One issue I have with that guide (although it is an excellent starting point for research and wrapping your brain around the basic concepts) is it gives some false impressions of baseline. For example, it estimates the RX360 to handle 675w. That would be capable of 2 gpu (310) and a base OC cpu (135w). Reality is... a big hell no. Zero headroom, no mention of accounting for environment factors (ambient temp, relative humidity), no accounting for static pressure vs cfm fan choice, and certainly doesn't address possible issues with someone just headed into world of OC. Could it do it? Probably. .. wirh perfectly matched fan setup running full blast and pump humming at max - damn jet engine lol. But a tenant of most watercoolers is balancing performance and noise. Adding an extra rad allows lower pump and fan speed thus quieter. Plus more headroom for oh crap moments. Rule of thumb i read, and seems to hold well, 1.5-2x 120mm per component cooled. Unless I missed these things somewhere, really should be added
The watercooling sticky does show the RX360 can dissipate that much heat because it can under ideal flow and fan conditions as proven by Skinnee and Martin. The biggest thing to learn from reading that guide is to determine
what you specifically should consider for your own watercooling loop...not following a cookie cutter design. The RX is actually a low FPI radiator, so you don't need high static pressure fans, meaning in turn, you don't need loud, expensive fans to get the best performance. Running 3x or 6x mediocre fans can net you fairly solid results if you find quality parts at a value. It's also a fairly unrestrictive radiator, meaning your flow rate should benefit well (most radiators qualify as the lowest restriction components in a loop, depending on make and model). You also have to account that total loop load TDP isn't going to be 100% thermal efficient when it comes to components being cooled. Power draw in watts and heat output in watts are directly related to one another, but that is never 100%. So, as a general rule of thumb, I typically suggest in finding total loop TDP (including overclocks) and drop that amount by 10-15% or so on average. This will give a bit more realistic TDP sum, but still is likely to be 'worst-case-scenario' as you are extremely unlikely to run 100% load on every component for extended periods of time (except when stress testing or benchmarking). If these are your primary intent, sure, over-radding is a recommended idea. Heck, over-radding is almost a religion. But, in most cases you have more radiator cooling headroom than you likely realize in most radiators and most custom watercooling loops.
Humidity isn't the issue, it's ambient air temperature. Humidity and dew point is going to impact what you as a human perceive the environment temp to be, but has a smaller impact on actual heat dissipation in this scenario.
Also, if you use the radiator thermal coefficient table to calculate radiator needs and potential, you'll find that your assessment of around 1.5x 120mm rad space per component is about right (depending on specific GPU in consideration, of course).