Thanks for replys, CJ, Ben and Benny.
I did some testing with average power draw, and average hash rate, over different power targets/OC
a while back, using nicehash (so varying payrates)

It gets complicated rather quickly, but I do see your points
What I'm wondering is where to draw the line, pushing it for example, from 218mBTC to 239mBTC (720sols to 770sols) on one 80ti would actually cost more in electric, but gets more mBTC
Not too concerned about the longevity as 5y warranty, and I only have a few, so return and replace would not cost me anything significant downtime.
UK, so even now in July heat is not an issue. (cry)
More cards on a circuit, smaller psu's and other electrical concerns, yes totally understand that, things are different for us all.
My calculations are probably off, I'll have to try them again, or keep doing them as things change and adjust the power accordingly. There are a lot of variables, I'm kinda trying to reduce the complexity some.
Regards
It's hard to tell because our cards might have different PCB and different power layouts and defaults. But suffice it to say you probably aren't going to want to run it at 100%, where that sweet spot is for absolute optimum will probably depend on the individual card, pay rate, projected increase in hash power, how bullish you are on the coin, temperatures, and probably other things.
You're obviously really trying to dial it in. Gotta go card by card I would imagine. Not sure how much variance there is between individual cards in your model.