Posted by casi 4 days ago
The boards are of course bigger, which I guess will be a non-starter for many. They'll also require active cooling unless you downclock them considerably. They often also require more power, although many mainboards can be configured down two a few watts of power consumption, depending on what peripherals and speed you need.
Yeah, it's more work than just getting a Pi, but given the current prices, it makes sense for me.
If you use the raspberry pi for homeservers or general compute (which I did/do sometimes) then the value proposition is quite poor regardless of pre AI prices unless you have very strict power requirements (in which case it probably fits the above use case already). The comparison with NUCs or mini PCs or laptop boards feels like it's a wrong comparison.
The chips used in PIs are repurposed Broadcom SOCs that have inferior bus speeds compared to a general purpose processor from intel or amd, even few generations old. For any general purpose computing these chips will always be slower. its like comparing apples and oranges. I never really understood why people did that.
Yeah, I don't disagree, but I would argue the Raspberry was always an applorange, and that's exactly what made it popular (together with the cheap price in the beginning, which made it basically an impulse buy). It brought people into microcontroller/embedded systems territory without having to actually deal with all of the stuff that made this so daunting. As you said, it is basically a normal Linux PC, just that it runs Arm and has GPIO. People who are really into electronics/embedded systems will use a proper microcontroller nowadays anyway, these have gotten so powerful and easy to use that there's really no reason for a Pi anymore. And on the other side, people who want a little server, a pi-hole, tailscale or octoprint or whatever, can just use an old used laptop/miniPC or similar. Since the Pi has gotten so expensive, I see basically no reason to get one anymore.
By the way, I got a zero 2W a few months ago and it costed me 47€
You can make such a dongle yourself from ~any MCU dev board, but there are also productized "USB GPIO breakout" adapter boards, e.g. the Adafruit MCP2221A.
The "Pi as an SBC with GPIOs" case now only makes sense if you really, really require the Pi's compactness and/or monolithic-ness (as might be the case for e.g. ruggedized embedded systems like robots.)
For many (maybe even most) people they do the same thing. Some just want a HTPC, basic desktop, or small server, no GPIO needed. Raspberry Pis were definitely cheaper early on because you could actually buy boards at or near MSRP, but over time the price rose and mini PCs became cheaper.
or maybe the mini pcs started becoming available at all. I think these were a concept since core 2 duo and first gen i3 era. before then the chips were too much wattage for a mini pc. and the first generation mini pcs were also costly for enterprises to adopt in bulk. its only after 4th and 6th gen intel that these were adopted in large quantities in enterprises and once they dumped those in the used market they became comparable in pricing and somewhat in power to a rpi.
Used powerful PCs have always been available for dirt cheap on eBay even when the first pi was launched. just that mini pcs are a more recent thing.
As far as mini-pcs... The first Intel NUC was a Sandy bridge, i.e. Gen2, FWIW. And we did have the 'Mac Mini' as far back as 2005, buuuuut that wasn't cheap because Apple.
I think SSDs were another big part of mini-PCs though; it's easy to forget for granted the risks of having a hard drive in a poor thermal environment, as well as the general size constraints one imposes. [0]
[0] - Surrrre, you could use a 2.5 inch HDD instead, but 2.5 inch HDDs almost always were slower than a 3.5 inch HDD (excluding the VelociRaptor, but that would only be more demanding as far as thermal...)
At first it was a blessing that the Raspberry Pi gained popularity as a general purpose computer, not just for education and embedded applications. The additional demand and popularity brought more money and development attention to the ecosystem.
At some point, which I could never really put my finger on, this general purpose computer popularity grew so large that the embedded systems and educational aspects of Raspberry Pi seemed almost forgotten by the mainstream users. At a local in-person meetup group it was a running theme for people to brag about how many Raspberry Pis they owned for different purposes and to boast about using them for applications where they didn’t really fit. In person I’ve heard more about Raspberry Pi NAS systems (where people are always disappointed with the performance and the price of forcing a Pi into this role) or Raspberry Pi Kubernetes clusters (which mostly sat idle and could have been replaced by a single PC with more RAM).
This reached such a critical mass that it almost seemed like the mainstream discussion forgot the original purpose of the Pi and thought it was literally just another computer, but small. Most people who tried to use it for this purpose were left slightly frustrated by how it didn’t perform like they wanted a computer to perform.
So now every discussion about Raspberry Pi attracts comments about how nobody should buy a Pi when you can buy a NUC or laptop or some other used piece of hardware and it’s so much “better”. This is the right advice for a lot of people who wanted a powerful general computer but bought a Pi instead, but it’s also furthering the misunderstanding of what the Pi is for.
The real advice is that everyone should be shopping for the right tool for the job. There are a lot of cool applications for the Pi where a NUC or old laptop or phone doesn’t make sense because you actually need the compact size, the low-level I/O, or some other feature. However anyone who can put a NUC or laptop side by side with a Raspberry Pi and consider them both viable options for their project probably shouldn’t have been looking at a single board computer to begin with.
There is a parallel situation happening with the “just buy an ESP32” arguments. If an ESP32 solves your problem, you also probably shouldn’t have been in the market for a Raspberry Pi.
IMO, the problem with the Pi GPIO is that it's a collection of GPIO pins and low pin count buses for controlling peripherals, rather than a singular bus. The fact that it runs Linux without an RTOS monitor program or a sidecar of sorts adds to the problem.
Suddenly the Kodi and the GPIO crowd (Arduino, ESP8266, ESP32, etc.) could work with the same board.
A new era, like the IBM PC or the iPhone. Changed computing forever.
EDIT: like other commenters noted, it seems like it made the intel NUC a reality.
Fantastic for one-off projects.
Nowadays they require cooling fans and cost absurd amounts considering the performance so there seems little point.
Most of my one off projects these days are now on microcontrollers.
In a better world we'd all have one in our homes, storing all our personal data and documents. The cloud would only be used for storing encrypted backups.
Many small businesses, kindergartens etc. could also have all their digital needs served this way.
Yes they would sometimes break down, but with the right setup, the fix would be a matter of booting a replacement from the right thumbdrive.
I have a Nextcloud server running on it and the maintenance is negligible.
During the recent PHP-vulnerability I simply turned it off. The next day I booted it, and it automatically installed the new Nextcloud image with the vulnerability fixed.
Makes me wonder if I could have it check for CVEs and automatically shut down vulnerable containers until updated images are out.
May I ask, how do you think kindergartens would benefit from a home server?
[0] https://www.servethehome.com/introducing-project-tinyminimic...
[1] https://www.amazon.com/GeeekPi-Cabinet-Equipment-RackMate-Ra...
This let me standardize on amd64 and simplified a bunch of my deploy code.
Of course, there are a wealth of hardware options out there, so it varies. I've heard as low as 4 watts.
btop says CPU takes 1W when "idle" (seeding some torrents), no idea what the rest of the board and NVMe could take.
Can you explain this a bit? What is the application/use-case where the power adapters are a problem?
Old mobile phones or laptops are riskier because of batteries. Some laptops can work without batteries as well, but do old phones?
Intel MacBook Pro from 2014 works but throttles the CPU to 800 Hz making it barely usable.
I have a ThinkPad x230 (3rd gen i7) which makes an incredible Windows XP nostalgia device. It’s basically the newest hardware that exists with full XP drivers (other than USB3 which I don’t need) so it’s massively OP, which was the point of that project. Basically, “what if I had access to insanely futuristic hardware in 2002.”
Obviously I wouldn’t sacrifice that hardware but it might be cool to try that on yet another laptop. (Hey, I don’t have a laptop collecting problem!)
they did, with modifications
I feel like The Internet has this urban myth that says keeping a charged battery plugged in all the time will eventually cause it to explode and burn your house down, guaranteed.
I get that keeping a battery at 100% charge is not great for it in terms of longevity (ability to hold a charge). And it's obvious that heating the battery up by running the laptop at full tilt inside a backpack or something will turn it into a spicy pillow.
But I struggle to imagine the failure mode that reliably shows up when keeping the laptop powered on 24x7 that is never a problem for 8-16 hours of continuous use while plugged in (a very typical work day for my laptops).
I ended up repairing the Mac, replacing the battery, and installing Chrome OS on it. (chrome sucks though, someday I’ll get it back out and switch to real Linux)
I had a MacBook Pro where the laptop worked but the battery became swollen so much that it distorted the case. If I wasn't there to notice it, it would probably burst at some point and Lithium is very flammable.
It's probably rare and overblown, but it can happen.
If that had been in a closet, it would have been months or years before I realized.
for example some alienware and XPS models rely on the battery for amperage capacitance when their big onboard GPUs get busy; allows them to run cooler/less-wattage/smaller wall-warts but it also makes for a platform that'll crash when it hits a high GPU usage without a decent way to make up the amps.
'spose you could use a bigger power supply if you get bios/uefi to play nice. hard proposition sometimes with Dell products.
Batteries degrade much faster when fully cycled repeatedly, and even then just about any laptop battery can easily make it 6+ years with low-to-zero risk of failure.
Apple documented the behavior in a support note. I’m not sure if the Apple Silicon models do the throttling as well. I guess the situation doesn't come up as often since the batteries are no longer user-removable.
> Unless you're just doing it for the love of the game
Isn't that way a lot of people do the home-lab thing and related projects what would need the Pi these ideas are talking about replacing? It is why I do it [or used to be before working in tech for decades killed a lot of the enjoyment I had for playing with tech].
> or happen to have a lot of laptop boards sitting around you can sell
You are the first in this thread to bring up what any of this is worth on a larger scale, selling the results. Everyone else is talking about putting together one simple computer or jumped-up microcontroller for a task of their own, and maybe more later for other projects. Some of us find this sort of thing fun, and in the shit world we are all living in right now having a bit of fun and achieving something [sort of] practical out of it, can be priceless.
For my malware home lab, I decided to just reuse old phones and routers that nobody wanted anymore (quite literally got them from the trash bins). Now they're running an ARMv8 kernel as the host kernel and the rest is emulated or containerized, depending on the use case.
Works like a charm for most tasks I throw at it, and many old cheap Android phones have at least quad or octa cores in them, making them pretty awesome for podman/docker usage, too.
The only thing you have to look out for is to remove the battery, and that you use a separate power circuit with a step down converter to power it. But other than that, old smartphones have become amazing for selfhosting stuff now.
> The open question is supply. Finding two good used phones is easy; finding a few hundred matching ones, with healthy batteries and predictable camera behaviour, would take more work. I feel increasingly confident that it’s possible.
RaspberryPies and such are commodity hardware. You can buy one, two, ten, twenty, a hundred, and it just scales along. All of those devices will, barring perhaps a few dead-on-arrival ones covered under warranty, just work out of the box.
A used smartphone will have issues which can vary from device to device, so scaling that would take a very dedicated refurbishing setup where the quirks are ironed out (and the unsalvageable ones discarded).
I would also expect that you are starting out at the other end of the bathtub curve: with new devices you will have a few that fail soon after unboxing and using them a bit, and then they will run fine for years until you start hitting the other end of the curve and wear-related faults pop up. With old cheap phones you would start out at that side of the curve, so if you buy one hundred of those, a bunch will fail at various points soon after deploying them.
The cost of that unpredictable behaviour seems high to me. So much so that I wonder if it wouldn't be cheaper to buy a lot of unsold stock of a certain type of budget smartphone from two or three years back. Those devices could run fine as long as they aren't connected to the internet directly.
Not my experience at all. Pies are hard to find, get more expensive each week. Some are impossible to find (Pi zero).
The most fragile part of a phone is the screen. If you don't care about breaking it, then access to the battery becomes pretty easy, and therefore easy to replace.
Unfortunately thanks to the EU's new tariff regime it would cost over 30eur to ship it here, but luckily I can smuggle one back from my next trip to the UK...
That said for 4 and 5s - I've not had a problem here (Finland) in literally getting them off the shelf of a local retailer (Partco) for the going rate anywhere else.
A bird [redacted] bush.Most/all phones have one cell and therefore no balancing magic. Two connections do the job of connecting to the battery cell to the rest of the phone. These two are easily identified with a DMM.
If the phone works with only these two connections, then we're done; the phone will run from a bench power supply and there's nothing left to pursue.
If it does not work, then the functions of any additional connections be further explored while still never unwrapping the battery. Serial data? NTC thermistor? Something else? Fine; record that and emulate it. NTCs are cheap and so are MCU dev kits. (And if we don't have a decent-ish oscilloscope yet, then this is a good excuse to get one.)
If you don't have any real training in electronics, please refrain from giving advice like this. It can start fires, especially when lithium batteries are concerned.
Do whatever the fuck you want and suffer the consequences yourself for doing stupid things you don't understand, but don't encourage other people who also may lack understanding to follow you over a cliff.
If fucking around with lithium batteries doesn't make you pause a bit, then you're just ignorant.
Still though, could be a lot of screwing around needed to get this to work. Likely the extra wires go straight to a chip which expects to be able to monitor the voltage of the battery, and which might be alarmed by a power supply that acts unnaturally for a battery.
Their board currently on sale is this. Based on Fairphone 2. Targeting commercial or industrial more than hobbyist though. I've been trying to convince them to sell the old phones as is, as well. https://shop.citronics.eu/
Thanks to Marc Lainez and his token spend we have a great Nerves system for running the Fairphone 3 by just programming Elixir. And people are flipping new (old) phones over to be put to work every week in the Nerves Discord.
We have a good UI framework in Emerge. We have media handling with Membrane. And Nerves handles the OS integration for stuff like networking. I thoroughly enjoy making up my own mobile OS for my own purposes.
You lose the GPIO/I2C/SPI/UART as easy extension points compared to the Pi. But the stuff that is there is already well integrated.
It makes more sense with high availability though. Their current situation seems rough.
It's not the dark days of android where something like CyanogenMod takes a whole team of experienced system and embedded engineers to port and you don't have to hunt for information on XDA anymore.
Oh and cooling. That's something that the Pi has an edge on.
Early Android hackers might remember something called the Android Arduino Mega ADK. I suspect nowadays if you need real time control you can probably just have smartphone control the peripherals via the ESP32 using either USB C or Bluetooth. Developing over the wire protocols is trivial since LLMs can hammer out protobuf definitions like there's no tomorrow.
Bot farms works that way.
I also don't have a heat gun, so I had to use a Ziploc bag with warm water.
Still though, they make it really hard
I don't have experience with the ziploc method, but heat guns can melt solder so I'm wondering if temp was not high enough with the ziploc.
Could you say more about this? What kind of setups work well? Is this really straightforward or are there any subtleties?
I am by no means an expert at this, but starting of with a device on the Postmarket OS list and Claude, you can get a lot of these old devices working with mainline Linux and nearly all drivers. The newer devices are a bit harder, but if you go back nearly a decade, they're very easy to get working.
So far I have repurposed a Fairphone 3, Nexus 5, Galaxy S20, and Galaxy Tab 3.0 7". All of them (except the s20) with all drivers working compared to stock. Others have done a lot too: https://github.com/mlainez/awesome-nerves-systems
For DIY projects you can use host USB, or get one of the Adafruit FT232H type boards for GPIO pins.
So Nerves folks typically run Elixir on the Erlang VM. So you have the Phoenix web framework, Membrane media framework, Nx (nx_eigen) for basic inference, Emerge does well for embedded UI.
I've played around with clustering a bunch of these devices over Tailscale across my home. Opening cameras from one device and streaming to anothers display, willy-nilly. All experimental for me.
I've put Homex and Espex on them to have them integrate with Home Assistant. Feeding sensor data up or whatever I feel like.
Elixir already has a production embedded ecosystem through Nerves and web services have been the bread and butter for a long time. So you can do almost anything.
Because the pi has a maintained kernel and a userspace I don't have to fight.
Everything without upstream kernel support is inherently vulnerable.
This is why most phones are unfortunately not general purpose computers. They don't run the software you want, but only work on closed binary blobs hardcoded against specific kernel versions.
In the long run I also imagine this sort of thing being something someone does once per phone model and uploads it.
Perhaps not in the first world reality you live in, where you can buy a rpi 4 for a couple of bucks. But in third world countries they aren't readily available, and usually the imported prices exceed that of a decent PC.
You can find an old used phone for almost half the price that is still good enough for many daily processing tasks.
> You can just as well say why use an expensive PI when I have an old computer with a 1680ti and 12 gigs of ram.
That is a logical fallacy. A high end PC and an old used phone that you can buy in Ebay are in entirely different price ranges.
> Like yeah okay but what does that have to do with Pi? Entirely different device classes imo
Entirely different device classes, yeah, but both can execute the very same functions. The difference is one is much cheaper and you get 2x the processing power than the "more specialized" other. You can use both as local llm, serv ers, home automation... and even GPIO if you expose some wires from the phone.
2. Again, "cheaper" is a valid qualifier if you compare costs. Phones vrs pi is not what I am talking about, free vrs second hand (not free) is what I am talking about. Something cannot be cheaper than free so comparing costs is kind of weird.
Sorry for any offense this semantic may have caused.
If you want to compare the expensive factor, at least use some kind of cost to compare like the price of a used phone etc or the cost of a basic server to see how they differ.
Used phones are good for tinkering, especially for "display with sensors and battery attached" usecases, but there are still plenty of downsides. And unfortunately Linux on phones is just not ready for primetime, last I checked.
I do use termux daily to ssh into other boxes, that's incredibly useful
If you can access it from within the sandbox, you don't need to be limited to keyboards - the HID framework allows for arbitrary I/O.
https://www.parkytowers.me.uk/thin/wyse/3040/
Atom x5 Z-8350, 2GB of RAM, 8-16GB of flash, some USB, and there's a board on that page you can have done up to do a slapdash SD adapter inside the case for more flash.
I particuarly think the economics of this are interesting, if you take the one plus 11 you can get one for around £200-250. The CPU is very capable and the alternatives are a PI 5 or N100. You get 16b of ram, a screen, gps, bluetooth a camera... the list goes on.
These devices could really shine as edge computing devices and I believe they are overlooked.
I'm honestly curious to see if bypass charging is on, but it's been a while since I touched the device itself and it has been fine do far. YMMV
That's not universally true - a lot of android phones have a "bypass charging" mode where you run directly from the power supply, and indeed you can run them with the battery removed.
Not sure if those phones would work without a battery and throttle the CPU, or if they would crash.
There are exceptions, so not all phones will, but more will. I hope there'll be more time/effort put into hacking removable battery phones to accept hardwiring a PSU onto the battery terminals over the next 5 years.
1. Batteries swell over time. Replacing them is easy, but it's still harder than changing one inside a UPS hat. Most phones won't even start without a battery.
2. Software support will be awful. Maybe you'll get lucky and your device will be able to run postmarketOS.
3. There's no modularity, you're stuck with the full package. Don't feel comfortable having a camera or a microphone on a device with no security updates? Too bad, there's not much you can do about it.
Don't get me wrong, if you understand these risks, are fine with them, and Android devices cover your use cases — that's great, go for it! I just find SBCs with decent software support to be a better option for general projects.
Then you can use any 4 volt power source.
#2 and #3: You install Termux.
I once tried powering an android tablet with straight DC, turns it is EXTREMELY sensitive to slow power supplies, ie. it required a very beefy one. I had to add capacitors even to get it to work, it was really weird. But considering a battery can dish out some serious amps....
Also 4V sources are not common.
There is of course the issue that a phone uses momentarily high current when transmitting. So there needs to be beefy capacitor too.
I have done this many times, of course. Especially on Ebooks, for which you cannot find replacement superthin batteries. You use original connector and any small lithiums in parallel. Discarded e-cigarettes are good source for small lithiums.