For Ryzen 2, I see 70-75C as ideal for an under air OC. That's under the heaviest possible loads. Linpacks, small ffts, anything AVX heavy.
Tjmax may be 95C but i find stability actually drops stiffly past 80C anyway. Not to mention temps tend to rise more sharply past that. Im pushing up on that with my cooling setup now. The fans max at 70C to hold that 80, so if i push any higher I'm looking at a runaway scenario. Pretty much guaranteed crash. Fine for my uses but impossible to properly stress test. I always base the temperature limit on the heaviest possible burner tests. That way I know I have some margin on the real-world side.
Now... that's absolute max for me. For what I do I never see nearly that. I wouldn't be comfortable pushing that high all of the time. Right now I see peaks of 65 while gaming. But if you look at the plots, those are small peaks. As the load kicks up temps rise to 65 but under that continued load they steadily drop and plateau at 50-55 with little spikes between. At that point my fans are only running at half speed. If the gpu were to max at the same time i might need that headroom to keep it all under control. It'll pretty much flatten at 65C with a little more fan rpms for really GPU heavy stuff this way. Still leaves me some headroom on the off chance that I need it, which I do occasionally need to be able to tap into it and be able to do what I need to do without hitting my cooling ceiling.
To me that's a reasonable equilibrium for a chip that's rated at 95C. Plenty of longevity headroom there without completely outpacing cooling during regular use. I'd have to say it depends on what you're trying to do though. I would shoot for a max further down from tjmax if i was planning on going heavier more often. Just because that sustained heat builds up in pcbs and could affect nearby components. Prolonged loads at 80C might just start to shave a little life off of something if given time. Tjmax is where throttling/shutdown occurs to prevent immediate damage. Doesn't mean damage can't accumulate from just below that if sustained. The occasional sustained load at 80C isn't enough to concern me. It's knowing it's going there often that would. Common sense says that temperatures where stability takes a dive are generally not good. That's your CPU/VRM trying to tell you that you're working it too hard.
I'm running what you could call a mid-range air setup. H5 sized 140mm air cooler, maybe a little larger. Midsize ATX with 120mm top/rear exhaust and two 140mm front intakes. If not for that i wouldn't let my cooler ever move 80C worth of heat energy. I'd worry too much about VRMs cooking. Its a 6 phase so it naturally runs a bit cooler and with those two fans pulling air directly off the heatsink it tends to stay quite low. Overall case air temperature is probably a little lower than my hottest components, so it all works out without anything overheating, including drives and PSU - which still run cool as can be. If I can sustain a MAX peak of 80C and still keep those conditions, I'll take the performance I get by having that as my max allowance, knowing I'll rarely ever need it to bear that.
So I guess short answer is that there's no magic number that anyone one the internet can give. Its totally situational and relative to the specced max of the chip. You start from tjmax and work your way down to what is actually stable and what your cooling manages well enough to keep case air temps under control. For most people this is going to fall significantly below the max. Takes some testing to get it worked out, but that's going to be your "safe" range. Really just gotta take the time to think not just about the CPU itself, but the build as a whole and start seeing where your weak points are. You can pick the highest temp that your weakest point allows for. Going higher is getting out of common sense territory and into "calculated risk." So long as you know, you can always monitor and switch OC profiles when you know heavy sustained loads are coming.
Rest is comfort zone imo. Past being below specced max, its a matter of how much performance you wanna compromise for peace of mind... or how much heat you can gain before performance drops and/or you hit diminishing returns. Depends on which way you swing. For me its diminishing returns. I'll take the max performance at the expense of extra heat, so long as the gains are still significant enough to justify the thermal load. Some would prefer lowest temperatures possible.
<- safe to ignore. 64C is great and it means you can push it further in terms of temps/voltage.
when using Waterchiller you dont need Rads but i placed one inside the case knowing that doing so temps would drop.
You don't have problems with condensation on the rad?