Thermal paste vs STIM solder: this is how they perform on the i7-8086K vs i9-9900K.
Contents
Just three days ago we offered you a review of the long-awaited Intel Core i9-9900K, a CPU that will be the top of the desktop range for the blue giant. Many of the reviews that have been published have given very even results in a very specific and controversial scenario, the temperature resulting from using solder vs. thermal paste, how much is the improvement?
Intel wouldn’t have done a great job with its STIM
As we say, the large number of reviews published all over the Internet leave no doubt that the Intel Core i9-9900K is a very hot CPU, especially to include solder between its die and IHS.
The guys at Techspot have gone further and wanted to compare what is the real improvement of using a solder vs. using the typical paste that Intel was administering to its CPUs in socket 1151.
We start from the premise that the i9-9900K is a microprocessor with 8 cores and 16 threads that comes very tight in frequencies, where it is capable of reaching 5 GHz in two cores while stock.
Thus, temperatures above 80 degrees will be obtained with high-end heatsinks or AIOs, while with RL custom we will be a little below.
As can be seen, with overclocking the temperatures only get worse and scale a lot, to the point of reaching 100 degrees with a 5 GHz configuration on all cores and it is only possible to exceed that frequency with liquid cooling even at the cost of being around these figures.
The comparison revolves around another CPU, the well-known i7-8086K, a six-core, 12-thread CPU that does not include soldering between its die and IHS. To make everything more fair, 2 cores have been disabled with their respective threads in the i9-9900K and they have matched frequencies and voltages to be able to measure the difference.
The motherboard used was an MSI Z390 Godlike while the stress software was Blender.
The results are worrisome under overclock
Starting at the lowest settings, both 4.5GHz and auto voltage CPUs achieved temperatures of 61 and 72º respectively with voltages of 1.16v for the i9-9900K and 1.26v for the 8086K. This reduction of 11 degrees of temperature means a 15% difference in favor of welding, but it is true that the voltage was 8% lower.
Going up to the section of blocked voltage at 1.35v and 4.5 GHz we again find significant differences: 87 vs 78 degrees, a difference of 9º in favor of the i9-9900K, which gives an improvement percentage of 10%.
If we go to overclock, at 5.1 GHz the differences are greatly reduced, where we find a temperature of 94 degrees for the 8086K and 91 degrees for the i9-9900K, only three degrees of difference.
But how can this be possible? To find an answer we will have to go to the video of the overclocker Der8auer, where you can see several details that glimpse said answer.
Higher die thickness, too much sealing compound and less thick weld
In the video you can see how Intel has made the die much thicker than that of the eighth generation of its processors, such as the 8700K, this implies a worse heat transfer from the silicon to the IHS. Remember that silicon has a thermal conductivity of 148 W/mK, while IHS copper gets 385 W/mK (although it has a very thin layer of electrolytic nickel (52.3 W/mK)).
In addition, and as seen in the video (as has been the case since the 3rd generation of Core CPUs) Intel applies too much sealing compound to its CPUs, making the IHS and the die not maintain good enough contact and affecting temperatures.
This seems to indicate that the greater thickness, the excess of sealing compound and a weld that does not seem to be very thick (with the consequent long-term problems in terms of microcracks) prevent improving temperatures. So much so that Der8auer himself, by removing the remains of welding and leaving the IHS clean, together with a lapping of the internal height of the IHS (to solve the lack of welding) together with the application of Thermal Grizzly Conductonaut, achieved a drop in noticeable temperature.
Many speculate on the reason for the increase in the thickness of the die by Intel, but the truth is that said increase is not seen curiously from the 2600K that also came soldered, so it is stipulated that said increase may come due to the need to use your STIM instead of conventional thermal paste.
To this we must add the fact that its welding seems to be less thick and therefore with less quantity, this may be a factor to take into account, since in many already made delids there is a poor distribution of it due to the die and therefore it is a direct failure of the blue giant.
