Depends, all the high-end AMD CPUs right now are 142W parts @ stock, so they all dissipate about the same amount of heat and run at about the same temp, albeit with variances. I think the 3950X actually runs lower temps than all the rest due to distributing power dissipation over a larger number of cores, but that is not always the case.
That being said, the temp you get reported from an AMD chip is a "hotspot" temp rather than a bulk die temp, so any of the 142W CPUs (3800X/XT, 3900X/XT, 3950X) can appear to be running hot when in reality it's due to power density. It doesn't take much of a heatsink to deal with the total heat flux; instead, you're encouraged to oversize your cooling to try to drive down delta T, which produces diminishing returns rapidly. Going from Wraith Prism to an NH-D15 to a 280mm AiO to custom water will only drop your observed temps in something like CBR20 by a few degrees per step. For example, I have an expensive custom water setup on a 3900X and get a temp of 60C in CBR20. I could put a Wraith Prism on there and it'd probably hit 75C. Performance is within 5% depending on other factors. It really doesn't matter what temp it's running so long as it's below 90C, since it's the same heat flux no matter what. At best, driving hotspot temps down buys you a few extra MHz of boost.
A 10900k is a completely different animal. For non-trivial durations, it can dump 250W, which makes it a lot "hotter" than anything AMD produces for AM4. Choose an undersized cooling solution - like an NH-D15 - and at best you are going to get some clockspeed losses as it throttles back clockspeeds. A 280mm AiO should be considered the minimum required to allow the CPU to reach its full PL2 value of 250W for a sustained workload of 56s - and that's only if it's following spec (some boards allow 250W indefinitely out-of-the-box). A CPU with a heat flux of 250W running a die temp of 70C is objectively "hotter" than a 142W CPU reporting a hotspot temp of 75C or higher.