There are 4 goals to accomplish for a successful delid:
(1) Replace Intel's TIM with liquid metal.
(2) Eliminate any space between the IHS and the Die.
(3) Ensure that the IHS does
not move once it's placed on the Die.
(4) Decrease the temperature differences between Cores.
First; you
must completely clean
all traces of the old sealant from the substrate (green PCB) and the IHS. The original temperature problem is not only caused by poor thermal conductivity of Intel's TIM, but it's
also due to the sealant being slightly too thick, which creates space between the IHS and the Die. The greater the space, the thicker the TIM and the higher the Core temperatures.
Second; when both surfaces are completely clean, as a visual test, you can place the IHS on the Die, where you'll notice that when viewed from the side, the IHS does
not sit on the substrate; it instead sits directly on the Die. You'll also notice that there's a space of only 100 to 200 microns between the IHS and the substrate. This verifies that there is
NO space between the IHS and the Die, which is exactly what you're trying to achieve. If the old sealant is not completely removed, you may end up with an imperfect delid job where you still have space between the IHS and the Die, with perhaps one end being higher than the other causing uneven Core temperatures.
Third; from the instant the IHS makes contact with the Die, the IHS must
NOT be allowed to move or rotate in
any way. This ensures that the liquid metal does not "smear" which can create voids or imperfection in an otherwise perfect application. Smearing the liquid metal will cause Core temperatures to be higher than if the IHS never moved. Moreover, when latching down the socket clamping lever, it's very difficult to keep an unsecured IHS from moving.
Fourth; smeared liquid metal can create poor thermal conductivity between the Die and the IHS in certain locations on the Die, which can cause exceedingly uneven temperatures between Cores. Intel's specification for DTS (Digital Thermal Sensor) accuracy is + / - 5°C. So, when all Cores are under a steady 100% workload, temperature differences between the highest and lowest Cores should not exceed 10°C.
Therefore, in order to avoid defeating the purposes of delidding, and to assure the best possible outcome, it's critical that the IHS does
NOT move from the moment it contacts the Die to when it's securely clamped into the socket. So yes, the IHS must be sealed. A bead of sealant under the corners of the IHS and under the clamping rails on the sides of the IHS is adequate. It's also important that you allow ample time for the sealant to completely dry before clamping the processor into the socket.
If the job is done well, you can expect a decrease in Core temperatures of about 16 to 20°C at 100% workload.
CT