> "...: It sounds like the key feature will be 'more': a faster CPU and faster IO, rather than new features."
Raspberry Pi Holdings is a embedded systems manufacturer for pity's sake; we don't need more from them, we need less. [EDIT] A faster Raspberry Pi 6 is encroaching on the territory of the Intel N150 and its successors and mainstream Linux distributions and that is a battle they would lose in terms of price and performance.
Give us a Raspberry Pi Zero 3W with proper sleep states to reduce sleep power consumption, lower idle power while awake, and 1 GB of RAM even if it doubles the price.
^^^ when I tell people tangential to the field that the latest pi needs considerations of cooling solutions and a beefy power supply (no more just any old micro usb cable into any old usb port), they're astonished. It was a "microcontroller" you could program in Python with a friendly Linux environment and is now an expensive, power hungry, hot computer with a microcontroller hanging off of it
I agree that Raspberry Pi is not a good general purpose computer, but some of these criticisms are starting to feel like a pile-on with partially incorrect information.
> the latest pi needs considerations of cooling solutions
FYI you can run the Raspberry Pi 5 without a fan or even a heatsink. It will safely throttle itself if it gets too hot.
If you're trying to get maximum performance out of it all the time, you will want a heatsink and fan. If you want to run some Python scripts in a Linux environment or even if you're doing heavy work and waiting longer is not a problem, you don't need extra cooling.
> and a beefy power supply (no more just any old micro usb cable into any old usb port)
This hasn't been true in 10 years.
Powering something off of any old USB port means it would have to fit within the 5V 500mA basic specification, which the Raspberry Pi 3 exceeded long ago.
> It was a "microcontroller" you could program in Python
It was never a microcontroller by any definition of the word.
Raspberry Pi uses USB-C and USB-C is more generous with power. Almost everything except computers provides 15W. With USB-C, devices request want they and there is no reason has to work with minimum possible.
The problem with the Pi5 is that they use weird profile, 5A/5V, that requires special charger. Most 5V chargers are 100W and beefy. If they hadn't cheaped out on power circuits, they could have used normal 30W charger. They should come out with new version that fixes that.
That’s crazy, 5A/5V is not and has never been in range of the USB power delivery spec. It only goes to 5 amps at 20V for the 100W output, or higher voltages in the newer Extended Power Range topping out at 5A/48V.
Keep in mind at least 2A of that is to power the four USB ports at their rated 500mA. So if you don't have four USB devices loading the USB ports at their max, you should be fine with 3A.
> Powering something off of any old USB port means it would have to fit within the 5V 500mA basic specification, which the Raspberry Pi 3 exceeded long ago.
Theoretically, devices like the iPhone have lower power consumption than that and loads of performance features like recording 4k 120fps video.
Of course, an iPhone costs much more than a RPi, and has much better economies of scale, so they’re not truly comparable.
> > It was a "microcontroller" you could program in Python
> It was never a microcontroller by any definition of the word.
I think the poster means people treat the R-Pi like an MCU that runs Python. The Arduino was popular at the time the Pi came out but limited. Once the Pi landed, it quickly filled the gap and the Arduino's popularity diminished to the point where it's now a corporate Pi clone.
Before we had the ESP32 type products that are now popular, using a Pi was often one of the only easy ways to make simple internet connected devices at home. I remember making smart lights and home sensors, speakers, clocks and all sorts of things using the Pi Zero W. It was cheap, small and easy to get up and running.
I and many other people very much used it like a microcontroller.
> Powering something off of any old USB port means it would have to fit within the 5V 500mA basic specification, which the Raspberry Pi 3 exceeded long ago.
In practice most USB wall warts will happily provide 2 amps or even more with no overcurrent protection at all. The 500 mA limitation, at least outside of ports on computers (where you'll maybe get a prompt on Windows and macOS), is theoretical.
> FYI you can run the Raspberry Pi 5 without a fan or even a heatsink. It will safely throttle itself if it gets too hot.
What's the point of doing so though? If you're doing this, you're obviously using the wrong device. If all you need is to run some python scripts in a Linux environment, you should use a Pi 3 or Pi 0w2.
Even for extended workloads, a Pi 5 without a heatsink is still a hell of a lot faster than a Pi 4. And as sibling says, most users appreciate bursty speed while not doing prelonged compute (see also fanless laptops).
(Disclaimer, I work for raspberry pi ltd, not views of employer etc.)
Most light workloads are very bursty. When you type a command or click on something you want latency to be low. Having the overhead to get it done quickly at the full clockspeed is good if you are latency sensitive.
Throttling has become a bad word. Some feel compelled to avoid it at all costs, doing things like buying big coolers and running synthetic benchmarks to avoid it. Unless you're doing sustained workloads where you need all of the performance, allowing a little throttling is fine.
I've been doing some heavy SDR lifting with a couple of my Pi 5s, and my own experience is that the active cooler works extremely well, and more often than not the fan can be shut off and it will work well as a passive cooler.
If you're consistently doing bursty loads that throttle, the thermal cycles are going to do some damage to your chip and it'll fail sooner than if you ran it cooler.
Do heat cycles mean that the device only lasts half as long as it might if it were kept at some constant ideal temperature? A third as long?
If so, then...so what? The tinkerers using these things aren't broadly concerned about these things. We're all out here running our real-live desktop, laptop, and pocket computers with bursty loads, dynamic clocks, and dynamic heat anyway, and these things are generally doing just fine. We aren't driving space ships with this stuff; it'll be alright.
I insist that there is no merit to holding a lowly Raspberry Pi board to a higher standard than we hold everything else.
Furthermore: If perfection is necessary for some kind of application, then maybe starting with a <$100 hobby SBC isn't the best move. It might be time to look within for a better pathway.
> It was a "microcontroller" you could program in Python with a friendly Linux environment and is now an expensive, power hungry, hot computer with a microcontroller hanging off of it
The Pi project was never originally a microcontroller - it was always a full-blown SBC you could program any way you want with some GPIO pins attached. People literally used them as (slow) home computers.
The company didn't sell its first microcontroller until years later in 2021 with the Pico, by which point we already had Pi 4. I do though think its a real shame prices for the SBCs have risen as they have.
At one point, the raspberry pi was a decent option if you wanted something hobbyist friendly that could toggle GPIOs and connect to the Internet (and later Bluetooth).
I suspect Espressif has mostly taken over that market now
The pi, particularly the pi zero is still useful if you need something that can run normal software but not a full mini pc. One example I've seen is using a Pi zero as a "wireless usb" where you can plug it in to a machine that accepts files over usb, and can now drop files on to it over the network.
Maybe you could do this with a ESP32 but it's easy on linux where you can use all the normal tooling and filesystem drivers.
ESP32 can run FreeRTOS, and I think people nowadays have no idea how much stuff we could do in MS-DOS PCs even with all their limitations for the epoch.
ESP32 hardware is much better than they used to be.
Not everything needs to be under Linux monoculture, thankfully.
It's still true that people, out of convenience and familiarity, used Raspberry Pi for tasks where a microcontrollers would have been perfectly adequate
There was definitely usecase overlap due to the presence of the GPIO, but huge numbers of Pis ended up doing things a microcontroller can't - stuff like the PiHole and Retropie projects, and never used their GPIO pins at all.
Thinking of any of the early Pis as microcontrollers ignores a huge amount of the ways in which actual end users interacted with the thing, and even the way it was sold and marketed. Upton was trying to replace early hacker-friendly home computers like the BBC Micro/Apple II, for a new generation.
> but huge numbers of Pis ended up doing things a microcontroller can't
Mate you know full well it was multiple things... Marketed to education as an actual computer, to the maker demographic as a microcontroller, as a way to learn coding but also robotics etc
>Thinking of any of the early Pis as microcontrollers ignores blah blah blah
This just clarified something for me. I've always been annoyed when I see a Pi with nothing connected to its GPIO header; why not just use a cheap thin client? Or an old laptop, for that matter? But that's missing the point. Here's the point:
Pre-Beagleboard-and-Pi, if you wanted an programmable thing to work with GPIO, you used an Arduino or a BASIC Stamp, or just a plain old PIC. But they wouldn't run a real OS.
Pre-Beagleboard-and-Pi, if you wanted an embedded Linux box, you used a WRT54G or a Soekris or an old laptop. But getting GPIO out of them was a PITA. (And often involved lashing an Arduino to the side.)
The Beagleboard (released in 2008), could finally do both. It had gobs of I/O and first-class support for it under Linux. It was pretty affordable. Then the Raspberry Pi came out in 2012, with a similar amount of GPIO, but demolished the price point to where it made sense to use it in place of a microcontroller.
That's really the magic of the Pi. You can keep one cheap gizmo around, and use it to solve (a large fraction of) two classes of problems. It doesn't fully replace everything a PIC or a PC can do, but it replaces an awful, awful lot of them.
> I've always been annoyed when I see a Pi with nothing connected to its GPIO header; why not just use a cheap thin client?
There have also been times when Pi's were cheap enough and x86 idled so power-inefficiently that you'd save money over a reasonable time horizon if you couldn't run your old laptops at full throttle.
Absurdly extreme example, but at one point I decided to replace a couple (maybe 3) RPi's with a single old Dell rack server off Ebay plus replaced my router with one running pfsense. I knew it would be mostly idle, that thing had 2 Xeon processors to replace 3 cheap ARM processors.
Between the 2 rack servers, my power bill went up by enough to buy a new Pi or two every month. It was like $80/month extra in power bills.
A $2 Chinese ESP32 dev board is more stout in many ways than the first Pentium machine I had, which sure did feel fast-enough to run the whole world with at that time. :)
In this amazing world full of such inexpensive choices, it seems so bizarre to me to demonize any of them. They're all excellent in some way, and it's OK that there's more than one.
This isn't football, handball, American politics, or Highlander: There can be more than one. It's OK.
This is what I keep repeating in many threads about ESP32, the little thing is actually quite powerful, especially for those of us that experienced MS-DOS PCs.
On the other hand, the RP2350 actually is a microcontroller, and IMO a nice one for many purposes. PIO, high-quality datasheet, nice ecosystem, etc. And the Pi Zero 2(W) can do most things the Pi/Pi 2 could, with a smaller footprint and less power consumption. Variety is nice.
I wouldn't quite go that far. Linux runs on nommu systems. With some psram you should be able to get a version to run on at least the rp2350 using the riscv cores with relatively minimal fuss.
Personally I have never seen Raspberry Pi as anything but a small personal computer. I've maintained a home server which get upgraded as Pi's upgrade. It runs well and manages my media, files and now running hermes agent. I remember even the first time it was introduced to me was with look it runs linux and you can probably run VLC on it.
Do you have a reference for this? Looking around, I see it being beaten by other ARM SBCs, and even low end Intel devices.
Many years ago, I measured performance per watt of the original Raspberry Pi when they were still relatively new. The performance per watt lagged behind even a beefy Intel box since the original Raspi was so slow that it destroyed any gain it got from using so little power.
They've never been particularly low-power, in a performance-per-Watt sort of way compared to other offerings at whatever present day. In recent times, I've seen completely-believable reports that N150 boxes walk all over it.
They've never been particularly cheap, in a performance-per-dollar sort of way. Used machines from eBay, yard sales, and old broken laptops (that still compute!) have always been better. (They usually come out OK when new is compared to new, though, which is IMHO the only valid comparison.)
Those comparisons were never very favorable.
---
The parts where it shines are: Small; they come in two sizes, and those sizes are small and smaller. That was new, but it's been cloned all over the place.
Built-in GPIO that's meant for people to actually-use and tinker with. That was new, too, but it's also been cloned. (Also: These days, anyone can plug a cheap Pi Pico into any PC with USB and get a fairly intense amount of GPIO to goof with.)
Standardization, and the appliance-like behavior this can enable. Lots of folks, including kids of all ages, just download pre-built images and swap SD cards like they would cartridges in a Nintendo. That's not for me, but it's pretty neat.
Community. They've still got a lot of momentum by being first at these roles. That's good. It helps newbs (who at this point may have never had anything resembling a "real computer" to play with ever before in their lives) to get started.
Low-power enough: It's not ideal when chasing tiny Watts for battery or small solar power, but you probably won't notice it on an electric bill (and despite the bizarro-world cooling rigs people put together, a passive heatsink really is good enough to keep it running in-spec).
raspberry pi has terrible power management as well. turning off a core was impossible on previous generations for example. a few years ago rpi was the worst of all sbcs we measured for battery powered usages. this was for an actual embedded product so it wasn’t going to be used either way (ask broadcom for some chips HAHA)
but there is very little reason to use a rpi over other sbcs if you have a remote idea what you’re doing beyond hobby use
Performance per watt is nice but I’d be more inclined to talk about “problems being solved per dollar.”
If you don’t specifically have a project where you need the GPIO pins built in, I struggle to understand the use case proposition of a raspberry pi compared to a typical x86 mini PC or even just grabbing a think client desktop like a ThinkCentre.
Almost everything that is unique to a Pi compared to an x86 mini PC seems like it makes more sense with an ESP device.
When the Raspberry Pi was $35 and it ran a desktop OS and the cheapest alternative that did that was 5x the price that use case made sense.
Maybe a tick-tock release cycle (one with new features and some speed, the next with the ~same features and more speed) is where they're headed, and maybe that makes sense. They wouldn't be the first.
I'd love to see even-lower-RAM versions, though. Most of what I use Raspberry Pis for at home for is not RAM-hungry at all.
My Pi4 network router has 2GB because that was the smallest/cheapest version at release when I got it, but the system itself consistently only uses about 64MB of RAM. It'd do perfectly well and have a ton of breathing room with just 128MB of RAM (which will never happen, but if it did happen...).
I suspect the Pi4 that I use as a set-top box with Kodi would be fine with 512MB.
I've used Zero Ws for all kinds of things over the years and never felt RAM-starved with their little 512MB of RAM.
So I'm learning towards 512MB.
But sure: 1GB options would also be fine even if it does double the price. Our comments serve to demonstrate that there's room in the marketplace for different SKUs with different memory capacities. :)
Ideally each RPi generation should keep the same price (or lower now that it's gotten so high) but with better performance. If they can't do that they just shouldn't create a new generation.
> Raspberry Pi Zero 3W with proper sleep states to reduce sleep power consumption, lower idle power while awake, and 1 GB of RAM even if it doubles the price.
Yes, that would be kind of a dream device, perhaps also if it could suspend the os when asleep so it doesn't have to boot every time but I guess that might the standard way of doing it.
I'd be interested in seeing them partner with some of the other interesting and open-ish players, like Pine64 or Radxa, to push for a more standardised, unified SBC landscape.
But that's not as good for PR as "bigger, faster, better" even if that comes with the problems you mentioned.
I wonder how they are positioned now in the market.
During covid I wanted a small low power always on server. I thought about Raspi, but at the time it was expensive and I went with an intel nuc, for a similar price.
Now if I wanted to do hobby electronics, I heard I should look into esp32 or stm32..
I use a Pi 4B as a 24/7 home server in a country where domestic electricity is expensive (and worsening).
Each Pi release is more powerful, but uses more energy. I found the Pi 4B to be the sweet spot for me, because it is the earliest model to support USB booting, gigabit Ethernet, and offer > 1GB RAM.
Perhaps a used one would fit your purposes and budget?
I currently use it to run PiHole, serve media via SMB, host Postgres & Redis, and run some custom written Dockerized apps. Home Assistant to possibly follow, too. The current load seems reasonable in htop, but I haven’t looked into burst scenarios.
The things that keeps stopping me from using one for my server is the IO. the PCIe lane on the Pi 5 was a great addition, but it isn't quite sufficient.
Raspberry Pi isn't in direct competition with N150's.
Their niche is the industrial/embedded space. For that market, power consumption doesn't matter. What matters is that each model is guaranteed to be available till a specific date.
> Raspberry Pi isn't in direct competition with N150's.
They may not have intended to be in direct competition, but in the current crisis conditions they are priced about the same as equivalent RAM/storage N150s, have even worse supply issues than the N150s, and have worse performance/watt than the N150s.
Its mighty hard to recommend them for new projects at the moment (nor any of the Pi clones, which are also rocketing in cost and dwindling in availability)
Among the people making things like a DIY NAS, who want fast USB, lots of cores and RAM, small-ish, not-too-bad power consumption, running Linux; and not caring much about GPIOs or passive cooling, it’s in competition with the N150
Among people who want GPIOs and network connectivity, a low price and an open, microcontroller like experience, not caring much about USB speeds and lots of cores and running Linux and suchlike, it’s in competition with the esp8266 & esp32. And the previous generations of RPi.
The Pi SBCs are squeezed in between the SFFs and the microcontrollers for many use cases. It's only a good fit if there is a need for the GPIOs/Linux/Display. For anything else, there are better products out there.
Is there some serious astroturfing going on with the N100/N150, or am I just jaded?
I have a bunch of old intel atom boards laying around. The Intel Compute Stick (TM) burnt out its flash root drive in a few months. The C2000 board I had burnt out the clock pin to drive the bios. I have a Clover Trail with a PowerVR GPU (I thought I was getting an intel GPU because it was branded Intel Graphics or similar, but nope!) that lost Windows support very quickly after launch, and has no GPU drivers for any other OS.
Instead of being fooled 4 times in a row, I looked into using an N150 for a NAS, but this time I held off a bit until after launch so I could research it first.
Lo-and-behold, they all have crazy PCIe / memory subsystem data corruption issues. I guess there are some chicken bits for the OS developers to set if the kernel can stay up long enough after boot without a panic.
Why would anyone buy this for a NAS / embedded use case?
All CPUs from all vendors have tons of bugs like this, which are mitigated in the operating system kernels, e.g. in the Linux kernel.
I am pretty certain that the Linux kernel must also contain specific code for various quirks of all Arm CPUs that have been used in the various Raspberry Pi models.
Intel had indeed several bugs that were more ugly than usual in their recent CPU models, like also MONITOR not working correctly in Lunar Lake, but even so, Intel still has better documentation for their CPU bugs than most vendors of Arm-based CPUs.
In any case, the bug that you linked was solved in the kernel years ago and it affects a privileged instruction that cannot be used in user programs. It does not have any direct relationship with memory and PCIe corruption. Memory corruption can occur inside the operating kernel only in certain circumstances, when the kernel changes the mapping of global memory pages and then writes the new pages, but the writes go to the old pages. However, this could happen only until 3 years ago, before the bug was known.
I've been running an N100 for 3 years with a 5 bay external enclosure over USB 3.2 Gen 2 and ZFS, and have not had any issues. It is pretty phenomenal, pulls about the same power, and costs around the same as an RPi 5 but provides substantially more compute and throughput.
I think exactly the opposite: we have no shortage of embedded crap we can buy; what is useful is dismembering intel. It would be better if the pi were risc v but this will do for now.
A very small point, but pulling from a feather form factor BOM to compare.
$0.12 for microUSB female connector (rated 1A)
$0.26 for a USB-C female (rated 3A). Needs 2 x resistors (< $0.01), 20% larger board area
I think the power capabilities are the biggest item. If you want to pull higher current from a laptop for development or supply from a wall, you have to switch to USB-C.
I don't think either of these prices are that aggressive - pretty sure the cost comes down at volume.
I wonder if it would be worthwhile for them to produce both. Well, it will be hard to compare because the design cost doesn’t show up in the BOM, haha.
But it seems like it would be useful nowadays, since some laptop have mostly USB-C connectors, and USB-C to USB-C is pretty common. I’ve never seen a C to Micro. Do they even exist?
I have an unfair bias because I design PCBs as a significant part of my job, and switching out to USB on this board appears to be a non-issue.
I have a Pico in front of me, and there's plenty of room there for a USB-C footprint and the two 5.1k resistors. Given that, I cannot reasonably agree that the "design" stage is significant.
In other words, it's a change that I would make to my own board in 2-5 minutes because the stakes are low. My ballpark guess is that such a change at RPi would have to go through a proposal stage, a PCB change review, and then there would be dozens of places to update documentation.
Since backwards compatibility is non-optional, this would result in a separate SKU, which means that the whole distribution chain needs to be updated with a new product.
So, I acknowledge that when you're working at their scale any change like this is the definition of non-trivial. What I don't agree with is the conclusion that it's not still clearly the right thing to do.
Comparing an RP2350 to the ESP32 family (which is broad) is very much apples to oranges; they each have feature sets which make them ideal for completely different use cases.
I cannot fathom why USB-C would make any difference. It's not like USB-C is intrinsically better for this use case? If you're doing hardware development, your desk is likely full of in-development crap anyway and a micro USB cable more or less won't make any difference whatsoever, nor is the Pico likely to the only thing needing a micro USB cable.
Edit: one thing I can think of where micro USB connectors are better: if you broke off the connector, it's much easier to solder it back on.
Micro USB is relatively brittle and can only be inserted in one direction. It means you have to make sure you have the old cable for this one thing. There's no USB-PD and you can't do USB host mode.
USB-C is the opposite of all those cons: much more durable, not directionally opinionated, you can use one cable for everything, you can do USB host and USB-PD.
It's also clearly the future everything is standardizing on. There's value in embracing that, since it is a Reference Board.
For what it's worth there are third-party rp2350 boards with USB-C connectors if that's important to you. Heck, WaveShare has one with two USB-C connectors: https://www.waveshare.com/rp2350-usb-c.htm
I do think that you're missing my point, which is that we're significantly past the point in this wretched timeline where they should offer a USB-C version of the reference board for this MCU.
Oh, no, I'm with you. The fewer connector types we have to use, the better. My comment was meant as an FYI for you or anyone else who wasn't aware and not as a dismissal of your valid criticism.
They probably just bought a shit ton of micro USB connectors back in the day and want to use them up, or something silly like that. It would be funny if the EU forced them to switch to USB-C.
You can at least buy USB-C boards from other vendors since they sell the rp2040/rp2350 separately. If you want wifi it gets a little more complicated unfortunately.
i understand that if they implement a port, it is usually well done (besides the unfortunate 27.1w 5a bs for the rpi5). so unlike cheapo electronics that do come with barely working type c.
however, while in one hand we are happy to (albeit temporarily) raise prices based on ram situation, in terms of design, there is simply not enough money for the port. especially when now they are adding ecosystem items that do cost money to develop and maintain.
this explanation made sense pre-ipo but no longer imho.
The good reason is that there are plenty of third party boards that already offer what you want. There’s very little to be gained by an ‘official’ one. The next one probably should have C, if just because it is the Euro standard, but no urgent need to backport.
I don't need a Pico or any other 3rd party board. I drop SC-1511/12s packages on my PCBs as needed.
What I am saying is that the reference board for an RP2350 should have a USB-C port in 2026. It's not aesthetics or even convenience (and it's definitely not price) so much as establishing best practices for how a part should be used.
A big part of that is to acknowledge the context in which a part exists, and in this case, it's both a fact and a very good thing that the world has embraced USB-C. It's even being regulated in many cases.
I'm not saying that you can't smoke, just that maybe you shouldn't do it when you're volunteering as a Big Brother.
While we're at it, their reference board also doesn't have a reset button, it just has one for boot. It's perhaps one of the most inconvenient official dev boards I've ever used in modern times.
It's not $0.50 extra. It's $0.16 extra for a USB-C port assuming you bought the USB-C port on mouser at 10k quantities and threw the micro USB port away.
I just don't get it. Anyone who wants to save a few pennies just buys the chip directly. Their Pico board is primarily for prototyping and one off products, where quality of life is everything and 16 cents is nothing. The adapter cable probably costs more than the amount they saved. That's a dick move.
Unpopular opinion but I actually enjoy using USB Micro B more than USB-C. That's because USB-C is much more complicated and there're non-standard-compliant cables floating in the market. USB-C also has some fancy mode like voltage selection. If it got screwed up somehow leading it to supply the wrong voltage, it could fry your Pico board.
> When asked about the Pi Zero 2W, Eben said the substrate supply is constrained—basically, so many AI chips are being made that even older chips using older process nodes have to fight for the actual silicon wafers to use to make the chips.
I keep hearing voices invalidate each other, is the bottleneck the raw silicon substrate, or fab capacity?
What purity levels are required for say Pi Zero 2W?
The volume of monocrystalline silicon used in solar panels is orders of magnitude greater than the volume used in IC's / RAM production.
Is the actual bottleneck 11N + grade silicon wafers while 6N to 9N grade used in solar panels remains unaffected?
The flagship Pi boards are also hitting the thermal design ceiling of what makes sense for this size or platform. Pi 5 is IMO already on that edge because it pretty much needs active cooling and a dedicated power supply. Going past this point means it will compete with much more powerful platforms and likely loose due to it’s architectural limitations.
The 8GB Pi 5, at $170 [1], is encroaching on Jetson Orin Nano Super's $240 price point [2]. But the Jetson has a faster CPU (newer a78ae cores rather than a76) and, obviously, a whole-ass GPU.
Nvidia's software platform for the whole Jetson series was, at least in my experience, absolutely awful on the Jetson Nano and Orin boards I worked on. Has that improved at all? I did not appreciate that the only option they provided was a full desktop version of ancient Ubuntu... and even flashing the OS image was a bizarre process.
Edit: looks like they at least have a better headless option now.
Nowadays upstream Linux with UEFI mostly works, with their out of tree drivers. I’ve managed to make it work in NixOS with the stock kernel. Look at the open embedded L4T project, they have some recipes for building that. No need to use nvidia’s kernel anymore!
Also, supposedly on the second half of 2026 they were going to be moving even more stuff out of their Jetson-specific drivers as they already do for their slightly newer chips (so you could use the standard drivers, and standard CUDA builds). Let’s see how that turns out.
They’ve got external kernel modules that they use for some hardware that is not supported upstream, namely the GPU, but also some things like Ethernet I think. Everything else is upstream though, so the situation is similar to a regular NVIDIA GPU.
Who would’ve thought that NVIDIA started upstreaming stuff once they realized how much money Linux is making them?
Yeah, one of my bigger complaints especially on the Nano was the GPU only had really limited model support (iirc, mostly tflite but maybe I'm misremembering) and it sounds like the newer ones are more normal. That and what seems from the docs to be better headless support would be major improvements. Going further to mainline distro support would make them interesting to me again.
I was always disappointed by the Nano as it was a pretty capable device, but it seemed like not many people picked it up as a platform for cool things which I always attributed to the software.
It used to be that raspberry pi was a cheap pc. Well it's not longer cheap.
And at their price point, you could just get a mini PC and have better performance, or if you want to use it as a microcontroller, you can just use an arduino, esp32, or an actual microcontroller for a fraction of the price and power consumption.
The form factor. It's tiny compared with mini PC. Unlike other SBC, the software support's good that it just work as advertised (GPIO, graphic hardware acceleration, etc.). If you're using it for work, it also has stable supply chain / commitment that you can ensure to be able to obtain a unit in at least the next few years.
But you're right. It's pretty overpriced for its performance. If I'm after the GPIO I'd just use a microcontroller. If I'm setting up a headless server I'd just use pretty much any other SBCs or mini PC. I'm personally not a big fan of Raspberry Pi.
The cheap Raspberry Pi's still available as Raspberry Pi Zero tho, if that's what you're after.
Hard to care anymore. They made it abundantly clear they dont care about their original audience of makers and education anymore now that the industrial market are so reliant on them. Not to mention the whole crapping all over their customers on social media and then accusing them of some sort of coordinated 'attack' just made it clear they're clueless about their audience.
The price increases were the end of the Pi being a viable option for most application in my eyes.
The PI5 was a real downgrade for me with its lack of proper hardware h264 decoding as in the earlier versions (playing back h264 is my primary use case.) I will buy a PI6, if that comes back, else I stay with my reliable, and passively cooled, PI3.
I have bought an rpi at every generation. And I still have yet to find an actual use for them.
Everything they do from a compute perspective is just better with a mini pc or old laptop with a mobile spec chip.
Everything they do from a programmability perspective is just better with a microcontroller specific to the task.
I just don't see the actual market position for these things. They were supposed to be a cheap board, but you can't actually buy them cheaply because the vendors upcharge so much.
Rapid prototyping: I created a PoC to take a webcam snapshot every half hour and upload it to a server in an afternoon. Freelance project. Could it have been done with a microcontroller? Yes, but not in 4 hours.
Local digital displays in a gym: we built a system with a number of overhead 60" TVs, each with a pi on a VESA mount to show scheduling and workout information in a gym for a client.
HVAC controller: bid on a project where the customer's original concept was a Pi managing a rooftop HVAC system for large buildings. They outgrew the pi and wanted a new solution.
Data aggregator: collecting sensor information via BLE (bluetooth) and uploading to an internet server over Ethernet.
Remember that "cheap" to a consumer doesn't have the same definition for a business. To most profitable businesses, a $100 computer that fits on the back of a TV and consumes a fraction of its power is cheap. In fact, one of the reasons that Pi's were so hard to find for a while is that the Raspberry Pi foundation was prioritizing industrial/commercial customers over hobbyists.
How do you find these kind of jobs? That’s something I think I would enjoy to do but I’m not sure where I would start looking for customers. Is it world of mouth?
For a hobbyist, they're quite nice for "I want to SSH into a thing, write my tools on Linux, and still have access to SPI / I2C / GPIO, USB, and whatever hats can plug into that." The Hat form factor, while not technically great and frequently overpriced, is also nice for distribution; both inside a team commercially and on the web as a hobbyist, it's a lot easier to share software and say "hey, buy this pi, this hat, and run this" than "fab this PCB / solder this bundle of nonsense to an ESP."
For a company, they're also nice for "I want to make an IoT device that's heavier weight than an ESP32 and/or I only want to hire Linux Application People and not Firmware People; what's the cheapest Linux module I can get that's widely supported, backed by a real company, and has regulatory approvals" - Pi Zero W. My understanding is this exact pattern is why it's harder to get them as a hobbyist.
I use them widely in automotive reverse engineering; ESP32 can and does work just as well or better for an end product, but for experiments it's really nice to have a self-contained appliance to SSH into and use SocketCAN on rather than some bespoke firmware project to manage and iterate on.
Given the price and availability issues I suspect the market is "correcting" a bit and companies are hiring Firmware People and switching to true MCUs in places they'd previously have avoided doing so, but it was definitely a thing for a long time.
I think I found a use case, but would love to be proven wrong with some faster / cheaper approach:
I've tired of buying S1 compatible sonos speakers on ebay, so I'd like to build a speaker enclosure with a WiFi device that has a high-quality DAC and the ability to use pipewire or similar to do real-time DSP and multi-room audio sync.
A RPI + third party hat should work well for this, or so I am told.
Now I have a wifi radio running in homeassistant.
Running my own (music only) radios with azuracast.
Loud enough for high quality 30+ year old 10kg 100/140 Watt speakers.
Next step is making a remote for it, cheap ikea one with two buttons or a tap dial with more.
This sounds like a great approach to me. I've thought some kind of pi zero 2w based smart speaker system would be an awesome project for a while. I am having a difficult time imagining a cheaper option, especially if you want something like the hat ecosystem.
> I have bought an rpi at every generation. And I still have yet to find an actual use for them.
It's amazing how well these fit into the category of products that people feel compelled to buy, play around with, and then forget about.
Flipper Zero is another product that landed in the same space.
What's sad is that Raspberry Pi does have a lot of legitimate use cases and people who want to use them, but the supply has always been swamped by all of the demand.
I think at this point the brand reputation and software quality are a big selling point.
If you're trying to build a couple of units of some embedded thing where you need to toggle some GPIOs or serial devices in response to requests over the network, but don't have the expertise or resources to do it with a microcontroller, a Pi is a great option - you know you'll have software support, and you know that the vendor will be making the exact thing you bought for 5-10y.
For hobbyist stuff at home, I agree, though. A mini PC is probably better for homelab stuff, and an RP2350 or ESP32 is probably better for anything embedded or battery powered that you want to do.
Development speed is also one of the forgotten axes. I mentioned upthread a system I built that was done in less than a day. There's no microcontroller solution I know of that would have let me deliver it that fast.
I've been playing with Circuit Python (variant of Micro Python, etc.), and it seems quite great actually. For small projects it's far simpler to use than C/C++, although if you're using some peripheral that isn't just toggling pins or simple serial commands you need a library (module). But then in that case you'd probably want a library in C as well. Definitely recommend it :) (it works on ESP32 boards as well as similar recent microcontrollers)
* Replacement controller for my UFO Catcher - It has WiFi, easy to update, and I can operate the machine remotely with it. It's bolted to the back of small touchscreen that lets me change the machine settings as well.
* Remote printer access - I can monitor from the USB cameras and gather statistics about the prints.(I suspect a lot of 3D printing enthusiasts use them for this purpose.)
Having a small low power computer has been useful for me in those instances.
They have great software support and Google-ability, which also means LLMs understand how to work them. I use them in scenarios where I want both Linux and GPIO. Specifically, right now I have 4 Pi 5s running kiosk displays with RGB LEDs around the outside of each. The displays show a web page, requiring something powerful enough for modern Firefox, and the LEDs are synchronized to the state of the web pages. You could do this with a mini PC and a microcontroller, but it’s just way easier, and cheaper, to use a Pi.
My opinion is that Raspberry Pi has to release an NPU, and start/revolutionise the open source NPU communities and tech. Raspberry Pi's has to find itself useful in vision AI applications (extremely common in industries these days). Without it I think Arduino with its new Qualcomm boards will kill it.
I run the media lab at one of Europe's must prolific art universities. The variant I tend to use most is the 3B+.
Reasons:
- full sized HDMI connector
- headphone connector
- good bang for the buck
If I had one wish for any new product in the Raspberry line it would be: Do the Raspberry Pi 3++ or something. Same thing. Faster, but with USB-C power connector, 4K Video resolution, 2× USB-C I/O, 2× USB-A peripherals and maybe M.2 support.
Same here. The 2b, 3b and Zero 2 have the best cost-benefit ratio. I recently migrated the Oberon System 3 to those three boards and particulary the Zero is a great host for this bare-bone system: https://github.com/rochus-keller/oberonsystem3native/. I hope the Raspi company postpones the end of life date of the 3b and Zero 2.
I always wanted to wrangle ~10 of these together and practice distributed computing for the price of ~$200. That was a 2014 dream. I guess instead someone else ran with all of Jensen's polygon drawers and Hynix's backlog. Que sera
Zero 2s can still hit that price, can't they? With much better CPUs than the 2014 dream.
And if you're not worried about CPU you can get a pile of Pi Picos or ESP32s.
Actually, considering that the original zero had a better CPU than 2014 Pis, I'm surprised you didn't get a 10 pack of those when they were extra cheap.
Would love to see actual security focused hardware/software features, like full OP-TEE, fTPM (or a more ideally a real physical TPM), and similar. For example, so that the OTP isn't the only way to store a disk encryption unlock key.
The existing secure boot mechanisms aren't bad, but allowing for more than one public key hash in OTP would be nice, too.
These kinds of things are expected to be on modern embedded SOCs and SOMs now.
You do know those are trivially bypassed with a signal processor, right? If physical access is outside your threat model, that's OK, but it makes (for example) the forced Win11 upgrade for DRM^H^H^H boot integrity enforcement seem ridiculous.
Yeah, fair enough. "Compliance" is probably the phrasing I should've used, rather than "security".
I've been curious for a while about the overall taxonomy of security, especially for embedded platforms. It seems like the only hope is defense in depth, given the power glitching attacks and the like that you can find demonstrated.
Specific to the Raspberry Pi, I believe I even saw a thread at some point where one of their firmware engineers was making the case that secure boot on the Pi 5 was equivalent to a TPM in almost any reasonable threat model, since, in either case, you were out of luck if an attacker had physical access and was willing to put in enough effort.
Normal secure boot does not use the TPM. Secure boot is the proactive process of ensuring only allowed code loads and executes.
The TPM is used for measured boot, the post process to understand what actually was booted and if the right set of things were booted then to allow unlocking of specific items like keys.
Both are important but they are not the same thing.
I was thinking about the RPi 6 yesterday whilst realising I couldn't set up my RPi Zero 2W anymore - the OS has become burdensome - tied strictly to an imager, that gives me an allergic reaction. Yes - they did all this for the uninitiated - but for Raspberry Pi OS Lite - bring back this experience: dd the image, write ssh into the boot drive, SSH in - change password, fully set up in almost zero fuss or effort.
Then I actually couldn't set the thing up because of the mini HDMI connection - I have a mini to HDMI cable, but to use my portable screen with it I need mini HDMI to MINI HDMI. Don't get me started on micro HDMI - almost everyone of of those connectors I've bought slips off or breaks in the device. Every time I go to set up an RPi5 I end up having to order another one of those tiny connectors.
Full HDMI for all new devices please. Even if the second display can't be connected.
These days a 175 GBP N95 from a no-name Chinese OEM on Amazon, with 16 GB of RAM and a 500GB SATA SSD is way better value and performance - and importantly - zero fuss - standard setup.
> __These days__ a 175 GBP N95 from a no-name Chinese OEM on Amazon, with 16 GB of RAM and a 500GB SATA SSD is way better value and performance - and importantly - zero fuss - standard setup.
"These days" seems to be over for now. The prices went up there, and now the only affordable option is an old laptop's motherboards with DDR3 sticks. But passively cooled PCs are rare, so the Raspberry Pi still makes sense. As a bonus, it is easier to power and find a small UPS that will keep it running for hours.
They aren't actually shipping the new RP2350 silicon revision on Pico 2 boards yet. If you want the errata fixes, you've gotta source the chips and make your own boards.
I recently found out about the Radxa Dragon Q6A. A Qualcomm chip with faster CPUs, a good GPU, a DSP and AI accelerator, and a hardware video encoder seems very compelling. It even supports Windows if you want that for some reason.
For me, Pi6 shall focus on NPU to run edge models instead of getting into minipc space, e.g. 10TOPS built-in, that will be slightly better than Rockchip 3588 which was produced 4 years back at 6TOPS, and way more powerful than NXP's 2TOPS. 10TOPS is the sweet spot for edge AI as far as I can tell.
As someone who has just had to migrate a project off of radxa due to ongoing supply issues... good luck getting your hands on more than one or two samples of those :/
The only way I'll buy another raspberry pi is if they come with a power supply that's guaranteed to work with them. I got tired of the random reboots in the night and replaced my media center/NAS with an old Nuc.
Official power supply is sold separately. Pis work fine with unofficial USB-PD power supplies but will limit maximum current for USB devices. I have two Pi 5's and one 4 and they all are perfectly stable when powered with an old Lenovo laptop charger.
It's still a huge pain point. I keep a magic "correctly sized, but provides clean power" USB charger with my Pi in a drawer, because I know that, if I lose that charger, I'm buying a replacement. It's unclear to me why they're so borderline. Maybe USB-C would help?
I don't know much about older models but on 4 and 5 they chose to use a weird power standard to reduce cost. Just buy the official charger. They're not expensive.
> "...: It sounds like the key feature will be 'more': a faster CPU and faster IO, rather than new features."
Raspberry Pi Holdings is a embedded systems manufacturer for pity's sake; we don't need more from them, we need less. [EDIT] A faster Raspberry Pi 6 is encroaching on the territory of the Intel N150 and its successors and mainstream Linux distributions and that is a battle they would lose in terms of price and performance.
Give us a Raspberry Pi Zero 3W with proper sleep states to reduce sleep power consumption, lower idle power while awake, and 1 GB of RAM even if it doubles the price.
^^^ when I tell people tangential to the field that the latest pi needs considerations of cooling solutions and a beefy power supply (no more just any old micro usb cable into any old usb port), they're astonished. It was a "microcontroller" you could program in Python with a friendly Linux environment and is now an expensive, power hungry, hot computer with a microcontroller hanging off of it
I agree that Raspberry Pi is not a good general purpose computer, but some of these criticisms are starting to feel like a pile-on with partially incorrect information.
> the latest pi needs considerations of cooling solutions
FYI you can run the Raspberry Pi 5 without a fan or even a heatsink. It will safely throttle itself if it gets too hot.
If you're trying to get maximum performance out of it all the time, you will want a heatsink and fan. If you want to run some Python scripts in a Linux environment or even if you're doing heavy work and waiting longer is not a problem, you don't need extra cooling.
> and a beefy power supply (no more just any old micro usb cable into any old usb port)
This hasn't been true in 10 years.
Powering something off of any old USB port means it would have to fit within the 5V 500mA basic specification, which the Raspberry Pi 3 exceeded long ago.
> It was a "microcontroller" you could program in Python
It was never a microcontroller by any definition of the word.
Raspberry Pi foundation has released microcontrollers that run MicroPython in a very user-friendly format https://www.raspberrypi.com/documentation/microcontrollers/m...
Raspberry Pi uses USB-C and USB-C is more generous with power. Almost everything except computers provides 15W. With USB-C, devices request want they and there is no reason has to work with minimum possible.
The problem with the Pi5 is that they use weird profile, 5A/5V, that requires special charger. Most 5V chargers are 100W and beefy. If they hadn't cheaped out on power circuits, they could have used normal 30W charger. They should come out with new version that fixes that.
That’s crazy, 5A/5V is not and has never been in range of the USB power delivery spec. It only goes to 5 amps at 20V for the 100W output, or higher voltages in the newer Extended Power Range topping out at 5A/48V.
5A/5V what are they smoking?
Probably the same stuff they smoked when they cheaped out on a resistor on the raspberry pi 4 and made it so chargers thought it was an audio device.
Keep in mind at least 2A of that is to power the four USB ports at their rated 500mA. So if you don't have four USB devices loading the USB ports at their max, you should be fine with 3A.
> Powering something off of any old USB port means it would have to fit within the 5V 500mA basic specification, which the Raspberry Pi 3 exceeded long ago.
Theoretically, devices like the iPhone have lower power consumption than that and loads of performance features like recording 4k 120fps video.
Of course, an iPhone costs much more than a RPi, and has much better economies of scale, so they’re not truly comparable.
> Theoretically, devices like the iPhone have lower power consumption than that and loads of performance features like recording 4k 120fps video.
iPhones use cutting edge (expensive) manufacturing processes for their chips.
They also have batteries inside to cover the power demands while doing intense work like recording 4K 120fps video.
I don't understand what point you're trying to make. The Raspberry Pi is not in the same class of device as an iPhone.
An iPhone will absolutely draw more than 2.5w under load.
Yeah an iPhone 17 Pro can get over 15W if you’re really getting at it.
> > It was a "microcontroller" you could program in Python
> It was never a microcontroller by any definition of the word.
I think the poster means people treat the R-Pi like an MCU that runs Python. The Arduino was popular at the time the Pi came out but limited. Once the Pi landed, it quickly filled the gap and the Arduino's popularity diminished to the point where it's now a corporate Pi clone.
Before we had the ESP32 type products that are now popular, using a Pi was often one of the only easy ways to make simple internet connected devices at home. I remember making smart lights and home sensors, speakers, clocks and all sorts of things using the Pi Zero W. It was cheap, small and easy to get up and running.
I and many other people very much used it like a microcontroller.
> Powering something off of any old USB port means it would have to fit within the 5V 500mA basic specification, which the Raspberry Pi 3 exceeded long ago.
In practice most USB wall warts will happily provide 2 amps or even more with no overcurrent protection at all. The 500 mA limitation, at least outside of ports on computers (where you'll maybe get a prompt on Windows and macOS), is theoretical.
Irrelevant...
They need to stick within limits to avoid liability (and avoid arseholes blaming them on forums).
Someone would get a problem with their computer or TV, and then blame Raspberry Pi (or create a problem).
> FYI you can run the Raspberry Pi 5 without a fan or even a heatsink. It will safely throttle itself if it gets too hot.
What's the point of doing so though? If you're doing this, you're obviously using the wrong device. If all you need is to run some python scripts in a Linux environment, you should use a Pi 3 or Pi 0w2.
Agree with your other points.
Even for extended workloads, a Pi 5 without a heatsink is still a hell of a lot faster than a Pi 4. And as sibling says, most users appreciate bursty speed while not doing prelonged compute (see also fanless laptops).
(Disclaimer, I work for raspberry pi ltd, not views of employer etc.)
Having the burst performance is helpful.
Most light workloads are very bursty. When you type a command or click on something you want latency to be low. Having the overhead to get it done quickly at the full clockspeed is good if you are latency sensitive.
Throttling has become a bad word. Some feel compelled to avoid it at all costs, doing things like buying big coolers and running synthetic benchmarks to avoid it. Unless you're doing sustained workloads where you need all of the performance, allowing a little throttling is fine.
I've been doing some heavy SDR lifting with a couple of my Pi 5s, and my own experience is that the active cooler works extremely well, and more often than not the fan can be shut off and it will work well as a passive cooler.
If you're consistently doing bursty loads that throttle, the thermal cycles are going to do some damage to your chip and it'll fail sooner than if you ran it cooler.
Yeah, in 20 years instead of 100 years.
I'm going to replace it before then.
So that means what, exactly?
Do heat cycles mean that the device only lasts half as long as it might if it were kept at some constant ideal temperature? A third as long?
If so, then...so what? The tinkerers using these things aren't broadly concerned about these things. We're all out here running our real-live desktop, laptop, and pocket computers with bursty loads, dynamic clocks, and dynamic heat anyway, and these things are generally doing just fine. We aren't driving space ships with this stuff; it'll be alright.
I insist that there is no merit to holding a lowly Raspberry Pi board to a higher standard than we hold everything else.
Furthermore: If perfection is necessary for some kind of application, then maybe starting with a <$100 hobby SBC isn't the best move. It might be time to look within for a better pathway.
> It was a "microcontroller" you could program in Python with a friendly Linux environment and is now an expensive, power hungry, hot computer with a microcontroller hanging off of it
The Pi project was never originally a microcontroller - it was always a full-blown SBC you could program any way you want with some GPIO pins attached. People literally used them as (slow) home computers.
The company didn't sell its first microcontroller until years later in 2021 with the Pico, by which point we already had Pi 4. I do though think its a real shame prices for the SBCs have risen as they have.
At one point, the raspberry pi was a decent option if you wanted something hobbyist friendly that could toggle GPIOs and connect to the Internet (and later Bluetooth).
I suspect Espressif has mostly taken over that market now
Yeah every "hardware hacker" I know has pile of ESP32 at home now instead of pile of raspberry pi's.
The pi, particularly the pi zero is still useful if you need something that can run normal software but not a full mini pc. One example I've seen is using a Pi zero as a "wireless usb" where you can plug it in to a machine that accepts files over usb, and can now drop files on to it over the network.
Maybe you could do this with a ESP32 but it's easy on linux where you can use all the normal tooling and filesystem drivers.
ESP32 can run FreeRTOS, and I think people nowadays have no idea how much stuff we could do in MS-DOS PCs even with all their limitations for the epoch.
ESP32 hardware is much better than they used to be.
Not everything needs to be under Linux monoculture, thankfully.
The Pi Pico is exactly that.
It's still true that people, out of convenience and familiarity, used Raspberry Pi for tasks where a microcontrollers would have been perfectly adequate
There was definitely usecase overlap due to the presence of the GPIO, but huge numbers of Pis ended up doing things a microcontroller can't - stuff like the PiHole and Retropie projects, and never used their GPIO pins at all.
Thinking of any of the early Pis as microcontrollers ignores a huge amount of the ways in which actual end users interacted with the thing, and even the way it was sold and marketed. Upton was trying to replace early hacker-friendly home computers like the BBC Micro/Apple II, for a new generation.
> but huge numbers of Pis ended up doing things a microcontroller can't
Mate you know full well it was multiple things... Marketed to education as an actual computer, to the maker demographic as a microcontroller, as a way to learn coding but also robotics etc
>Thinking of any of the early Pis as microcontrollers ignores blah blah blah
Same applies to thinking of them as SBCs
This just clarified something for me. I've always been annoyed when I see a Pi with nothing connected to its GPIO header; why not just use a cheap thin client? Or an old laptop, for that matter? But that's missing the point. Here's the point:
Pre-Beagleboard-and-Pi, if you wanted an programmable thing to work with GPIO, you used an Arduino or a BASIC Stamp, or just a plain old PIC. But they wouldn't run a real OS.
Pre-Beagleboard-and-Pi, if you wanted an embedded Linux box, you used a WRT54G or a Soekris or an old laptop. But getting GPIO out of them was a PITA. (And often involved lashing an Arduino to the side.)
The Beagleboard (released in 2008), could finally do both. It had gobs of I/O and first-class support for it under Linux. It was pretty affordable. Then the Raspberry Pi came out in 2012, with a similar amount of GPIO, but demolished the price point to where it made sense to use it in place of a microcontroller.
That's really the magic of the Pi. You can keep one cheap gizmo around, and use it to solve (a large fraction of) two classes of problems. It doesn't fully replace everything a PIC or a PC can do, but it replaces an awful, awful lot of them.
> I've always been annoyed when I see a Pi with nothing connected to its GPIO header; why not just use a cheap thin client?
There have also been times when Pi's were cheap enough and x86 idled so power-inefficiently that you'd save money over a reasonable time horizon if you couldn't run your old laptops at full throttle.
Absurdly extreme example, but at one point I decided to replace a couple (maybe 3) RPi's with a single old Dell rack server off Ebay plus replaced my router with one running pfsense. I knew it would be mostly idle, that thing had 2 Xeon processors to replace 3 cheap ARM processors.
Between the 2 rack servers, my power bill went up by enough to buy a new Pi or two every month. It was like $80/month extra in power bills.
Then the esp8266 came along, offering wifi and gpios and an arduino-like programming experience at a price point of literally $1
And the ESP32 is offering CPU power on par with many of the embedded-x86 SOCs that powered some of the early wifi routers...
A $2 Chinese ESP32 dev board is more stout in many ways than the first Pentium machine I had, which sure did feel fast-enough to run the whole world with at that time. :)
In this amazing world full of such inexpensive choices, it seems so bizarre to me to demonize any of them. They're all excellent in some way, and it's OK that there's more than one.
This isn't football, handball, American politics, or Highlander: There can be more than one. It's OK.
This is what I keep repeating in many threads about ESP32, the little thing is actually quite powerful, especially for those of us that experienced MS-DOS PCs.
On the other hand, the RP2350 actually is a microcontroller, and IMO a nice one for many purposes. PIO, high-quality datasheet, nice ecosystem, etc. And the Pi Zero 2(W) can do most things the Pi/Pi 2 could, with a smaller footprint and less power consumption. Variety is nice.
And to the GP's point - the Pi Pico can be programmed in Micro Python.
The Pico is amazing and is by far the best thing they've ever done.
Something can't be a microcontroller if it runs Linux.
I wouldn't quite go that far. Linux runs on nommu systems. With some psram you should be able to get a version to run on at least the rp2350 using the riscv cores with relatively minimal fuss.
Is that a good idea? Well, not really.
You can run Linux on this
https://dmitry.gr/images/uARM_big.png
Bare metal or bare bootloader only right?
Or stuff like FreeRTOS, NuttX, Zephyr, and co.
Personally I have never seen Raspberry Pi as anything but a small personal computer. I've maintained a home server which get upgraded as Pi's upgrade. It runs well and manages my media, files and now running hermes agent. I remember even the first time it was introduced to me was with look it runs linux and you can probably run VLC on it.
Performance per Watt still outranks any other (quasi-)mainline linux device
Do you have a reference for this? Looking around, I see it being beaten by other ARM SBCs, and even low end Intel devices.
Many years ago, I measured performance per watt of the original Raspberry Pi when they were still relatively new. The performance per watt lagged behind even a beefy Intel box since the original Raspi was so slow that it destroyed any gain it got from using so little power.
EDIT: One set of benchmarks I found as an example: https://bret.dk/raspberry-pi-5-review/#Performance-Per-Watt
They've never been particularly low-power, in a performance-per-Watt sort of way compared to other offerings at whatever present day. In recent times, I've seen completely-believable reports that N150 boxes walk all over it.
They've never been particularly cheap, in a performance-per-dollar sort of way. Used machines from eBay, yard sales, and old broken laptops (that still compute!) have always been better. (They usually come out OK when new is compared to new, though, which is IMHO the only valid comparison.)
Those comparisons were never very favorable.
---
The parts where it shines are: Small; they come in two sizes, and those sizes are small and smaller. That was new, but it's been cloned all over the place.
Built-in GPIO that's meant for people to actually-use and tinker with. That was new, too, but it's also been cloned. (Also: These days, anyone can plug a cheap Pi Pico into any PC with USB and get a fairly intense amount of GPIO to goof with.)
Standardization, and the appliance-like behavior this can enable. Lots of folks, including kids of all ages, just download pre-built images and swap SD cards like they would cartridges in a Nintendo. That's not for me, but it's pretty neat.
Community. They've still got a lot of momentum by being first at these roles. That's good. It helps newbs (who at this point may have never had anything resembling a "real computer" to play with ever before in their lives) to get started.
Low-power enough: It's not ideal when chasing tiny Watts for battery or small solar power, but you probably won't notice it on an electric bill (and despite the bizarro-world cooling rigs people put together, a passive heatsink really is good enough to keep it running in-spec).
raspberry pi has terrible power management as well. turning off a core was impossible on previous generations for example. a few years ago rpi was the worst of all sbcs we measured for battery powered usages. this was for an actual embedded product so it wasn’t going to be used either way (ask broadcom for some chips HAHA)
but there is very little reason to use a rpi over other sbcs if you have a remote idea what you’re doing beyond hobby use
Performance per watt is nice but I’d be more inclined to talk about “problems being solved per dollar.”
If you don’t specifically have a project where you need the GPIO pins built in, I struggle to understand the use case proposition of a raspberry pi compared to a typical x86 mini PC or even just grabbing a think client desktop like a ThinkCentre.
Almost everything that is unique to a Pi compared to an x86 mini PC seems like it makes more sense with an ESP device.
When the Raspberry Pi was $35 and it ran a desktop OS and the cheapest alternative that did that was 5x the price that use case made sense.
Can show us where to find a new ThinkCentre for the price of a new Raspberry Pi 5 kit?
The ones I find on Lenovo's website don't seem to get any cheaper than $649.
I’m talking about used on eBay, obviously.
Or do a search for “mini PC” on aliexpress.
They'll do whatever they do.
Maybe a tick-tock release cycle (one with new features and some speed, the next with the ~same features and more speed) is where they're headed, and maybe that makes sense. They wouldn't be the first.
I'd love to see even-lower-RAM versions, though. Most of what I use Raspberry Pis for at home for is not RAM-hungry at all.
My Pi4 network router has 2GB because that was the smallest/cheapest version at release when I got it, but the system itself consistently only uses about 64MB of RAM. It'd do perfectly well and have a ton of breathing room with just 128MB of RAM (which will never happen, but if it did happen...).
I suspect the Pi4 that I use as a set-top box with Kodi would be fine with 512MB.
I've used Zero Ws for all kinds of things over the years and never felt RAM-starved with their little 512MB of RAM.
So I'm learning towards 512MB.
But sure: 1GB options would also be fine even if it does double the price. Our comments serve to demonstrate that there's room in the marketplace for different SKUs with different memory capacities. :)
We need a 5 lite. The specs of the 5 without the massive power draw and heat generation
At least now they literally cannot pull more power over USB 5V, because they maxed out the spec. It's even slightly beyond it.
Ideally each RPi generation should keep the same price (or lower now that it's gotten so high) but with better performance. If they can't do that they just shouldn't create a new generation.
Sure (inflation adjusted), but I see no harm if they also provide higher specced SBCs.
I don't think the Raspberry PI foundation has the buying power to control RAM prices.
The price was creeping up long before the DRAM shortage. They're totally doomed now.
> Raspberry Pi Zero 3W with proper sleep states to reduce sleep power consumption, lower idle power while awake, and 1 GB of RAM even if it doubles the price.
Yes, that would be kind of a dream device, perhaps also if it could suspend the os when asleep so it doesn't have to boot every time but I guess that might the standard way of doing it.
I'd be interested in seeing them partner with some of the other interesting and open-ish players, like Pine64 or Radxa, to push for a more standardised, unified SBC landscape.
But that's not as good for PR as "bigger, faster, better" even if that comes with the problems you mentioned.
I wonder how they are positioned now in the market.
During covid I wanted a small low power always on server. I thought about Raspi, but at the time it was expensive and I went with an intel nuc, for a similar price.
Now if I wanted to do hobby electronics, I heard I should look into esp32 or stm32..
I use a Pi 4B as a 24/7 home server in a country where domestic electricity is expensive (and worsening).
Each Pi release is more powerful, but uses more energy. I found the Pi 4B to be the sweet spot for me, because it is the earliest model to support USB booting, gigabit Ethernet, and offer > 1GB RAM.
Perhaps a used one would fit your purposes and budget?
I currently use it to run PiHole, serve media via SMB, host Postgres & Redis, and run some custom written Dockerized apps. Home Assistant to possibly follow, too. The current load seems reasonable in htop, but I haven’t looked into burst scenarios.
Well... "hobby electronics" is extraordinarily broad :)
It depends massively on what kind of DX you want. If you want to work with a 'regular' operating system, you're looking more in the RPi direction.
If you want to write straight-up C firmware, then yeah, the esp and stms are both great.
The things that keeps stopping me from using one for my server is the IO. the PCIe lane on the Pi 5 was a great addition, but it isn't quite sufficient.
Raspberry Pi isn't in direct competition with N150's.
Their niche is the industrial/embedded space. For that market, power consumption doesn't matter. What matters is that each model is guaranteed to be available till a specific date.
> Raspberry Pi isn't in direct competition with N150's.
They may not have intended to be in direct competition, but in the current crisis conditions they are priced about the same as equivalent RAM/storage N150s, have even worse supply issues than the N150s, and have worse performance/watt than the N150s.
Its mighty hard to recommend them for new projects at the moment (nor any of the Pi clones, which are also rocketing in cost and dwindling in availability)
Among the people making things like a DIY NAS, who want fast USB, lots of cores and RAM, small-ish, not-too-bad power consumption, running Linux; and not caring much about GPIOs or passive cooling, it’s in competition with the N150
Among people who want GPIOs and network connectivity, a low price and an open, microcontroller like experience, not caring much about USB speeds and lots of cores and running Linux and suchlike, it’s in competition with the esp8266 & esp32. And the previous generations of RPi.
The Pi SBCs are squeezed in between the SFFs and the microcontrollers for many use cases. It's only a good fit if there is a need for the GPIOs/Linux/Display. For anything else, there are better products out there.
It is for those of us that rather play around with an ARM CPU, and first class Linux support.
IMHO people expected a Pi that offered similar performance as a Mac M1 but with Linux as first class citizen
I mean, they're still on some ancient node like 28nm last time I checked.
Is there some serious astroturfing going on with the N100/N150, or am I just jaded?
I have a bunch of old intel atom boards laying around. The Intel Compute Stick (TM) burnt out its flash root drive in a few months. The C2000 board I had burnt out the clock pin to drive the bios. I have a Clover Trail with a PowerVR GPU (I thought I was getting an intel GPU because it was branded Intel Graphics or similar, but nope!) that lost Windows support very quickly after launch, and has no GPU drivers for any other OS.
Instead of being fooled 4 times in a row, I looked into using an N150 for a NAS, but this time I held off a bit until after launch so I could research it first.
Lo-and-behold, they all have crazy PCIe / memory subsystem data corruption issues. I guess there are some chicken bits for the OS developers to set if the kernel can stay up long enough after boot without a panic.
Why would anyone buy this for a NAS / embedded use case?
> I looked into using an N150 for a NAS [...] Lo-and-behold, they all have crazy PCIe / memory subsystem data corruption issues.
Source? I've never had a single problem with PCIE on N100/N150/N200.
I have had a ton of issues with drive corruption on the pi, both via USB3 and PCIE.
https://www.phoronix.com/news/Linux-6.4-Lands-PCID-INVLPG
https://forum.opnsense.org/index.php?topic=48343.0
All CPUs from all vendors have tons of bugs like this, which are mitigated in the operating system kernels, e.g. in the Linux kernel.
I am pretty certain that the Linux kernel must also contain specific code for various quirks of all Arm CPUs that have been used in the various Raspberry Pi models.
Intel had indeed several bugs that were more ugly than usual in their recent CPU models, like also MONITOR not working correctly in Lunar Lake, but even so, Intel still has better documentation for their CPU bugs than most vendors of Arm-based CPUs.
In any case, the bug that you linked was solved in the kernel years ago and it affects a privileged instruction that cannot be used in user programs. It does not have any direct relationship with memory and PCIe corruption. Memory corruption can occur inside the operating kernel only in certain circumstances, when the kernel changes the mapping of global memory pages and then writes the new pages, but the writes go to the old pages. However, this could happen only until 3 years ago, before the bug was known.
so a single 3 year old errata for the n100 that was long since patched?
I've been running an N100 for 3 years with a 5 bay external enclosure over USB 3.2 Gen 2 and ZFS, and have not had any issues. It is pretty phenomenal, pulls about the same power, and costs around the same as an RPi 5 but provides substantially more compute and throughput.
I think exactly the opposite: we have no shortage of embedded crap we can buy; what is useful is dismembering intel. It would be better if the pi were risc v but this will do for now.
A Raspberry Pi with sleep and hibernation is like asking Valve to make Half Life 3. They just can't. It doesn't compute.
The entire raspi foundation is a marketing department. Raspi products is making zero sense ever since the Pi 2.
> encroaching on the territory of the Intel N150.
It’s nerdier to say you built your homelab on a Raspi, and that’s what keeps the foundation afloat.
Pi's refusal to drop a USB-C on Pico due to cost increases is a terrible call IMO.
I seriously cannot fathom being someone doing development who wouldn't pay $0.50 extra to purge the last micro USB from their desktop.
A very small point, but pulling from a feather form factor BOM to compare.
$0.12 for microUSB female connector (rated 1A) $0.26 for a USB-C female (rated 3A). Needs 2 x resistors (< $0.01), 20% larger board area
I think the power capabilities are the biggest item. If you want to pull higher current from a laptop for development or supply from a wall, you have to switch to USB-C.
I don't think either of these prices are that aggressive - pretty sure the cost comes down at volume.
I wonder if it would be worthwhile for them to produce both. Well, it will be hard to compare because the design cost doesn’t show up in the BOM, haha.
But it seems like it would be useful nowadays, since some laptop have mostly USB-C connectors, and USB-C to USB-C is pretty common. I’ve never seen a C to Micro. Do they even exist?
I have an unfair bias because I design PCBs as a significant part of my job, and switching out to USB on this board appears to be a non-issue.
I have a Pico in front of me, and there's plenty of room there for a USB-C footprint and the two 5.1k resistors. Given that, I cannot reasonably agree that the "design" stage is significant.
In other words, it's a change that I would make to my own board in 2-5 minutes because the stakes are low. My ballpark guess is that such a change at RPi would have to go through a proposal stage, a PCB change review, and then there would be dozens of places to update documentation.
Since backwards compatibility is non-optional, this would result in a separate SKU, which means that the whole distribution chain needs to be updated with a new product.
So, I acknowledge that when you're working at their scale any change like this is the definition of non-trivial. What I don't agree with is the conclusion that it's not still clearly the right thing to do.
One caveat is that the pico isn't so old that micro usb wasn't a weird choice even when it came out
> I’ve never seen a C to Micro. Do they even exist?
They do, in spades: https://www.amazon.com/3FT-Micro-Data-Charge-Cable/dp/B0DDWH...
I look forward to the day when they're no longer necessary.
I've always used USB-C to USB-A dongle + USB-A to micro B.
But then you loose market share to ESP32 where people just get USB C on their own.
Comparing an RP2350 to the ESP32 family (which is broad) is very much apples to oranges; they each have feature sets which make them ideal for completely different use cases.
I get my USB-C connectors at around $0.08 at low volumes (LCSC).
https://www.mouser.de/ProductDetail/Same-Sky/UJ20-C-H-G-MSMT...
I'm on a phone that makes it a pain in the ass to switch languages. Do it yourself.
I cannot fathom why USB-C would make any difference. It's not like USB-C is intrinsically better for this use case? If you're doing hardware development, your desk is likely full of in-development crap anyway and a micro USB cable more or less won't make any difference whatsoever, nor is the Pico likely to the only thing needing a micro USB cable.
Edit: one thing I can think of where micro USB connectors are better: if you broke off the connector, it's much easier to solder it back on.
Micro USB is relatively brittle and can only be inserted in one direction. It means you have to make sure you have the old cable for this one thing. There's no USB-PD and you can't do USB host mode.
USB-C is the opposite of all those cons: much more durable, not directionally opinionated, you can use one cable for everything, you can do USB host and USB-PD.
It's also clearly the future everything is standardizing on. There's value in embracing that, since it is a Reference Board.
USB-C is a better connector in every aspect. Micro USB is a terrible connector. It's that simple.
The economics don't add up either, because an adapter cable costs money too.
> If you're doing hardware development, your desk is likely full of in-development crap anyway
Sure when you're at home, if you have a stable home. It's one more cable to bring if you don't, or if you ever travel and stay in a hotel.
For what it's worth there are third-party rp2350 boards with USB-C connectors if that's important to you. Heck, WaveShare has one with two USB-C connectors: https://www.waveshare.com/rp2350-usb-c.htm
I was just needing exactly something that for my project, thank you!
This way, I can have my laptop running codex control over a target laptop, by pretending to be a keyboard.
I am aware, thanks.
I do think that you're missing my point, which is that we're significantly past the point in this wretched timeline where they should offer a USB-C version of the reference board for this MCU.
Oh, no, I'm with you. The fewer connector types we have to use, the better. My comment was meant as an FYI for you or anyone else who wasn't aware and not as a dismissal of your valid criticism.
They probably just bought a shit ton of micro USB connectors back in the day and want to use them up, or something silly like that. It would be funny if the EU forced them to switch to USB-C.
You can at least buy USB-C boards from other vendors since they sell the rp2040/rp2350 separately. If you want wifi it gets a little more complicated unfortunately.
i understand that if they implement a port, it is usually well done (besides the unfortunate 27.1w 5a bs for the rpi5). so unlike cheapo electronics that do come with barely working type c.
however, while in one hand we are happy to (albeit temporarily) raise prices based on ram situation, in terms of design, there is simply not enough money for the port. especially when now they are adding ecosystem items that do cost money to develop and maintain.
this explanation made sense pre-ipo but no longer imho.
There are broadly available third party RP2350 boards with subc and a variety of additional capabilities if that is important for you.
https://shop.pimoroni.com/en-us/collections/rp2350
https://www.sparkfun.com/sparkfun-pro-micro-rp2350.html
https://www.dfrobot.com/product-2913.html
https://www.seeedstudio.com/Seeed-XIAO-RP2350-p-5944.html
Not the point, though I appreciate folks listing links in an attempt to help.
What I am saying is that we're well into 2026 and there's no good reason for RPi not to offer a USB-C version of the reference board for this MCU.
The good reason is that there are plenty of third party boards that already offer what you want. There’s very little to be gained by an ‘official’ one. The next one probably should have C, if just because it is the Euro standard, but no urgent need to backport.
The point is just blowing past you, friend.
I don't need a Pico or any other 3rd party board. I drop SC-1511/12s packages on my PCBs as needed.
What I am saying is that the reference board for an RP2350 should have a USB-C port in 2026. It's not aesthetics or even convenience (and it's definitely not price) so much as establishing best practices for how a part should be used.
A big part of that is to acknowledge the context in which a part exists, and in this case, it's both a fact and a very good thing that the world has embraced USB-C. It's even being regulated in many cases.
I'm not saying that you can't smoke, just that maybe you shouldn't do it when you're volunteering as a Big Brother.
While we're at it, their reference board also doesn't have a reset button, it just has one for boot. It's perhaps one of the most inconvenient official dev boards I've ever used in modern times.
It's not $0.50 extra. It's $0.16 extra for a USB-C port assuming you bought the USB-C port on mouser at 10k quantities and threw the micro USB port away.
I just don't get it. Anyone who wants to save a few pennies just buys the chip directly. Their Pico board is primarily for prototyping and one off products, where quality of life is everything and 16 cents is nothing. The adapter cable probably costs more than the amount they saved. That's a dick move.
Unpopular opinion but I actually enjoy using USB Micro B more than USB-C. That's because USB-C is much more complicated and there're non-standard-compliant cables floating in the market. USB-C also has some fancy mode like voltage selection. If it got screwed up somehow leading it to supply the wrong voltage, it could fry your Pico board.
That and a reset button.
> When asked about the Pi Zero 2W, Eben said the substrate supply is constrained—basically, so many AI chips are being made that even older chips using older process nodes have to fight for the actual silicon wafers to use to make the chips.
I keep hearing voices invalidate each other, is the bottleneck the raw silicon substrate, or fab capacity?
What purity levels are required for say Pi Zero 2W?
The volume of monocrystalline silicon used in solar panels is orders of magnitude greater than the volume used in IC's / RAM production.
Is the actual bottleneck 11N + grade silicon wafers while 6N to 9N grade used in solar panels remains unaffected?
The flagship Pi boards are also hitting the thermal design ceiling of what makes sense for this size or platform. Pi 5 is IMO already on that edge because it pretty much needs active cooling and a dedicated power supply. Going past this point means it will compete with much more powerful platforms and likely loose due to it’s architectural limitations.
The 8GB Pi 5, at $170 [1], is encroaching on Jetson Orin Nano Super's $240 price point [2]. But the Jetson has a faster CPU (newer a78ae cores rather than a76) and, obviously, a whole-ass GPU.
[1] https://www.microcenter.com/product/673711/raspberry-pi-5
[2] https://www.microcenter.com/product/691058/nvidia-jetson-ori...
Nvidia's software platform for the whole Jetson series was, at least in my experience, absolutely awful on the Jetson Nano and Orin boards I worked on. Has that improved at all? I did not appreciate that the only option they provided was a full desktop version of ancient Ubuntu... and even flashing the OS image was a bizarre process.
Edit: looks like they at least have a better headless option now.
Nowadays upstream Linux with UEFI mostly works, with their out of tree drivers. I’ve managed to make it work in NixOS with the stock kernel. Look at the open embedded L4T project, they have some recipes for building that. No need to use nvidia’s kernel anymore!
Also, supposedly on the second half of 2026 they were going to be moving even more stuff out of their Jetson-specific drivers as they already do for their slightly newer chips (so you could use the standard drivers, and standard CUDA builds). Let’s see how that turns out.
> with their out of tree drivers
rephrase please?
They’ve got external kernel modules that they use for some hardware that is not supported upstream, namely the GPU, but also some things like Ethernet I think. Everything else is upstream though, so the situation is similar to a regular NVIDIA GPU.
Who would’ve thought that NVIDIA started upstreaming stuff once they realized how much money Linux is making them?
ahhh, it's out-of-tree as in one word
I thought it was "out of three" or something
human parser error :/
Yeah, one of my bigger complaints especially on the Nano was the GPU only had really limited model support (iirc, mostly tflite but maybe I'm misremembering) and it sounds like the newer ones are more normal. That and what seems from the docs to be better headless support would be major improvements. Going further to mainline distro support would make them interesting to me again.
I was always disappointed by the Nano as it was a pretty capable device, but it seemed like not many people picked it up as a platform for cool things which I always attributed to the software.
Yup. At this moment in the embedded world Jetson Orin Nano is the best value for money purchase. Absolute beast at the price point.
The 16GB Pi 5 makes the Orin Nano look afforadable.
It used to be that raspberry pi was a cheap pc. Well it's not longer cheap.
And at their price point, you could just get a mini PC and have better performance, or if you want to use it as a microcontroller, you can just use an arduino, esp32, or an actual microcontroller for a fraction of the price and power consumption.
So, what do people actually do with these pies?
I've got one set aside to play with v4l2m2m. Once that's working, it should improve video encoding support in my OSS project (https://github.com/mickael-kerjean/filestash)
The form factor. It's tiny compared with mini PC. Unlike other SBC, the software support's good that it just work as advertised (GPIO, graphic hardware acceleration, etc.). If you're using it for work, it also has stable supply chain / commitment that you can ensure to be able to obtain a unit in at least the next few years.
But you're right. It's pretty overpriced for its performance. If I'm after the GPIO I'd just use a microcontroller. If I'm setting up a headless server I'd just use pretty much any other SBCs or mini PC. I'm personally not a big fan of Raspberry Pi.
The cheap Raspberry Pi's still available as Raspberry Pi Zero tho, if that's what you're after.
Getting a mini PC with ARM and first class Linux support from the OEM.
Exactly. I think they are a bit more long lasting, but that is debatable.
Hard to care anymore. They made it abundantly clear they dont care about their original audience of makers and education anymore now that the industrial market are so reliant on them. Not to mention the whole crapping all over their customers on social media and then accusing them of some sort of coordinated 'attack' just made it clear they're clueless about their audience.
The price increases were the end of the Pi being a viable option for most application in my eyes.
The PI5 was a real downgrade for me with its lack of proper hardware h264 decoding as in the earlier versions (playing back h264 is my primary use case.) I will buy a PI6, if that comes back, else I stay with my reliable, and passively cooled, PI3.
I have bought an rpi at every generation. And I still have yet to find an actual use for them.
Everything they do from a compute perspective is just better with a mini pc or old laptop with a mobile spec chip.
Everything they do from a programmability perspective is just better with a microcontroller specific to the task.
I just don't see the actual market position for these things. They were supposed to be a cheap board, but you can't actually buy them cheaply because the vendors upcharge so much.
Here are a few places that I've found pi's:
Rapid prototyping: I created a PoC to take a webcam snapshot every half hour and upload it to a server in an afternoon. Freelance project. Could it have been done with a microcontroller? Yes, but not in 4 hours.
Local digital displays in a gym: we built a system with a number of overhead 60" TVs, each with a pi on a VESA mount to show scheduling and workout information in a gym for a client.
HVAC controller: bid on a project where the customer's original concept was a Pi managing a rooftop HVAC system for large buildings. They outgrew the pi and wanted a new solution.
Data aggregator: collecting sensor information via BLE (bluetooth) and uploading to an internet server over Ethernet.
Remember that "cheap" to a consumer doesn't have the same definition for a business. To most profitable businesses, a $100 computer that fits on the back of a TV and consumes a fraction of its power is cheap. In fact, one of the reasons that Pi's were so hard to find for a while is that the Raspberry Pi foundation was prioritizing industrial/commercial customers over hobbyists.
How do you find these kind of jobs? That’s something I think I would enjoy to do but I’m not sure where I would start looking for customers. Is it world of mouth?
For a hobbyist, they're quite nice for "I want to SSH into a thing, write my tools on Linux, and still have access to SPI / I2C / GPIO, USB, and whatever hats can plug into that." The Hat form factor, while not technically great and frequently overpriced, is also nice for distribution; both inside a team commercially and on the web as a hobbyist, it's a lot easier to share software and say "hey, buy this pi, this hat, and run this" than "fab this PCB / solder this bundle of nonsense to an ESP."
For a company, they're also nice for "I want to make an IoT device that's heavier weight than an ESP32 and/or I only want to hire Linux Application People and not Firmware People; what's the cheapest Linux module I can get that's widely supported, backed by a real company, and has regulatory approvals" - Pi Zero W. My understanding is this exact pattern is why it's harder to get them as a hobbyist.
I use them widely in automotive reverse engineering; ESP32 can and does work just as well or better for an end product, but for experiments it's really nice to have a self-contained appliance to SSH into and use SocketCAN on rather than some bespoke firmware project to manage and iterate on.
Given the price and availability issues I suspect the market is "correcting" a bit and companies are hiring Firmware People and switching to true MCUs in places they'd previously have avoided doing so, but it was definitely a thing for a long time.
I think I found a use case, but would love to be proven wrong with some faster / cheaper approach:
I've tired of buying S1 compatible sonos speakers on ebay, so I'd like to build a speaker enclosure with a WiFi device that has a high-quality DAC and the ability to use pipewire or similar to do real-time DSP and multi-room audio sync.
A RPI + third party hat should work well for this, or so I am told.
I have spent 100$+ on multiple esp32s3 amplifiers instead with multiple software solutions ready to flash and be up and running in no time.
https://sonocotta.com/esp-products/
Now I have a wifi radio running in homeassistant. Running my own (music only) radios with azuracast. Loud enough for high quality 30+ year old 10kg 100/140 Watt speakers. Next step is making a remote for it, cheap ikea one with two buttons or a tap dial with more.
This sounds like a great approach to me. I've thought some kind of pi zero 2w based smart speaker system would be an awesome project for a while. I am having a difficult time imagining a cheaper option, especially if you want something like the hat ecosystem.
> I have bought an rpi at every generation. And I still have yet to find an actual use for them.
It's amazing how well these fit into the category of products that people feel compelled to buy, play around with, and then forget about.
Flipper Zero is another product that landed in the same space.
What's sad is that Raspberry Pi does have a lot of legitimate use cases and people who want to use them, but the supply has always been swamped by all of the demand.
I think at this point the brand reputation and software quality are a big selling point.
If you're trying to build a couple of units of some embedded thing where you need to toggle some GPIOs or serial devices in response to requests over the network, but don't have the expertise or resources to do it with a microcontroller, a Pi is a great option - you know you'll have software support, and you know that the vendor will be making the exact thing you bought for 5-10y.
For hobbyist stuff at home, I agree, though. A mini PC is probably better for homelab stuff, and an RP2350 or ESP32 is probably better for anything embedded or battery powered that you want to do.
Development speed is also one of the forgotten axes. I mentioned upthread a system I built that was done in less than a day. There's no microcontroller solution I know of that would have let me deliver it that fast.
I've been playing with Circuit Python (variant of Micro Python, etc.), and it seems quite great actually. For small projects it's far simpler to use than C/C++, although if you're using some peripheral that isn't just toggling pins or simple serial commands you need a library (module). But then in that case you'd probably want a library in C as well. Definitely recommend it :) (it works on ESP32 boards as well as similar recent microcontrollers)
My use cases:
* Replacement controller for my UFO Catcher - It has WiFi, easy to update, and I can operate the machine remotely with it. It's bolted to the back of small touchscreen that lets me change the machine settings as well.
* Remote printer access - I can monitor from the USB cameras and gather statistics about the prints.(I suspect a lot of 3D printing enthusiasts use them for this purpose.)
Having a small low power computer has been useful for me in those instances.
> I just don't see the actual market position for these things.
Isn't it mainly for learning and hobby-ism?
I have one running Klipper on a modded Ender3.. but I agree with the sentiment.
They also make OK Kodi/Libreelec boxes due to better documented video decode silicon (Intel still better at that).
Which mini PCs are available at the same price with ARM CPUs?
I would tend to agree unless you need those gpio pins. Then there are few other good options.
They have great software support and Google-ability, which also means LLMs understand how to work them. I use them in scenarios where I want both Linux and GPIO. Specifically, right now I have 4 Pi 5s running kiosk displays with RGB LEDs around the outside of each. The displays show a web page, requiring something powerful enough for modern Firefox, and the LEDs are synchronized to the state of the web pages. You could do this with a mini PC and a microcontroller, but it’s just way easier, and cheaper, to use a Pi.
My opinion is that Raspberry Pi has to release an NPU, and start/revolutionise the open source NPU communities and tech. Raspberry Pi's has to find itself useful in vision AI applications (extremely common in industries these days). Without it I think Arduino with its new Qualcomm boards will kill it.
I really really wished they released a Raspberry Pi running a RISC-V 23 processor :(
I run the media lab at one of Europe's must prolific art universities. The variant I tend to use most is the 3B+.
Reasons: - full sized HDMI connector - headphone connector - good bang for the buck
If I had one wish for any new product in the Raspberry line it would be: Do the Raspberry Pi 3++ or something. Same thing. Faster, but with USB-C power connector, 4K Video resolution, 2× USB-C I/O, 2× USB-A peripherals and maybe M.2 support.
Same here. The 2b, 3b and Zero 2 have the best cost-benefit ratio. I recently migrated the Oberon System 3 to those three boards and particulary the Zero is a great host for this bare-bone system: https://github.com/rochus-keller/oberonsystem3native/. I hope the Raspi company postpones the end of life date of the 3b and Zero 2.
Looks like the RPI5 is like $200 in Canada now. We might have lost the plot
The Steam Deck price in Canada went up by something like 50% yesterday also - it's now over $1100 for the 512GB OLED, up from just under $700.
Yeah, we have. We decided the world's RAM production capability would be better suited for datacenters so we can cheat on our homework.
I always wanted to wrangle ~10 of these together and practice distributed computing for the price of ~$200. That was a 2014 dream. I guess instead someone else ran with all of Jensen's polygon drawers and Hynix's backlog. Que sera
Zero 2s can still hit that price, can't they? With much better CPUs than the 2014 dream.
And if you're not worried about CPU you can get a pile of Pi Picos or ESP32s.
Actually, considering that the original zero had a better CPU than 2014 Pis, I'm surprised you didn't get a 10 pack of those when they were extra cheap.
Would love to see actual security focused hardware/software features, like full OP-TEE, fTPM (or a more ideally a real physical TPM), and similar. For example, so that the OTP isn't the only way to store a disk encryption unlock key.
The existing secure boot mechanisms aren't bad, but allowing for more than one public key hash in OTP would be nice, too.
These kinds of things are expected to be on modern embedded SOCs and SOMs now.
A physical TPM with their overall high-quality software support would be awesome.
I've spent far too much time messing around trying to get TPMs working over SPI or I2C to meet security requirements with 4Bs and 5s over the years.
You do know those are trivially bypassed with a signal processor, right? If physical access is outside your threat model, that's OK, but it makes (for example) the forced Win11 upgrade for DRM^H^H^H boot integrity enforcement seem ridiculous.
https://pulsesecurity.co.nz/articles/TPM-sniffing
Yeah, fair enough. "Compliance" is probably the phrasing I should've used, rather than "security".
I've been curious for a while about the overall taxonomy of security, especially for embedded platforms. It seems like the only hope is defense in depth, given the power glitching attacks and the like that you can find demonstrated.
Specific to the Raspberry Pi, I believe I even saw a thread at some point where one of their firmware engineers was making the case that secure boot on the Pi 5 was equivalent to a TPM in almost any reasonable threat model, since, in either case, you were out of luck if an attacker had physical access and was willing to put in enough effort.
Normal secure boot does not use the TPM. Secure boot is the proactive process of ensuring only allowed code loads and executes.
The TPM is used for measured boot, the post process to understand what actually was booted and if the right set of things were booted then to allow unlocking of specific items like keys.
Both are important but they are not the same thing.
The article you link to explains how to defeat the sniffing with TPM 2.0. But also, there’s no reason a physical TPM has to be a separate IC package.
It seems like esp32 boards have taken over the pi's original market as developing for them has gotten easier while pis have gotten more expensive
These two products don't even have the same use case.
I was thinking about the RPi 6 yesterday whilst realising I couldn't set up my RPi Zero 2W anymore - the OS has become burdensome - tied strictly to an imager, that gives me an allergic reaction. Yes - they did all this for the uninitiated - but for Raspberry Pi OS Lite - bring back this experience: dd the image, write ssh into the boot drive, SSH in - change password, fully set up in almost zero fuss or effort.
Then I actually couldn't set the thing up because of the mini HDMI connection - I have a mini to HDMI cable, but to use my portable screen with it I need mini HDMI to MINI HDMI. Don't get me started on micro HDMI - almost everyone of of those connectors I've bought slips off or breaks in the device. Every time I go to set up an RPi5 I end up having to order another one of those tiny connectors.
Full HDMI for all new devices please. Even if the second display can't be connected.
These days a 175 GBP N95 from a no-name Chinese OEM on Amazon, with 16 GB of RAM and a 500GB SATA SSD is way better value and performance - and importantly - zero fuss - standard setup.
> __These days__ a 175 GBP N95 from a no-name Chinese OEM on Amazon, with 16 GB of RAM and a 500GB SATA SSD is way better value and performance - and importantly - zero fuss - standard setup.
"These days" seems to be over for now. The prices went up there, and now the only affordable option is an old laptop's motherboards with DDR3 sticks. But passively cooled PCs are rare, so the Raspberry Pi still makes sense. As a bonus, it is easier to power and find a small UPS that will keep it running for hours.
They aren't actually shipping the new RP2350 silicon revision on Pico 2 boards yet. If you want the errata fixes, you've gotta source the chips and make your own boards.
I recently found out about the Radxa Dragon Q6A. A Qualcomm chip with faster CPUs, a good GPU, a DSP and AI accelerator, and a hardware video encoder seems very compelling. It even supports Windows if you want that for some reason.
And poor software support.
The thing I've always wanted in a Raspberry Pi is an SBC and microcontroller on one board.
For me, Pi6 shall focus on NPU to run edge models instead of getting into minipc space, e.g. 10TOPS built-in, that will be slightly better than Rockchip 3588 which was produced 4 years back at 6TOPS, and way more powerful than NXP's 2TOPS. 10TOPS is the sweet spot for edge AI as far as I can tell.
Designing a chip for one person is doomed to fail as a business.
More than one person is buying Rockchip powered systems. A new high-end RPi has to at least match the hardware features of a RK3588 or RK3688.
The raspberry pi CEO said that some of the boards will be delayed because of RAM prices.
https://www.notebookcheck.net/Raspberry-Pi-discusses-Zero-3-...
As someone who has just had to migrate a project off of radxa due to ongoing supply issues... good luck getting your hands on more than one or two samples of those :/
The only way I'll buy another raspberry pi is if they come with a power supply that's guaranteed to work with them. I got tired of the random reboots in the night and replaced my media center/NAS with an old Nuc.
Official power supply is sold separately. Pis work fine with unofficial USB-PD power supplies but will limit maximum current for USB devices. I have two Pi 5's and one 4 and they all are perfectly stable when powered with an old Lenovo laptop charger.
It's still a huge pain point. I keep a magic "correctly sized, but provides clean power" USB charger with my Pi in a drawer, because I know that, if I lose that charger, I'm buying a replacement. It's unclear to me why they're so borderline. Maybe USB-C would help?
I don't know much about older models but on 4 and 5 they chose to use a weird power standard to reduce cost. Just buy the official charger. They're not expensive.