I've had my eye on this thread out of interest to see how it would progress...
First, I thought I'd mention a little how I've experimented with 2 particular Thermalright coolers out of curiosity VS the NH-D15S with 2 fans - TR is REALLY popular over on Overclock.net.
Both were claimed by users - not the manufacturer - to be 'Noctua Killers'. Cpu is a 7820X, BTW.
A)Silver Arrow IB-E Extreme Rev. B
320w TDP. Equipped with 2x 2500rpm fans, this cooler requires that the chassis fans are also high rpm, so I replaced all chassis fans with IPPC 3000s.
I also tested it with regular Noctua fans, as well as the cooler fans.
B)True Spirit 140 Power
Whopping 360w TDP. Doesn't require any special chassis fans, but I wanted to cover all bases - tested with stock fan, tested with IB-E fan, swapped chassis fans in/out, etc.
No matter what I did, neither of these coolers actually beat the NH-D15S(dual fan).
I even opened a thread over on Overclock.net to try and get to the bottom of why.
The conclusion I came to from that thread - one or more of the following:
1)Poor surface mating. The flat base of the 2 TR coolers on the 7820X's slightly concave/convex(I forget which) IHS. The NH-D15S' base is also slight in one direction or the other.
2)TR lied about their coolers' specs. No way, right? Surely their engineers know better than lil ol' me...
3)Those folks touting about the performance of TR's coolers were full of it.
What does the above have to do your situation though? Well, I'd lean towards point 1 here.
-We already know the Antec NX300 doesn't have great airflow potential, but that shouldn't really choke the LGMRT, as it gets air from not just the front, but the top as well.
-The LGMRT is a big ol' brick of a cooler. I wouldn't expect a 3700X to come remotely close to being able to saturate it, but here you are, running around 85C with it. I would've expected results in the 70s...
-The center paste application isn't really ideal on Ryzen 3000; due to the multiple dies beneath the IHS, that method might not cover them. The thin spread and 2 dot(if you know where the dies are located) methods should be more effective.
-The motherboard doesn't appear to be cranking excessive Vcore through the cpu.
I feel like I forgot something, but it's late. Maybe I'll remember after a nap.