Side-channel vulnerabilities routinely attack performance optimizations that reveal information about the data and/or code.

How much of the difference is attributed performance lost to recently patched side-channel vulnerabilities?