Sick of watercooling...

[David Hess]

I had to get used to the Ryzen as well
There apparently really is no issue with the new AMD processors when they hit 90 degrees. I think they throttle soon after, at 91 or 92 degrees.

It is just a question on longevity.  If you only expect to get 5 years out of your CPU then 90C is fine.

My old 120 watt Phenom 2 940 never ran above 65C, but my 65 watt Ryzen 3100 and 3700X hit 75C.  With the stock heatsink, the 3100 thermal throttled at 99C.

I think the difference is that the large die size of the 940 yields a proportionally lower thermal resistance.

[Ranayna]

Yes, the die size is the very likely reason why the Ryzen gets hotter.
If i remember correctly, the 5800x has only one CPU die. That model has the reputation of being a "hothead", the Ryzen that is most difficult to cool.

Longevity... Sure, higher temps accelerate aging processes. We will have to see how relevant this really is in practice. The 5xxx are about two years old. I hear nothing yet about dieing CPUs in significant numbers.

[SiliconWizard]

Air bubbles should not form in a pump. The pump should be at the lowest point of the system after the reservoir. Any potential air bubbles would be caught in the reservoir, and the pump always has enough suction head.

Air bubbles do not necessarily form "in the pump". They may form anywhere in the loop at some point. If so, they will eventually get to the reservoir indeed, sucked in by the pump. Depending on the reservoir, pump design, turbulences, etc, the air bubbles may get close enough to the pump to make it stall. I again suggest looking at how most of these pumps are designed to get a better understanding. This is a very common issue.

Now of course not all reservoirs and pumps are made equal, and the best ones (pump+reservoir) can go for over $200. So uh, yeah. But sure better gear and proper installation can make things significantly better.

Air bubbles are a lot less likely to form (if at all) in AIOs, but as we saw, those can have other issues.

And otherwise - one simple trick to deal with a stalled pump is just to tap it a couple times. It usually works.

[MathWizard]

Interesting that you're running into failures. My custom loop has been running 24/7 for about 5 years now with no issue, maintenance about every 2 years.

Maybe look into how data centers to watercooling?

And I get annoyed taking out my gpu for cleaning

[SiliconWizard]

What are the failure statistics on AIOs anyway?

I don't have any, but i'd say stay away from those that do no have proper fittings for the tubing. Those that just use rubber tubbing shrunk around small fittings without anything to keep them tight in place - that's a recipe for failure after a couple years. I think Be Quiet AIOs are rather well made. Alphacool ones too. But they are not cheap. Been pretty disappointed by Corsair's stuff.

Do favor compression fittings. Unfortunately, those are more common in custom loops than in AIOs. I think Alphacool AIOs have compression fittings, IIRC? But they also are "refillable" AIOs. And again, not cheap.

Leaks at the CPU waterblock are less common from what I've seen. Not that they never happen either. Or from the radiator's tubing - that would likely come from corrosion, which would mean you didn't use a proper coolant or failed to replace it (if not an AIO) in due time. Obviously never use tap water.

But the stories of ruined graphics cards due to leaks are everywhere, and I did witness one myself.
That part could be mitigated by putting a sheet of metal or plastic over the card to catch any potential leak, but doing so usually hinders cooling for the card, so to be done carefully.

[wraper]

I may just end up building less critical stuff out of air / water coolers like dummy loads or such vs.

What's the point? You don't need any water cooling or heat pipes for such application. If you are into such total heat dissipation to even consider water cooling, you should use multiple pass transistors anyway. Which means that heat dissipation will be spread out. Things like water cooling or heat pipes only are needed when you have a very small area with a lot of heat generated.

[SiliconWizard]

I agree there. If you are the designer of your application and said application can be implemented spreading the heat sources around, passive cooling will be no problem and much easier to handle.

The real problem with those powerful CPUs is that a very high power is dissipated over a very tiny area. Indeed not what you would typically have to deal with for a "dummy load".

[SiliconWizard]

Just to add a little something - if you're interested in designing your own cooling systems for whatever application, then you could also consider what I suggested earlier: using heat pipes to conduct heat, and dissipate the heat further away from the source. This is what liquid cooling achieves "easily" (but with maintenance that comes with it), but you can also do this without any liquid.

This is how heat is dissipated in modern laptops, for instance. You can still use heatsinks and fans, or go fanless depending on the context. But the benefit is that the heatsink(s) and fan(s) can be put further away from the components to cool down, which would give you flexibility.

Actually, this is the main benefit of watercooling for computers IMO. Sure, the liquid has a good thermal capacity, but good heatpipes can be pretty close, without the hassle.