Case: Streacom DB4
Mobo: ASRock AB350 mini-ITX
CPU: AMD Ryzen 5 1600
CPU Cooling: Stock DB4 heat pipes + Streacom LH6 Cooling Kit
RAM: 32GB Corsair Vengeance LPX DDR4
SSDs: 1TB Samsung 960 Evo NVMe + 1TB Samsung 860 Evo SATA
GPU: Asus Phoenix GeForce GTX 1050 Ti 4GB
GPU Cooling: Streacom DB4 GPU Cooling Kit
PSU: Streacom ZF240

No fans. No moving parts at all. 0dB

Two more images showing the CPU and GPU heat pipes...

The stock DB4 heat pipes, and optional LH6 heat pipes, installed:


The GPU Cooling Kit installed:


Full write-up over on WordPress...

Congrats on your build. What soak temps are you seeing with typical loads?

Thanks Steve. Haven't gotten around to any rigorous thermal testing yet. Waiting for Ubuntu 18.04 LTS to drop (in four days) before I do a clean install and migrate everything I need over from the old machine. Light use and an IR thermometer suggests good conduction to all of the spreaders and no obvious hotspots on the mobo.

My old machine had a Win8 partition (primarily for gaming) that I could easily switch over to and do ad-hoc hardware/thermal testing with the usual suite of tools. This one is going to be pure Ubuntu so I think it's time I move out of my comfort zone and learn how to properly test thermals in a Linux environment.

I've completed the first round of CPU thermal testing — a full write-up can be found here.

The final results (normalised to an ambient temperature of 20⁰C) are:

Thanks - great write up. I'm always interested in the soak time curves as it provides a good feel for how well the platform can cool and where some of the limitations may be. It looks like the limit of transfer with 6 heatpipes is 42C to the exterior of the case (same temp for 75% and 100% loading). Streacom claims the 6 heatpipe config can cool up to 110W and you've got a 65W device in there. Have you considered a CPU overvolt/overclock just to see the thermal performance? It's for science! Maybe hook up a power meter so you know the delta between stock and OC CPU.

Also, what does the thermal curve look like once you unload the CPU? Pretty steep?

I'm not sure that saying "the limit of transfer with 6 heatpipes is 42C" is the best way to describe heat transfer, but I see what you're getting at.

Yes, that was the plan. A 95W TDP 1600X is basically an overclocked 65W TDP 1600. I adopted a conservative/safe strategy by starting out at 65W and am going to push it up a bit and see what happens when I get closer to 95W.

To be honest, though, I don't really need the extra FLOPS for anything. The stock 1600 is as performant a CPU as I need. That said, there's no harm in giving it a go. Once I've got the results I can always restore it to stock if I want. No point wasting electricity if you don't need to.

Will do. I've already taken some stock power readings. Power consumption (for the whole computer, excluding monitor):

• Computer turned off: 12W
• CPU (stock Ryzen 5 1600, 6 cores, 12 threads, 3.2GHz) idling: ~35W
• CPU loaded to 25%: ~55W
• CPU loaded to 50%: ~65W
• CPU loaded to 75%: ~70W
• CPU loaded to 100%: ~75W

Here's what it looks like when the CPU unloads after sitting on 100% for about an hour:



The sudden change in angle intrigued me, so I did some other tests. Full write-up here.

Sneak peek: A side-by-side view of the first 60s after un/loading the CPU from/to 100% looks like this:

Yeah, you are seeing the capability of the CPU cooler header to quickly transfer heat up to a point and then the slower tail of the heat pipes transferring heat to the case. Pretty typical of passive designs..partly due to the design of the heatpipes/interface and partly due to the mass/design of the case. The quick fall portion is the best indicator of a well done interface/installation of the CPU heatspreader to CPU cooler.

Ah... All of my previous machines have been air- or water-cooled and their curves did not have pronounced kinks. Learn something new every day.

Well, that's reassuring! Being my first truly passive build I wasn't sure what to expect.

Is a ±1⁰C/s rate of change something that is worth looking at improving, or would you be happy with that in a machine of your own?

Given how much effort (and thermal paste) it takes to assemble a DB4+LH6 I'm not keen on re-doing it in the short term, but if AMD releases a compelling upgrade to the Ryzen 5 1600 by 2020 then I may as well investigate the option (assuming there is enough potential upside to make the effort worthwhile).

It's fine as-is.