Memory bandwidth directly effects any CPU processes which are bandwidth limited. That is why higher speed RAM leads to bigger scores on the performance tests.
If you are running lots of plugins - convolution and algorithmic reverbs etc - unless the whole workspace the plugin is using is in the L2/L3 caches then data is running up and down to main memory. In many situations, most of the CPU cycles for each core in fast modern CPUs are spent waiting for data - the CPU is spinning its wheels. The lower latencies of the caches are aimed at minimising this wastage as far as possible but anything that involves cache misses brings the much larger latency to main memory into play. A big DAW mix, with lots of tracks, with effects, samplers, synths all working at once - is keeping hundreds of tasks running in parallel and keeping all their outputs synchronised and doing this will inevitably involve swapping lots of stuff in and out of main memory.
