I’m currently running Home Assistant on a Raspi 3b+, which is reaching its limits of only 1GB RAM.
Now I’m looking for the most economical option (purchase price + power consumption) to upgrade - and ideally to have peace of mind for the next few years.
The idea was to get a second-hand thin client and install HAOS on it (either directly or via Proxmox). In addition to HA (currently about 70 ‘devices’ incl. Zigbee, no cameras; a few automations), it would maybe later run an SMB server or a DNS ad blocker at most - nothing really demanding.
What hardware would you recommend now (Autumn 2024)?
Budget would be under €100
Suggestions I’ve found online so far:
Dell Wyse 5070 - 4GB DDR4, 16GB EMMC - about €50
HP T640 - 8GB DDR4, 32GB SSD M.2 - about 90€ (probably already much more than I need → and correspondingly higher power consumption)
I can only relay my experience. Can’t recommend what is best as only tried one solution.
Shorthy after covid when RPi prices were silly I decided to upgrade my system by selling 2 rpi4’s and replacing them with second hand HP t630’s. Since then I have been runing proxmox with HAOS, Pihole, and openmediavault with 4 usb ssd drives. Its been faultless, I upgraded it to 16gb ram and 250gb ssd. I then bought another winch now runs Frigate with 9 cameras and a coral a=e key thingy. This also runs Immich with a library of 22k images.
So I can recommend HP t630’s and I guess a t640 will be more than usable.
Used thinclients have the best price/performance ratio in my view. I’m using a Praim Atomino. Replaced the internal sata connected flash with a 128gb ssd. Consumes a little more than the raspi, but only some watt (4-8) so no problem but the performance are superb.
The only drawback of thinclients that I found is the lack of WiFi embedded. Small usb adapter can fix this quite cheaply if needed.
Not using proxmox as mine has only 4GB Ram and I was already using HA core, so using docker instead on a Ubuntu server os.
I bought a second as a hot standby and use rear as a backup tool. Daily refresh of the standby system. I excluded cluster configuration as the hassle is not worth in my view.
