Hacker News new | comments | ask | show | jobs | submit login
Ask HN: How will it go for Raspberry Pi in the coming years?
46 points by DrNuke 8 months ago | hide | past | web | favorite | 75 comments
In the consumer and the small business markets, we are having computing power being more and more shifted from desktops or laptops to tablets and smartphones, we are even seeing full computing watches these days. In your opinion, what about DIY, IoT, small robotics, etc. and the generalist barebone computers they use, like Arduino, Raspberry, etc.? Will they become more and more powerful for a new race to bigger, maybe embedded systems or will they stay small, open, very cheap and marginally energy-hungry?

These things are a god-send for the auto industry in my opinion...specifically professional drivers.

as a lead automotive mechanic for a semi truck shop, I bought about 5 pi's for our techs. We use them for reading OBD diagnostics and doing resets or long-term vehicle observation for certain problems. Theyre cheaper than most of the commercial rent-seeking computer tools we use and if a tech accidentally crushes one, its cheaper than dirt to replace.

We even have a customer who keeps long term backups of log books and GPS data on a pi running off her cigar lighter, as well as her favorite audio books. Trucking companies dont care if you whine the most about your miles, but send them a ZIP file of 6 years of logs and the CEO himself will write you a pay raise. best of all, she tucks the pi in her shirt when she switches trucks. Try doing that with a qualcomm or a laptop!

Just curious what software are you using for the ODB diagnostics? I'm currently using a cheap ELM327 reader hooked up to an Android tablet with Torque. I have wanted to switch to using a Pi with software I could modify myself, but I guess my previous search was a little short.

It's a dream, but I would love to see some open-source software that would be equivalent to the diagnostic tablets from some of the tool trucks.

For when you need it only for OBD-II, this might be a better/cheaper option: https://www.ebay.com/itm/OEM-Bluetooth-Mini-ELM327-V2-1-OBD-... (it connects to your phone via bluetooth, it needs no external power and there's a bunch of apps on the Play store that you can use).

A word of warning for that customer using it for backups — pi’s have a habit of corrupting sd cards once in a while, or the sd cards just go bad. Make sure to have backups of the backups!

This is a perennial problem for the handful of Pi's I have deployed around the house (homebrew DAC, weather station, boiler manager (wip), workbench terminal). They're all 'purpose built' and sit almost entirely idle while fulfilling a single, usually non-interactive task. And after 18 months or so, they all start failing the same way:

* Random userspace crashes (most recent one was a reproduceable crash in the bowels of the python runtime) happen with greater frequency

* (Usually in the process of debugging the above) strange filesystem artifacts show up (files disappearing after reboot, etc)

* Booting ends up taking 5+ minutes, if it succeeds at all.

The solution to the above ends up being 'rebuild the box on a new SD card'. This is usually on 'brand name' (Kingston, etc) SD cards bought through proper retail channels, so while they're not high end cards, they're at least not (likely) counterfeit either. I still don't have enough data to know if simply dd'ing the old card over is sufficient, or if a full ground-up rebuild is sufficient.

Would love advice on this, especially around brands of cards with known good lifetimes in similar environments.

This was recommended in a previous HN post. https://www.digikey.com.au/product-detail/en/atp-electronics... I would consider using a SBC with onboard eMMC/NAND flash, such as the beaglebone or orange pi.

What kind of software are you using for reading OBD?

This looks like the kinda setup:


Also a search for "reading OBD diagnostics with raspberry pi" brings back some promising results.

Not OP, but I found this Python library for reading data from a car's OBD port that people seem to be having success with, including on the Pi: https://python-obd.readthedocs.io/en/latest/

how do you deal with power surges during start up and shut down. I imagine the inductive loads/peak inrush would be massive?

Most of the comments so far seem to have missed the point of the RPi specifically, which is that it was always intended more as an educational tool than commercial solution. The idea of the RPi foundation is to promote experimenting and hacking by selling a board cheap enough that you don't care too much if you ruin it.

This was explicitly a response to the rise of smartphones(/smart-whatevers) which turned computing into magic by sealing off all the components and preventing any hands-on work - students were enrolling in to computer science degrees with nearly no understanding of how a computer is fundamentally put together.

There is no reason for RPi to change from their very successful mission - they will keep putting out the best hardware they can squeeze into their fixed price point.

But just because its primary purpose is education doesn't mean it will/would/should always be that. Its sort of like curb cuts, something that starts out trying to fix one problem but then its quickly seen how many other problems it helps fix or other capabilities to enable.

The original point of the RPi is meaningless if the majority of the use cases are something completely different.

For me, doing embedded consulting, the Pi is a cheap, simple way to get a touchscreen and/or network capability without jumping through the hoops of Arduino add-ons.

I don't think that's true when the Rpi is entirely under the control of a non-profit. They don't care how many you buy, you are only further funding their goal and so far that goal has trumped everything/anyone else. Which I think is great.

I'm more of a Beagle Bone person, so I hope you don't mind me answering the question from that standpoint.

The sadness I have about the bone is that it is a means to an end. Purchase bone, insert into host board. Voila, IoT. It gets you to an MVP really fast, but limits your design to that form factor. Migrating away is difficult.

The Arduino, at least, served as training wheels. The end goal was to design an atmega328 into your circuit board, not glue another pcb on top. (Just remembering, there is some progress here with the pocket beagle... Though it does raise supply chain concerns.)

Also unlike the bone, the Arduino opened up a pathway for many other architectures and families of microprocessor, lowering the barrier to entry for 32 bit processors and ARM, keeping the platform relevant.

The bone CPU is getting a little long in the tooth, and there doesn't seem to be a path forward to the next generation of processors in the same SWaP C space. There are others, but the cost or power consumption tripples. The bone is at a real sweet point for price and capabilities. I'd love to see a lower power generation with some fixes to annoying problems (like the RTC).

In their defense, there are two or three orders of magnitude more complexity to an ARM SoC than an 8bit micro. The bone community really is on point with making a stable platform with regular releases, especially given the complexity of supporting a SoC+peripherals with the kernel, device tree, U-Boot, Debian, etc.

You mentioned Pocket Beagle, but did you know the OSD335x[1] is readily available on Mouser and DigiKey? I certainly would have specified one instead of a BBB if it was available during our product development cycle.

[1] https://octavosystems.com/

From the interviews I've heard with the founder, rPi set a price point first, then developed the best hackable computer they could to match it. If they keep up with that (as they have so far) then expect something more powerful and possibly smaller at the same price. Although I think the rPi zero is probably at the lower end of how small you can go without making tinkering cumbersome with all those gpio pins and such. although, people also like ATTiny, so who knows :)

The PiZeroW is hits such an amazing triple point of price/functionality/size that although I love the Tinker Pro for coming close in size/price and the ease of Arduino programming for simple things the PiZeroW just gives you SO MUCH functionality.

Slightly off topic - The one-per-customer policy on the PiZero and PiZeroW seems very limiting to me for someone who would like to develop more than one pi-based hobbyist project (or multi-pi projects like a cheap kubernetes cluster to play around with that stack). Does anyone know if there are plans to drop this restriction anytime soon, or do hobbyists just need to accept that to do multiple projects they need to buy the more expensive models? (Not that I can complain that $25-$35 is expensive!)

I doubt that they can manufacture enough Pi Zero's to match demand without purchase restrictions. Not only that, but you also get people buying them in bulk and reselling them.

I don't think it's one per customer. More like one per transaction. I buy mine from Microcenter and probably have 10 by now.

You can buy as many pizerow as you like if you buy a bundle that includes the official case. That raises the price to around 18 usd.

Their competitor, Libre Computer, has mused that given their investment in the graphics stack, they might be headed in the GPGPU / AI direction. Think Raspberry Pis that you can cluster together for deep learning, or individual boards that kids can tinker with to learn about AI. The blog post also predicted the 3B+ fairly exactly:


Money quote:

> We expect the Raspberry Pi 4 to be based on a new Broadcom SoC based on Cortex-A55 built on < 28nm fabrication process. We can call this BCM2839 and it will feature the much alluded to VC5 GPU and a new form factor. Since that hardware is more than two years away, it is difficult to predict the direction of the industry and thus the exact specifications. We expect such a Raspberry Pi 4 board to utilize dual 32-bit LPDDR3 or LPDDR4 to cope with the bandwidth requirements of OpenCL and neural network acceleration. The exact release date for such a specimen can vary dramatically depending on Broadcom’s tapeout schedule. It should be relatively soon if the Raspberry Pi Foundation expects to release hardware in 2019.

Please avoid code snippets for blockquotes. It's very annoying to read this with the `overflow-x: scroll` that HN's CSS adds (particularly on mobile). Just put a ">" in front of the paragraph, like so:

> We expect the Raspberry Pi 4 to be based on a new Broadcom SoC based on Cortex-A55 built on < 28nm fabrication process. We can call this BCM2839 and it will feature the much alluded to VC5 GPU and a new form factor. Since that hardware is more than two years away, it is difficult to predict the direction of the industry and thus the exact specifications. We expect such a Raspberry Pi 4 board to utilize dual 32-bit LPDDR3 or LPDDR4 to cope with the bandwidth requirements of OpenCL and neural network acceleration. The exact release date for such a specimen can vary dramatically depending on Broadcom’s tapeout schedule. It should be relatively soon if the Raspberry Pi Foundation expects to release hardware in 2019.

I still struggle as to why this is still a thing on HN after all these years. Just link to a gist if you have code you want to share that is wrapping too long.

Ok, done.

Let me get this straight, they think the Raspberry Pi - a primarily low-cost device aimed at teaching kids the fundamentals of computers/computing - is going to ship something more expensive (it will be) to better support AI? Upton has repeatedly made it clear that keeping the price low is key and anything that threatens that needs to add substantial value for Foundation's teaching goals. I think it's fair to say that OpenCL and neural networks are well beyond that.

I'm curious, which part of their 3B+ prediction was accurate? It all seems totally wrong and looks like they forgot why the thing exists in the first place. The VC4 doesn't even have hardware support for H.265 decode (I assume they mean decode). You can do it in software with a much faster clocked ARM but that's not what they are saying and you won't get 4k. Why RPF would even care about 4k support I have no idea. I'm also not sure where they got the idea that power consumption is important either. 2GB of RAM is a fantasy considering the substantial cost brings little gain. Making these sort of predictions is so difficult but they really were way off.

I'd love to see more focus on embedded boards with decent GPUs. About a month ago, I was looking for a hobbyist embedded Linux board with Vulkan support. I couldn't find a single one. There's one in development, but it can't be purchased yet: https://www.96rocks.com

The difference between Arduino and Raspberry Pi is already one where the Arduino (at least the basic, most popular(?) model) is just a board supporting a small, highly energy efficient microcontroller that you program in C, while the Raspberry Pi (even the Zero) comparatively uses a ton of power and runs a full Linux distro. That difference is huge, and seems to already be in the two different directions you're postulating.

That said, Arduino (the company) has released more boards with faster, more powerful processors over the years, and I wonder if they'll ever give up the flagship Uno model. I sure hope not, since as bjt2n3904 mentioned, it's a great stepping stone for small, cheap, efficient microcontrollers. It's kind of like learning a programming language and realizing how that knowledge applies to most/all other programming languages, whereas so much of the RPi is abstracted behind Linux.

Raspberry Pi is for hacking. It's not really a commercial platform. You can technically use it as such, but it's more cost-effective to design your own boards at that point.

If you've got enough users. I argue that, for a (possibly initial) batch of less than 1000/2000 boards, a rpi (either the zerow, or the 3) are simply great.


I shipped an RPi-based project a few months ago that was a one-off. Could have been done with Arduino, but it was far easier (read: less expensive to develop) on a Pi because it needed a touchscreen UI. I'm unlikely to ever build more than one more unit like it.

I'm currently quoting a design I'd like to use the Pi on because it will need to be connected to WiFi or Ethernet and I doubt I will build more than 20 or so. No way in hell I could make back the board design cost at that quantity. Were it not for the networking, it wouldn't even need a CPU!

RPi and BeagleBone hit a sweet spot for a certain type & quantity of designs that nothing else (of that popularity) meets.

For wifi-connected IoT thing I really like the ESP8266 chip. You can get it in a small form-factor board such as the "Wemos Mini D1" for approximately €3.0.

Programmable via the arduino-studio, widely-supported, and very reliable. I've built a bunch of projects with them, and couldn't be happier.

Those aren't commercial products by colloquial definition. Sure you made money, but they're not up for sale with thousands of customers.

That's true for the Arduinos of the world. But you need very very high volume to make designing a PCB with a complex BGA like the Pi worth your while. The industry standard is to use a system on module, and there are plenty of those from a long list of manufacturers. The Pi foundation has one too, what they call compute module.

Meh. That is true if you are selling thousands or more of an item, but for small companies, custom boards can be a nightmare. I worked for a company once that put literally two full weeks of debugging into a problem caused by a flaky diode batch.

I'll raise you two full weeks of debugging and then the project got canceled because redesigning the board was needed and it was too expensive ;)

I'm doing quite a few of low volume (mostly bussiness/industrial helpers) arm linux projects on Pis now. Some of them stay on Pis, if rebooting the board every 2 hours is feasible, some move to custom boards after the prototyping is done. Still, it's very nice to have an off the shelf documented board to start stuff on.

What issues have you seen that require a 2-hour reboot?

Stability was one of my concerns with my last Pi project, so during development I basically never powered it off or rebooted (except to test power-up behavior) so I could see if I encountered any problems. No problems came up, but others' experiences are always good to learn from.

Was it a project in a climate controlled area, like your home? Mine were in a factory, and after a while the stuff that they were supposed to connect to over USB simply vanished. A reboot (of just the Pi, not the other device) fixed it. So we just scheduled a reboot in 2 hours when the Pi started up.

More or less: at home, but in a basement room with temperature swings. For the application it was fine: it was going to be used in a temp controlled environment anyway. My current project I'm thinking of using BeagleBone because I trust their temperature range specs.

Is it possible your Pi was suspending USB as part of power management? I've had that problem with some other devices.

In my stuff we never did a two hour reboot, but we did do a daily one. Sometimes you just run into memory leaks some of which aren't even your fault.

I don't know if they needed a two hour reboot, but the customer wanted to be sure. For sure they stopped seeing usb devices after less than 24 hours. It helped that they were just reading and transmitting some values every 10 minutes or so, the reboot didn't hinder that.

If the availability of the pi zero W is any indication I expect trouble ahead. Trying to buy the zero W in bulk through channels is practically impossible, and the vendors sell them only in kit bundles at many times MSRP at 10$.

Instead of the pi zero, check out the CHIP Pro (https://getchip.com/pages/chippro) ... a little more expensive, but available in quantity and designed to be pick & placed (or hand-soldered!), runs linux, and does most of the things a pi can do (sans video)

Stay away from CHIP, they appear to be bankrupt and/or fraudulent, they took my money and never delivered.

Fwiw I don't think it's even possible to order anything from them at the moment. I can't figure out how to (but also have heard other stories of late or non-delivery). Seems like a cool product until you realise it doesn't exist.

They withdrew sales of the original C.H.I.P. months and months ago, without really explaining why. It's quite puzzling, because I ordered four in November 2015 and got them as promised about six months later. I guess maybe they made an initial batch, fulfilled that, then just ran out of steam and haven't made any more? It's disappointing to hear other people haven't received theirs.

Huh, weird. I ordered some CHIP Pros and dev boards in February last year, and got them in ~2 weeks. I haven't ordered from them since, so maybe they went defunct or something. I dunno. That's too bad. :(

I ordered 5 CHIP in Feb 2017 and never got anything. Estimated shipping was Q1 2017.

I sent several mails and got no reply.

The email address <ahoyahoy@nextthing.co> is now bouncing mails: "Recipient address rejected: User unknown".

I think the future is only going to get brighter for Raspberry Pi because its niche is as a hacker's machine. It's so inexpensive and versatile for putting together proofs of concept and I think that's where it's real value lies. If you can show a solid working prototype of something on a Pi then the assumption is that a dedicated hardware solution will run even faster... assuming you don't just make a custom case and mounting set for your ad hoc solution and call it good. What I think would kill Raspberry Pi is trying to make it more advanced and powerful than it needs to be, trying to make it a full powered computer rather than a hacker's delight.

Hopefully it will become more open and less reliant on binary blobs.


It really doesn't matter where else they "go" until they fix that right there.

(Why? Because on my workbench right now is the pi, instead of a pi; and that makes all the difference.)

That's exactly my problem as well. I wanted to hack a Pi Zero W into a presentation machine (for giving talks) and I didn't want to ruin my only Pi Zero W on account of how expensive it would be to replace (with shipping etc, it comes out to 30 euros or so).

Only for bare pizerow. Buy it along with official case, you can order as many as you like. Price raises to 18 usd per unit.

The price was the selling point. By the time you get to $20-ish, there are better choices for an SBC depending on the task you need it for. (Unless you are new to the scene and need the RasPi community to get going, which admittedly is excellent)

It looks like "$10 wifi computer" was more of an add (loss-leader) than a real product. There still seems to be a gap between ESP32/8266 wifi enabled micros and $20 linux SBC's with built-in wifi that needs filling.

Yeah, I wish they would just charge me the real price of the Pi, and let me buy as many as I want. No need to buy a case with it.

I've got 3 pi's in my desk drawer not in use. There's multiple outlet's that sell more than 1 per customer.

If you could tell me where I can get z-dubs 10 at a time for the advertised $10 price with combined shipping, I'd be eternally grateful.

Do they still enforce that? When I was working with Pi's a few years back I never had any trouble getting 5 or 6 at a time.

I had trouble getting more than one zero at a time when it was new. Maybe it's still the same in some places.

I think I've only ever bought Pi's by the handful...

Only for bare pizerow. Buy it along with official case, you can order as many as you like. Price raises to 18 usd per unit.

I'm not sure I understand the question. Of course they'll become more powerful--that's what happens. They'll stay focused on what they're good at.

The specs are always going to get better. For me it's all about physical size for the Pi. Given say 10 years... what would you imagine the smallest form factor you can get the Pi functionality into for the same cost.

I would bet we'll see a version with an integrated SIM card, hopping on the IoT-anywhere train. Twilio was a big first mover with their Twilio mobile service, and I can see people deploying rPi's and the like to more out-of-home applications.

Orange Pi have one, the Orange Pi 2G-IOT. It's really cheap, might be worth checking out. Supposedly runs Android or Linux.

The rPI mostly failed in its original mission as an educational tool. In practice, it proved too complex and fragile to use as a microcontroller but not powerful enough to substitute for a desktop PC. A lot of important lessons were learned and put into practice in the BBC Micro:Bit project. For teaching basic programming and electronics, a dinky little Cortex M0+ with a few LEDs and IO pins makes far more sense than a repurposed mobile phone chipset.


rPI has spawned a whole ecosystem of hacker-friendly hardware. Like Arduino, it generated demand for a particular form-factor - a lot of people bought rPIs simply because they were cool and cheap, then figured out how they could be useful.

The rPI has clearly moved towards being more powerful and power-hungry, with the latest model drawing about twice the power under load than the original. Documentation and kernel support remain a sticking point, especially for the off-brand PIalikes with Chinese chipsets.

The development of these boards is really dictated by the rest of the market. Nobody is spinning silicon specifically for dev boards, so they'll continue to use whatever cheap chips are on the market. A lot will depend on how well the foundries cope with <14nm process nodes.

They are very powerful right now, how much powerful do you wish for them? What I like about them is CHEAP, lower energy use. So long as they maintain that, then it's all good. I bought 6 pi zeros with wifi, $5 each.

I wish I could make use of those small computers better, but so far my main use case is for xbmc for my dumb TV, and I'm loving it.

Hopefully USB3 will come to the RPi... It would make it more useful as a file server.

I use a BananaPi for this (comes with a SATA bus).

And get the USB and Ethernet onto separate buses

Maybe just add USB-C with PCI-Express lanes in there. This would allow Ethernet or anything else to be plugged into that should it be needed.

I wanted this as well; from what I read from the forums, the chip does not support it so it'd need to either switch the chip or add another one, going way beyond the price point. Hopefully the integration for new chips becomes better and there's more demand so that it becomes a reality.

This is the one feature I'm hoping to appear.

Applications are open for YC Summer 2019

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact