Yeah, you kind of do since it is a contributor to a gap between the ihs and the die. It is raising the ihs up off the die a tiny bit of you leave it on.
Time it takes? With a tool? 30 seconds to pop the top off, 30 minutes to an hour to clean the black goop off, a minute or two to apply new paste, and then 8-24 hours while your glue cures to hold it on again.
With a razor blade? An hour or so for the first step if you’re being very careful with it. I have done one with a razor. I wouldn’t again.
Make sure you select an appropriate replacement material for under the IHS. Many pastes will pump out due to the thermal expansion and contraction of the die and/or the IHS unlike what you see between IHS and heatsink. This is a huge pain to deal with so choose well. I like to use a liquid metal option (CLU) in this role because I know it will last.
Delidding can drop temps 10 C to 20 C, depending on CPU. In my case, I didn’t have a tool so I sent my 8700k over to Silicon Lottery to have them delid and bin it. I’ll probably build the system this weekend and see if I can tweak the settings a bit to get lower voltage.
My "gain" for CLU-re-lid was 12C. I'm a consummate DIY'er, and the cost of the "tool" mentioned is insignificant.
I initially justified paying Silicon Lottery the $50 to avoid an inconvenience. It only helped my decision that I bought the CPU from Silly-Lotts: there would be a delay in the delivery of the CPU whether I bought it from them or from NewEgg (etc.) An additional justification was the inference that the person or persons behind Silly-Lotts had performed the operation 100s or 1,000s of times already.
Further, as substantiated mostly by Silly-Lotts mention of refining the gap between IHS and CPU die, it seemed important to get this aspect right the first and only time, so going into a DIY episode without previous experience with it prompted me to lean further toward "outsourcing" the de-lid/re-lid.
absolutely, specially if the CPU is kinda old. I own a 6600k since november 2015 and the TIM was severely degraded (thanks Intel! /s), to the point that I've improved 30ºC in package temps after delid + liquid metal. This is how the TIM looks and performs in a 2-year old 6600k:
and this is how the same CPU with the same overclock and Vcore performs with liquid metal between IHS-die (I still used Artic Silver 5 seating the cooler, no double liquid metal interface):
Cores #0 and #2 are no longer toasting at +16ºC over cores #1 and #3, and I can push the OC all the way up to 1,45V 4.8Ghz without being thermally constrained. I'd say delidding these CPUs is a must
Are you using an AiO, custom-water, or heatpipe cooler?
I'm at 4.7 Ghz on a CLU-relidded Skylake i7-6700K. My peak package temperature under IntelBurnTest "High" are about 18C higher than you show with the 6600K @ 4.8. So while I don't question that your 50C result is accurate, I question instead any assumption I could make that the screenie was taken with a heatpipe cooler comparable to mine.
However, I am running with HT Enabled.
So if you're under water-cooling, which AiO or custom components?
Even so, my tests under "Maximum" IBT and Affinitized LinX "max prob size; max memory" show peak package C about 2C lower than an unmodified retail processor at 4.7 cooled with an EXOS external dual-radiator assembly as shown in a published review.
So . . . I'm just curious. . . . .