@oldschool, if you do delid, once delidded, use a credit card to gently scrape off all the black adhesive, that took me 15-20 minutes. Most then use liquid metal clu/clp and put IHS back on, ie ihs just held in place with cpu bracket. My temps went down 22C with IHS back on at 1.3v, prime load. If you go bare die, you can lower temps another 3-4C (if testing at high wattage, like 1.3v with prime type loads), but to go bare die have to mod the socket and mod the waterblock, hence not really worth going bare die, too much hassle for too little gain. Not to mention all the issues of chipped dies and intermittent ram channels working from uneven pressure.
@ShrimpBrine, Intel soldered cpus have 87 w/mk thermal interface indium solder. When you delid the 4790k, you are removing 3.5 w/mk intel polymer paste that is applied thick (ie two problems poor thermal conductance tim1 and thick bondline of tim1). Replacing it with 40 w/mk liquid metal tim with a thin bondline, can improve temps 15-20+C (higher overclock/dissipated watts tested, the greater temp gain). And then you can improve temps another 3-4C more by removing IHS and using liquid metal tim between die and waterblock, but with increased hassle of modding socket/waterblock (many now have tested and gotten same few C temp gain between IHS on w/liquid metal vs IHS off).
I have delidded soldered ihs for testing purposes, but would now never delid one for temp gain, as it already has maximized tim1 interface. Dont see the point of delidding solder cpus...for intel anyways.
soldered intel IHS has roughly same temps as running bare die with liquid metal
soldered intel IHS has better temps than running bare die with paste.
The reason is die is small with hot spots, and paste tim at 3.5 w/mk is a huge roadblock when the surface area is very small (individual hot spots on die). With solder the heat is conducted at 87 w/mk into copper ihs (400 w/mk) and heat is spread more evenly via coppers high conductance, and now you have a huge surface area of ihs before hitting crappy paste tim between ihs and die. Tons of people (including myself) have tested bare die vs IHS on, and get same few C difference. And for tim2 (between ihs and waterblock) the tim material isnt as important since surface area is much larger (hot spots somewhat mitigated and ihs larger than die)