Higher speed DDR3 memory can make a difference in Haswell too. I get a pretty nice bump to minimum framerates by replacing a DDR3-1600 kit with a DDR3-2400 kit. It's almost 10% in GTA V and more than 15% in Fallout 4 for me in the problem areas that dragged my framerate to its lowest, I mean these framerate differences were easily reproducible for me in both games as they happened in very specific spots. I was pretty shocked to see the difference in
Digital Foundry's tests and it motivated me to buy a Z97 board and the fastest RAM with good price to performance to try to make up a bit for running a locked 3.6 GHz Haswell quadcore (Xeon E3-1231v3). (I chose DDR3-2400 since it cost almost the same as DDR3-1600 but DDR3-2666 was way more expensive than DDR3-2400)
And as I said before, it made a pretty big difference. It doesn't do anything for games like Shadow of Mordor or Far Cry 4 that are straight gpu bound by my GTX 970, but GTA V can hit the cpu hard in areas and Fallout 4 hits the cpu hard everywhere. I'm kind of disappointed none of the tech sites uncovered this until after I had already bought an H81 board and a locked processor though, otherwise I would have probably bought Z97 from the start and a i5-4690k or i7-4790k (I built my system a little after Devils Canyon came out).
But it does nothing for my average framerates, since I'm mostly gpu bound in GTA V and since Fallout 4 runs with automatic vsync to 60 fps. But having much smaller dips in framerate at the worst cpu bound areas of those two games is pretty noticeable to me.