watching it from the start to the end? Glitching out during one part (power shortage/overload/overheating => freeze or throttle) could affect the average score quite significantly. Not sure about 2017 but some years back unclean driver reinstalls could do silly unpredictable things like slower performance with no obvious glitching, so you wouldn't even know something was wrong if you didn't know how fast the relevant card was supposed to perform. Or perhaps something's overriding DirectX version selection or you have some quality enhancements enabled in drivers and overriding the benchmark's own settings? I'd grab some other tests to see, perhaps switch all driver settings to controlled by application.