Is turning the crank faster (faster cpu) ever limited by passing information into the CPU, or better put, how does the local Cache  capacity affect performance versus swapping ram at memory bus speed?

For instance, I do a lot of master sketch modeling, with lots of constraints that require a lot of calculations. My models are generally not more than 30MB, maybe more at the 15MB size. Generally I assume I need to max CPU, but I figure I also want to max local Cache to keep as much information ready for the processor at a time since RAM bus speed is only 1ghz.  When it comes to the tradeoff of i7 6700K vs i7 5930K we get more CPU cycles/sec but also less cache.

Is the CPU ever data starved when running these constraint calculations?

For example, if I were working on my 15MB model:

If i were to use the 6700K with 8MB cache my naive understanding would be that I can only load half my model at a time onto my cpu. So my cpu is constantly swapping back and forth at 1ghz with my ram.

But, if i were to use the 5930K with 15MB of cache I could potentially fit the entire model on my cpu, limiting how much swapping is happening with the ram.

Now, there are actually a number of other service happening all the time in windows, so presumably my model is being split up across ram and cache.  But, say I've got a 30s load time on a model, is my cpu ever starving for information?  I have some models that take 10minutes to load, so I would assume those are swapping all the way back to ram or even the HD.

I wonder if anyone has an answer to this tradeoff?