I've always thought of the value proposition of the framework as being it's capacity for upgrade.

You should obviously start from a level of compute that meets your requirements, but the additional cost for the framework was always the price of the ability to change out parts later (at the very least the idea of it)

It feels like an apples to oranges comparison but with the only success criteria being the things an apple is designed for (pun intended).

If performance to price ratio was the only criteria they are interested in , the framework shouldn't even be in the race.