KPTI + Retpoline Linux Benchmarking On Older Clarksfield / Penryn ThinkPads
First up was the SQLite test that for some systems has seen a measurable impact due to the KPTI + Retpoline overhead on I/O / system calls. But with these ThinkPads, the T61 didn't see any change, the W510 was 10% slower, and the X1 only 3% slower. The impact on I/O tests have found to be correlated depending upon the speed of the storage device if it's already a bottleneck, which may explain why the T61 didn't see any impact on performance with its old, original hard drive.
With FS_Mark the results were close to the same while in the larger test case the X1 Carbon with its solid-state storage saw a 6% drop in performance.
Similarly with CompileBench, the X1 Carbon with its fast SSD storage saw the most significant impact from booting the patched kernel.