Dont waste the m.2 slot on intel smart response, use a cheapo ssd for that on a sata port.
I've been all through this. When I first built the 2600K system in my sig, it was the initial configuration. I had a 600GB WD VelociRaptor cached to a Patriot Pyro 60GB SSD. The VelociRaptor died, and there had been other problems later proving as a consequence of a secondary power-saving BIOS setting that confused the issue. But the storage configuration was all replaced by an 840 Pro.
Intel had sold their ISRT caching as a proprietary feature of the chipset and controller. Samsung had pushed RAPID (caching to RAM) as proprietary: it wouldn't work with a Crucial or Intel SSD. And finally, Marvell's Hyper-Duo tiered caching depended on their controller as well.
But there was really no technological barrier to making all of these solutions independent of the hardware brands and designs.
I found two software programs which do just that. PrimoCache, by Romex Software (Shanghai), and SuperCache from SuperSpeed in Massachusetts:
http://www.romexsoftware.com/en-us/primo-cache/index.html
http://www.superspeed.com/desktop/supercache.php
I tried both, but I might have foreseen my ultimate choice just by reading the overview capabilities of each. PrimoCache at least gives you the option for a slightly more hazardous caching of deferred writes. And SuperCache doesn't replace ISRT or Hyper-Duo with an "L2" or SSD caching option.
PrimoCache does it all. You can cache several HDDs to a single SSD, when Hyper-Duo and ISRT require at least one SSD per HDD. You can cache the SSD-HDD cached combination to RAM. It doesn't limit you to the ISRT 64GB SSD limit for the caching SSD. You could, if you wanted, cache a 6TB HDD to a 250GB SSD.
It is storage-mode agnostic, unlike ISRT which requires RAID-mode, and unlike Samsung RAPID which requires AHCI-mode. You can have one controller configured for RAID mode, and another controller configured for AHCI mode. You can cache both the RAID array and a single AHCI HDD/SSD as separate caching tasks.
Better than that, even with the deferred-writes "walk-on-the-wild-side" configuration, it actually adds to performance, stability and reliability by allowing RAM-cache contents to be written to disk at shutdown or restart, and resurrected at boot time. The extra time at boot-up or waking from hibernation to reach the desktop is a small price to pay for the advantage prevailing during the desktop session. If there's a 50% hit-rate for the cache at restart or shutdown, an hour or two after boot-up restores a high hit-rate that may have taken many more hours of session usage to achieve before the previous shutdown.
Of course, your hardware must all be in good working order, and that caveat applies to ISRT, RAPID, Hyper-Duo or anything else.
For the time being with Gen2 and Gen3 processors and chipsets, I've begun to use Primo for most -- not all -- of the home workstations.
If you're using RAPID with a Sammy drive, you would simply turn it off. You can deploy RAPID alongside Primo -- they will co-exist comfortably. But why do that? You'd only leave Magician installed to provide information about a Sammy SSD under SMART.
I can move my lifetime licenses to Z170 systems, which of course could be fast enough to only reduce the usefulness of Primo, but would not eliminate its advantages altogether.
The only drawback, for penny-pinching enthusiasts determined to get what Intel and Samsung seem to provide for "free," is the price-tag of about $30 for a single-PC lifetime license.