Rpi5 to be released in October

https://www.makeuseof.com/raspberry-pi-5-launch-announced/

4 and 8gb of ram. Better cpu, ac wifi.
I’m only running some test instances on rpi, my prod oh setup is running on a lenovo tiny m920q.

Will not get an rpi5 for 60-100euros.

Word on the street is that it will be released at 23rd of October.

/S

5 Likes

This is good timing. I’m right at the limit of RAM supported on my VMs and need to move some stuff off to other machines. I was going to get an RPi 4 to add to my stack but this new 5 looks even better.

Eventually I’d like to decommission my Proxmox server and just use a stack of RPis to save physical space.

2 Likes

I just ordered mine! :slight_smile:

1 Like

I’m intrigued by the m.2 hat they’re working on, which could put all of the RPi storage debates to rest. Part of me thinks they should have just built it in, but making it an add-on almost certainly keeps the price lower.

I don’t need another RPi at the moment, but will likely pick one up within the next year.

And it’s not only m.2, but a PCIE slot, you could technically use it for many other things.

2 Likes

The eMMC is great, but it indeed contributes to costs. I am glad PI got a less problematic storage support. Making a flash disk image and burning it should not be that hard, you just need an adapter to plug it into usb. If they will make possible to boot from disk without all the crap (boot.txt) flashing will be super easy and less engaging for end users.

For these who wish to have both emmc and m2 option I can recommend other hardware I been playing around: URVE Board RPI - rpi3+ hardware replacement.

I saw a lot of reviews yesterday, I loved this one with the CEO talking about the changes https://www.youtube.com/watch?v=CmFEdtkyVQM .

Also the processor implements the ARMv8.2-A architecture which supports some features that allow to run IA models much faster, so llama and whisper models should be more usable now. I’m looking forward to seeing what people do.

As much as I love RPI:es I’m have gone the opposite route, going from a cluster of 4 x RPI 4 to Tinys
Lenovo M920Q, HP T620, HP T730 etc.
The HP T620 for instance, draws around 4W measuring in socket, price is about 30-50 euros on ebay including storage and memory.
A RPI 5 will cost about 100 euros new and then you have to add storage to that price.

I run most services I self host including openHAB in docker. I have also decommissioned my old promox setup. I can utilize the hardware a lot better by running docker. Next step for me is to set up Portainer to handle the containers, or possibly k8s.

3 Likes

not sure if there ever was an open-minded and respectful debate and when Pi OS officially supported booting from an SSD that debate was in my opinion needless.

My biggest limiting factor is RAM right now. My Proxmox machine (a Lenovo T150) is maxed out at 32 GB RAM and I’m right at the limit (I was over the limit earlier this week and had to turn down the amount of RAM used by ElasticSearch and switch from GitLab to Gitea).

Most of those thin clients and mini computers only come with 4 GB RAM (I just searched eBay and there is only one with 8GB in the US for sale that I can buy after a brief search). The processors are also pretty anemic, though that’s not a huge consideration for me, processor isn’t my limiting factor. And the storage is pretty low at 8 GB or 16 GB which would make deploying my Postgresql server and Plex somewhat challenging. They appear to only support USB 2 which further limits my ability to add external storage.

Unless I can find five with eight GB RAM or three with 16 GB RAM I think (or upgrade them, assuming that’s feasible, many are not upgradable), given my requirements, the RPi still comes out on top.

I already have the storage (if I can do the PICe to attach my HDDs so much the better), a stacking case, and all the other needed accessories (except the plug, see below).

The power supply issue is a concern though. They require a special plug which is not standard meaning, for now at least, you have to get the plug from them, and at least at Pi Hut they are sold out.

I have some room on my RPi 4 still so I might move Gitea and maybe Calibre over to it along side Vaultwarden, Heimdal, and Semaphore. That should free up a gig or two of RAM and make Zabbix stop complaining.

I run all my services in Docker except those that require access to Bluetooth or GPIO pins. It’s all managed through Ansible.

openHABian used to support running off of an external drive. It even had a migration tutorial. However, there were too many variables, RPi kept changing things, and it ended up being too complicated to maintain. So support for that was dropped and the SD card wear out problem was addressed in other ways.

There was no real debate. The people who volunteer to maintain it decided it wasn’t worth it any longer. It can still work and many still use that configuration. But if you do, you are on your own. The maintainers of openHABian can’t/won’t help you if/when you run into trouble.

Again, there’s not debate, there are just no volunteers to support it.

2 Likes

I meant in general, not specifically in the OH community. But yeah, it’s rarely open-minded or respectful of different needs and opinions. :wink:

Is the 4GB recommendation still in place? more power is good, but more RAM means it has to be a 64-bit OS… and openHABian still works best with 32-bit, I guess?

The recommendation from openHABian to use 32-bit comes from a time when RAM was the primary limiting factor. When all you have is 1 GB RAM and you run 64-bit, it requires more ram than 32-bit, you are making a bad problem worse.

If you have plenty of RAM to work with, there is no harm in using 64-bit. You have the RAM to spare. There is also some evidence that the first run of JS Scripting/Blockly rules is vastly improved (by 10x) when running 64-bit Java.

Most people also forget about the larger x86 processor architecture memory footprint.
An x86 processor needs 1,5-2 times the space for code and data that the same Java code on ARM will require. PLUS it is always 64 bit, once more doubling the amount of memory you need.
Check your Java process sizes if you don’t believe this is an important factor.
3-4 gig are pretty common values for a decent x86 OH installation while for 32bit ARM it’s rather 0,9-1,4 or so.

You might want to consider this post regarding 64-bit, especially if you are using JS script and want to reduce the 20 seconds delay of the first run of a script:

I do not follow this comment… is it saying, there is no openHABian anymore? Or it only offers SD card install?


I find your comments about PROXMOX interesting, as I thought about using it with a ‘tiny’ whatever machine. Meaning, I have now ditched the idea and keep OH on it’s own machine.

The later, only SD card install. Nobody wants to stop openHABian.

1 Like

I can’t imagine any benefit running it in an underpowered machine. It’s a Type 1 Hypervisor (sort of) and it’s best used on beefy machines, preferably a cluster of beefy machines, to more effeciently allocate and utilize resources. I mainly use it to separate some services onto separate VMs so I don’t lose everything on a reboot but still only need the one physical machine.

Snapshots and templates are nice too.

This topic was automatically closed 41 days after the last reply. New replies are no longer allowed.