I absolutely agree, and I did a comparison chart to show what tests benefits from NO AA (shown in green), and thus benefits from a faster card. In other words: If the score is not affected by the AA then the card is not running at 100% and the software is the bottleneck (shown in red).
(Edit: GPU_01 is not really limited by the software, but neither limited by hardware)
VAD's GeForce 970 on an i7-4790K:
I absolutely agree that these three tests affects the score WAY TO MUCH. And I will fix that together with other stuff in the upcoming Holomark 2.5 release. I'll also add the text string for fast copy-paste to the also upcoming comparison chart. (I'll make a spreadsheet like I had for Holomark1)