This should be fixed in the 13th gen and via firmware updates on previous gens:
I commented over there, but I couldn’t confirm such an improvement in my benchmarks.
As someone who has been half-reading this discussion as it developed, certainly not doing any testing myself, it would be very interesting to see this process and discoveries written up in a blog format. The tools used, methodology, etc. would an illuminating read. In combination with the recent blog post by frame.work which said,
To address every scenario, we have to go beyond the black boxes of CPU and retimer behavior that we have limited control over and modify the Expansion Cards instead. We found unexpected CPU and retimer behavior in which placing a HDMI or DisplayPort Expansion Card on the same side of the laptop as any card other than USB-C could keep subsystems powered, whether or not a display was connected. To solve this, we’ve modified these cards to now behave as if they are generic, non-display USB devices when no monitor is connected. This, in combination with our system firmware changes, allows full power saving behavior.
I think there is a lot of insight into ways to determine what closed-source personal computing hardware (low-level chips and buses) is doing without schematics, in the case of this thread, and how to work around it in the case of the blog post.