As we did in the past with Intel's Core i7-8700K, we again created absolutely identical test and measurement conditions. We use the same type of CPU cooler (Alphacool XPX), the same Thermal Grizzly Kryonaut TIM, and the Alphacool Ice Age 2000 chiller that, as always, provides exactly 20°C water temperature. We weigh the thermal paste (0.15 grams) on a laboratory balance to further ensure accuracy. Thus, our older test results are usable for comparisons to the Intel Core i9-9900K.
Thermal Grease Vs. Solder
We have an important preliminary about the change from the thermal paste TIM on the Core i7-8700K to the solder of the Core i9-9900K. Since the height of the CPU has remained absolutely the same with the new chips, one can also assume that Intel uses the same heatspreader as the older CPUs. The distance between the die and heat spreader was previously relatively high due to the design, as the chips have a relatively thick layer of thermal paste.
Therefore, we can assume the solder layer also turns out to be a bit thicker than it would actually have needed. This fact, and the significantly smaller heat spreader (surface area) compared to the LGA 2066 CPUs, will certainly explain why the results that follow are the same as they were. Good, but not perfect.
The Intel Core i7-8700K achieves just under 160 watts in the stress test with Prime95, so we ran the Intel Core i9-9900K with a similar load. The Core i9-9900K did not always use all its cores fully, so slight fluctuations occur despite the same average waste heat over the entire time. But these remain negligible. Logically, the Intel Core i9-9900K runs much cooler than with Solder TIM:
With nearly identical power dissipation and identical cooling conditions, we calculate a mean package temperature of 57°C for the Intel Core i9-9900K and 75°C for the Intel Core i7-8700K. This results in a delta of 18°. Tests of the delidded Intel Core i7-8700K show that this is not necessarily optimal – the delta was at least 20° (better by 2°). In either case, even the industrial solder solution is always worth more than the thermal paste of the previous CPU generation.
Leakage at different temperatures but same load
CPUs are thermistors, where the internal resistance decreases with rising temperature instead of rising. That makes it interesting to see how the temperatures, and thus the leakage currents (and consequently the power loss), develop with the same applied load. To measure this, we have the Core i9-9900K overclocked under Prime95 run once with chiller and with a normal AiO compact water cooling (Corsair H110i).
The result turns out as expected. While the chiller the CPU is at an average of 63° C, but the 90° C with the AiO is already near the absolute limit. Mind you, this result is at stock settings with an AVX load. Interestingly, when using SSE, the delta of 27°C remains nearly the same, as it is still 25°C on average.
But back to the AVX load, because what does the almost 30° C temperature difference ultimately mean for power consumption? Here, too, we are amazed to a certain extent, because between the 205W with chiller and the 229W with the AiO compact water cooling, there is a difference of 24W. We can only attribute that to the now stronger leakage currents.
You can see the remaining power consumption figures on the previous page, but under almost ideal conditions. The better the cooling, the better the power consumption. However, we were only able to record these large differences at package temperatures above 80°C, which then almost rise like an avalanche. This would dissuade us from air cooling, even if the Intel Core i9-9900K should not be overclocked any further. The 4.7 GHz all-core and a constant load are quite sufficient to make air cooling absurd.
MORE: Best CPUs
MORE: CPU Benchmark Hierarchy
MORE: All CPUs Content
Intel clearly needs 10nm and a new architecture to go back into the game. As is, I struggle to think of any reason to buy the 9900k.
(By the way, I totally saw this coming what with the crazy clock speeds they were pushing.)
Also, you tested this system on a 600$ motherboard... 600$ and a prenium cooling solution.
This system is above the 2000$ threshold compared to an AMD one barely hitting the 1000$.