And yet globally both of these companies spend 20% of CPU cycles on TLB misses. Should we turn off virtual memory protections and go back to raw physical memory

And yet globally both of these companies spend 20% of CPU cycles on TLB misses. Should we turn off virtual memory protections and go back to raw physical memory (or something?)