[BeeOnRope]

> there are no architectural changes between Broadwell and Haswell in the core

There were certainly some changes. The gather instructions were dramatically improved, taking ~5 uops instead of ~30 and with much better throughput.

Conditional moves only take 1 uop in Broadwell, down from 2.

Some other changes listed here:

http://users.atw.hu/instlatx64/HaswellvsBroadwell.txt

Intel's tick-tock model was never black and white: even the "ticks" (node shrinks) received some changes and even new instructions.

[DiabloD3]

Sorry, I should have written major changes. In practice, benchmarking identical Haswell vs Haswell Refresh (that, again, are effectively DDR3 Broadwells), such as with an E3-1230v3 vs a E3-1231v3, I did not see anything that wasn't within a reasonable margin of error.