Hi folks,
Hoping for a steer for a small HA (but also Frigate & transcoding) homelab project
Budget:
- ~AUD$350 / US$230 flexible by say $AU100 / ~US$50 for a better resourced solution but I’m not going to go crazy.
- I have not looked into running local voice + coral card / Open Vino but understand a little to know I’ll want it for good object detection.
Existing tech:
- Synology DS1019+. 16gb RAM. (Planned upgrade) from 1gbe to 5gbe {~3.4ghz utilised).
- DS1019+ specs (page 3)
https://global.download.synology.com/download/Document/Hardware/DataSheet/DiskStation/19-year/DS1019+/enu/Synology_DS1019_Plus_Data_Sheet_enu.pdf
The NAS runs:
- Plex (limited on the fly cpu transcoding).
- A dozen related media management containers
Networking:
- Behold my exceptionally technical drawing:
- I can’t see a reason I couldn’t put the NAS and this project in the office and upgrade the switches, but Dunning-Kruger makes me cautious.
- I have more space in the office and space where the ethernet terminates is very limited.
- House is small 160m2 lot, two storey, HA devices on both floors.
- Routers (current switch) have 1gbe ports. Will upgrade switches.
HA devices:
- Have an as-yet unused SMLIGHT SLZB-06
- A few Zigbeem2qtt devices. One Ikea thread powerpoint to kickstart Thread.
Goals:
Primary goal:
- Home Assistant (HAOS) (likely to deploy about 20 matter / thread devices but slowly and methodically as I learn integrations).
- Frigate (3 cameras max, am replacing my existing cameras so can get cameras that work best with Frigate).
- Have enough overhead so resources will not be an issue but not wild amounts.
Secondary goal:
- Media transcoding (via Tdarr). It does not need to be fast - Plex can transcode live while Tdarr takes longer to ensure subsequent plays do not need transcoding. Will not transcode 4k.
Tertiary goal:
- Offload Plex transcoding (or just move Plex off the NAS - I have Plex Pass)
- Lower decibels
- Energy efficient (efficient as in the difference between NUC vs an old inefficient desktop rather than squeezing a few percent between NUC models)
Hardware shortlist (I am struggling to set the right path here so I went to NUCs as my primary option for their compact size.):
- Beelink EQ14 - Intel N150 - RAM 16GB - SSD 500GB - 2.5G LAN/Dual - AU$330
- N1 Mini PC - CPU AMD Ryzen 7 5700U - RAM 16GB - NVMe M.2 512GB - 2.5G LAN/Dual - AU$380
- Intel NUC (8i3BEH) - CPU Core i3-8109U - RAM 32GB - NVMe 250GB + HDD 500GB - AU$259
- Intel NUC (8i5BEH4) CPU i5-8259U RAM 16GB - NVMe 256GB - AU$327
- Intel NUC (8i5BEH4) CPU i5-8259U RAM 8GB (free slot) - NVMe 256GB AU$327
I’m open to:
-
Any hardware options within budget well beyond my shortlist and beyond the idea of a NUC:
-
Dropping TDARR if that blow costs out the window.
-
Dropping Frigate for the same reasons (and deploy Thingio firmware on Wyze cam v3s).
-
Buying any brand from anywhere.
What I think I know:
-
Intel Quicksync is a big win for transcoding on the CPU.
-
AMD and Tdarr may not play as nicely as Intel. Not sure if that’s HA, Frigate, or Tdarr as I have done a lot of research and just encountered that. Believe it relates to the benefits of QuickSync.
-
USB passthrough required for HA so the DS1019+ is out of contention.
-
HAOS enables access to add-ons so plan on running that in Proxmox alongside Frigate and Tdarr (but again I suspect this may not be the best - I’ve just read of troubleshooting issues with Docker).
-
CPU is slower at transcoding than GPU, but better quality v file size.
-
Intel 150 NUCs seem to be less effective than the Intel 100 line, but not sure if this is just a myth.
-
Stick with Zigbeem2qtt and add Thread later. The Smlight M1 series is a good choice.
-
Home Assistant will run on a potato, but running Frigate with object detection will demand more resources, as will Tdarr.
Criticise this, me, be subjective, be unwavering, ignore OP and post memes, and do it all with my deep appreciation in helping me sort this out!
Cheers!
Darren
