I bought a large heatsink size 37*37*24mm and modified with my dremel. And then connect the heatsink and the soc with a pice of copper and 3M™ Thermally Conductive Adhesive Transfer Tape 8810.
I've learned that both the ap6255 module and the ddr ram onboard generates lots of heat.
ap6255 gets overheat under heavy load and stop working until core temperature drops under alert temperature without a proper heatsink
I'd say if anything the evolution of WiFi should tell us that bigger != better and more power != better reception and longer support-time != better wifi.
I've experimented with that topic A LOT with many different devices and routers - expensive and cheap ones.
Almost all RTL8811au and RTL8812au dongles just work well (there are obviously some black sheeps that overheat, and some that have a random MAC-Adresses assigned, but they're in a minority and can be dealed with) - even if the drivers have hiccups and are all over the place. The biggest issue is a lack of official support (especially regarding upstream support) requiring all kind of user-patches - who would wonder. I found the driver-variant from aircrack-ng to be the most stable (I think that's the one Hardkernel also uses): https://github.com/aircrack-ng/rtl8812au. Realtek is at it again, telling us their own drivers to be a code-mess rewriting them ... supplying ... tadaaaa ... nothing as an replacement in the meanwhile ...
A year later ... we're still not there. Oh well! At least the RTL8812ae PCIe variant is already supported now. Hopefully it will get upstream support soon (tm).
Btw. If you're seeking small, inexpensive 5GHZ AC-WAPs (or Routers) on the other hand I can only recommend the Xiaomi Wifi 3G 2018 (with GBe, not 2017!) with Padavan Firmware ...
They're worth their money twofold over every other expensive Consumer-Routers I've owned before (Several Asus ACXXAU, TP-Link, Netgear Variants ...).
The thing is they don't even offer MU-MIMO, but that's fine - the bottlenecks for a lot of the consumer routers above will be their horrendus power-design, airflow and software.
They're overcramped with components that increase the heat and decrease performance, lifetime and reliability.
If you really need MU-MIMO then get something like an UniFi UAP-AC-PRO (usually for a flat with 100 feets that's simply a waste of money).
I'd say give it a few months / years and we might have upstream support (hopefully).
TLDR: RTL8811au and RTL8812au are the choice to go for SBCs if you need Wireless nowadays.
You'll still benefit from 2 Antennas without MU-MIMO because one is being used to send and one to recieve. A single antenna can just do one thing at once. 2x MU-MIMO only means 2 users can simultaniously send/recieve on 2 Antennas at once.
Powered via Micro USB from an USB port of the router next to it (somewhat recent Fritzbox). Storage was a 128 GB SD card, no peripherals, cpufreq settings limited the SoC to 1.1V (fixed 912 MHz) and so maximum consumption was predictable anyway (way below 3W). I could've allowed the SoC to clock up to 1.2GHz at 1.3V but performance with this specific use case (torrent server) was slightly better with fixed 912 MHz compared to letting the SoC switch between 240 MHz and 1200 MHz as usual.
I used one of my 20AWG rated Micro USB cables but honestly when it's known that the board won't exceed 3W anyway every Micro USB cable is sufficient.
By comparing https://github.com/friendlyarm/linux/blob/sunxi-4.14.y/arch/arm/boot/dts/sun8i-h3-nanopi-duo2.dts and https://github.com/friendlyarm/linux/blob/sunxi-4.14.y/arch/arm/boot/dts/sun8i-h2-plus-nanopi-duo.dts the only real difference is replacing the crappy XR819 Wi-Fi with RTL8189 (the other change being H2+ being replaced by H3)
NanoPi Hero also features RTL8189 Wi-Fi, is limited to Fast Ethernet and has an I2C accessible voltage regulator allowing the board to clock at up to 1368 MHz (at 1.4V VDD_CPUX): https://github.com/friendlyarm/linux/blob/sunxi-4.14.y/arch/arm/boot/dts/sun8i-h3-nanopi-hero.dts
Maybe @mindee is so kind to provide an early picture of the latter?