I wish I could use an external PSU for my desktop!
I'm like you; I want one box / lump, not two (mainly because I move the box around a lot). Consequently, I have a Mac mini. The price cheapest non-used Mac mini however has risen from $500 to $800 in about 3 years.
I'd prefer an external PSU.
I'm at work so I don't have time to do many more googles.
I myself have been building out a brickless skyreach s4 mini system inspired by this  (same author as this A300 unboxing)
If you integrate the power supply into the inside of the product, then you need to certify the product itself individually once for whatever standard, and potentially for each country you import into.
Hence Apple can do it partly because of a combination of scale and margin, where-as it's a lot harder for smaller manufacturers to do so in a cost effective or reasonable way.
Having said that, I agree with you, external PSUs drive me nuts. And I love products like the Mac Mini that have it integrated.
I bought the Intel version of the DeskMini for $130 a while back. It's about the same price as a normal case + mobo + psu, which it replaces. I put an i5 8400, an M.2 SSD, and some memory in it, and in 5 minutes you have a working system. No hassle, inexpensive, and small. Highly recommended unless you need a GPU.
If you get a case large enough to support a full-size GPU and SFX PSU (so around 7-8L in volume), you're golden. Smaller than that you start having to go with ITX GPUs or external PSUs.
I think every big motherboard maker has at least one ITX board for each new generation/chipset so I haven't had any trouble in the last 3-4 years since I switched over to ITX for my builds.
The only thing you really have to look out for is some of more recent extra tall + wide GPUs. But even some full-tower ATX cases can't fit these without vertical mounting.
The only compatibility issue that you need to worry about is cooler height and video card length, if you choose a very small case. But there are plenty of really nice ITX cases like the NCASE M1 that are reasonably small, with excellent compatibility.
If you want to get really exotic, there are cases that require DC to DC adapters for external power supplies, but if you stick with SFX internal PSUs you should be fine.
I've been waiting for the next run of the Skyreach S4 Mini to finish my sff build. It's gorgeous.
This helped me in the past for setting up my own NAS.
Asrock has an upcoming low-power fanless MB/barebone for embedded, http://linuxgizmos.com/worlds-first-amd-based-nuc-mini-pc-sh...
> ASRock Linux-ready “iBox-R1000” industrial PC and “NUC-R1000” mainboard provide the new AMD Ryzen Embedded R1000 SoC in a 4×4 NUC form-factor with up to 32GB DDR4, 2x GbE, 3x USB 3.1, triple 4K displays, and 2x M.2 slots.
For my current router I had to buy separate pcie cards to get proper throughput, the 1gbit cards on the intel motherboard could only sustain 300 mbit/s routing (600 mbit on the bus), but if only one card was used it could accept or send 1 gbit.
EDIT: smallformfactor.net link appears to work again.
The bet to prioritise server chips over consumer ones in fab priority may seem like a win from money side, but loss of clients will be much more impactful on them long term.
Just head over to eBay and you can find a bunch of used Asrocks with Socket A/462 for AMD Athlons from the mid-2000s.
It's way more power efficient to have an SFF do all the 24/7 IT-work and my gaming/main-PC only selectively powered up when I need it.
> The AMD Accelerated Processing Unit (APU), formerly known
> as Fusion, is the marketing term for a series of 64-bit
> microprocessors from Advanced Micro Devices (AMD),
> designed to act as a central processing unit (CPU) and
> graphics processing unit (GPU) on a single die.
My understanding is that it has always been a CPU and GPU combined, but is some marketing wangle as you suggest. I've owned one in the past and it was generally quite fast in general use, I believe it off-loaded simple GPU related tasks to the GPU part of the APU which ultimately reduced the graphics latency.
This seems to be quite similar to the Mali GPU which can be bundled on ARM packages to handle simple graphics . I believe this is how something like the Raspberry Pi can output 4k over HDMI despite not being particularly powerful.
Quite a lot of motherboards will do the (I think) "three beeps" which means it's detected a missing graphics card.
But if you leave it to go for a small while (20 seconds?), they'll continue on after the pause then boot/run without problem.
Useful for home servers.
On a similar note, I picked up a Gigabyte GB-BLCE-4105 a couple of months ago. Very small box with a quad core J4105 Celeron CPU. Tiny fan. Added 16 gigs of RAM and a 512GB NVME SSD.
Super nice home server, using a small amount of watts. Running a minimal Ubuntu install which uses kvm to host multiple isolated VM's for different applications (Unifi controller, domotica, smb, solar panel/energy control, etc.).
Gee, I'm just happy that they reaped the RS232 DB9 ;)
The article says "GTA V, 50FPS at 1080P" for Ryzen 5 2400G with integrated Vega 11 GPU.
BTW, on YouTube you can find some videos where that MB is piggybacked on huge VGAs: https://www.youtube.com/watch?v=pGM-Hf-CPuk
Supermicro is now shipping Epyc 3000 mainboards with ECC, e.g. 8 cores at 30W TDP with quad NIC and quad SATA.
Do you have a source for that? My current understanding is that Asrock only officially supports ECC RAM for Ryzen Pro APUs/CPUs, but non-Pro ones work fine with ECC enabled as well.
Particularly susceptible to memory errors are applications that keep data cached in RAM over longer periods of time, such as file / database servers.
One of the first Google hits: https://www.phoronix.com/forums/forum/hardware/processors-me...
AMD doesn't really have a good offering for small fast home servers that will sit in a closet.
Or 8-core Epyc embedded server motherboard with VGA BMC from Supermicro, https://www.servethehome.com/supermicro-m11sdv-8c-ln4f-revie...
That embedded server looks sweet but is overkill for most homes in terms of connectivity and price.
Also you can limit the power that the iGPU uses if you absolutely want, but as long as it is mostly idle I see no reason to do that.