Are you talking about 25% performance uplift in general or 25% performance uplift in gaming (average)?
The latter will be virtually impossible as most games aren't significantly CPU bottlenecked unless you run them at unrealistically low resolutions.
I have never seen a developer intending to write code to
love cache or
love frequency.
If you wanted to write code to
love cache, you would have to write bloated code or de-optimize existing code. Cache sensitivity, especially L3, is often regarded as a symptom of lack of optimization and bloated code.
Any coder should want to make their code scalable with CPU performance, so if the code scales with e.g. frequency it's an indicator of better code (or in your terms "loves" frequency
).
It's not a new platform or a new basic die, so the only thing to watch out for is whether the assembly of the extra cache is unproblematic.