The folks at Raspberry Pi are riding on a bit of a wave at the moment, with the launch of the Pi 5 with its PCIe and RP1 peripheral chip, the huge success of the RP2040 microcontroller, and the supply chain issues that dogged the Pi 4 and Compute Module 4 during and after the pandemic finally working themselves out. But as always there are plenty of would-be competitors snapping at their heels, so [Jeff Geerling] has posed the question of what it takes to make a Raspberry Pi killer. He’s in a good position to do this, as he’s amassed an impressive collection of every competing Compute Module board.
It’s a well-observed analysis of the world of small Linux SBCs, on hardware, software, community, and price, and we find ourselves pretty much in agreement with it. The Pi hardware has quirks and is rarely the best on paper when compared to the competition, but they win hands-down on distribution support and community. In a sense what you really buy when you get a PI is this, because Raspberry Pi OS will run on it for the reasonable future. Rival makers would do well to read his piece, because we sense that if one of them tried to give the Pi a run for its money away from the hardware it would make for a much better SBC ecosystem. Take a look at his Compute Module comparison below the break.
We recently took a look at the strategic importance of the Pi 5 and in particular the RP1.
so the upshot is… It takes everything that raspberry pi has done in terms of development, hardware peripherals, drivers, support, pricing and community since its release to create a pi ‘killer’, which, despite their best efforts (or lack of) no other manufacturer has managed to do, to date.
Sounds about right.
If even one of these manufacturers partnered with a distro to get full board support into a generic arm image, they would be able to provide some serious competition. Lots of people have experience with debian, for example, but I don’t think I could get regular debian from the debian website running on any of these boards.
It’s always some fork of debian/ubuntu/whatever with package repos and kernels from the SBC manufacturer, which maybe stay up to date for a year and then rapidly fall out of date. I don’t want the SBC manufacturer’s kernel, I want one from a distro.
I’ve taken a serious look at “non-pi boards with long term software support” and the best I’ve been able to come up with is a beaglebone from ebay running mainline NetBSD.
NetBSD runs on just about everything. That’s not surprising. :P
Armbian or DietPi?
Its most likely Armbian in any case. Dietpi is just unsupported Armbian packed with bloatware.
Armbian guys deserve a medal for their work.
Not quite the same thing, however Pine64 works heavily with Manjaro, as well as most mainstream distros (Debian, Ubuntu, Arch, Fedora), and the folks over with Banana Pi have been working heavily with OpenWrt (as well as a few mainstream distros such as Debian), and have been affecting Mediatek’s outlook upon open source development, to the point that Mediatek is fast on track to becoming quite FOSS-friendly, regarding the hurdles needed to develop for their SoCs.
Huh. Armbian is Debian and is supported on piles of boards. https://www.armbian.com/download/?tx_maker=friendlyelec
Latest Rasberian installs on the earliest RPi model 1A, how.many profit obsessed SBC manufacturers could surpass that achievement??
If you think the raspi foundation isn’t profit obsessed, you’ve got another thing coming. They’ve established an absolutely STUPENDOUS profit-protecting moat for themselves.
What you describe is a non-problem. Debian ARM port works very fine on many many of those SBCs.
The issue is the bootloader and kernel. Proprietary blobs for peripherials (nand, video processor etc) are killers too.
The pi got the kernel in upstream trees. The gpu is ugly, but foss. The bootloader is a blob, but you could argue, because its in an spi flash, it doesn’t count/matter.
Long as the bootloader can boot a standard kernel, most people let it slide.
Armbiam
It is very unlikely anyone else will be able to combine the chip maker support, free personnel and facilities, guaranteed sales into education, and tax breaks, and the instant user base the RPi had. And what kind of breakthrough would it take? And even then it won’t work because the RPi user base will just wait until the Pi people can copy whatever you did. Even a poor copy will thrill the enthusiasts.
Plus I think there is more alertness now to the kind of stealth dumping into overseas markets by the foundation masking as an educational resource.
Can you elaborate on this, or define more precisely as i cant seem to find much about this on the net.
Bizarrely, you still get media articles from so called tech websites with titles like “raspberry Pi Alternative”, which none of them are.
They should take a step back with their thermal designs. More power is not necessary for most use cases. The Pi 3 with passive cooling is still the most sensible for higher ambient temperatures.
Not really necessary, just reduce voltage and processor speed and passively cool the new Pi 4/5. It’s good to have a thermal solution designed by the company. They recognize some are using it for desktop or gaming/emulation, and I’d rather have a properly designed solution (did anybody else see the tests of Intel LGA775-LGA1368 ‘flower’ coolers? It took a surprising amount of money for an aftermarket cooler to beat them thermally, as simple as they were. Especially the 2 or 3″ tall ones with a copper core). The cooling solution is optional anyway.
How does one go about setting up a 4 or 5 to run needing only passive cooling?
Passive cooling with a relatively simple application is my use case. With the 3, the setup for my use case consists of just plugging in the power supply and using it.
https://forums.raspberrypi.com/viewtopic.php?t=250167&start=25
Doesn’t seem difficult, you already set up the software you need to use, whats a couple boot cfg settings?
As long as a Pi 3 is available and does what you want keep using it. It appears that a mild underclock/undervolt and a small heatsink stuck on it is plenty for a Pi4 or Pi5. It is more processor for those who need more processor. Its like me complaining the new i9 or Ryzen 9 run hot when I should be on an i3 or Ryzen 3 for my workloads.
Thanks.
I’ve had a pi4 running for several years with just stick-on heat sinks. It is exposed to room-temperature air with free airflow (covection). Right now it’s at 58 degrees, and never gets above low to mid 70s.
You don’t have to put any cooling on a Pi4 (and probably 5) at all, they can run without even a heatsink stable but thermal throttled – From Mr Gearlings tests IIRC the Pi 5 while thermal throttled had similar performance to a Pi4 that wasn’t, which means it would really be a great candidate for passive cooling in many projects.
From what I recall of playing with Pi3 those can run passive but are actually really really hot chips for the performance, so you probably would find a Pi4 gives you a noticeable jump in performance or substantial drop in CPU temperatures (depending on what your load is).
RPi 4/5 with capped frequency so it emit as much heat at load as RPi3 is still going to have better performance. There is no f’ing step back, only forward.
Like some of todays EU/Japan cars. The fact that you can have 150-200kW and torque of a tractor at your disposition doesn’t mean that at sane power levels they don’t have much better fuel consumption than cars from 30years ago.
I have entirely too many Pi 4s here – media centers running Kodi, my kids have a Pi 4 and a Pi 400, Kali running on an 8GB Pi 4, etc.
None of them use active cooling.
Initially, the 4 needed it – but after an early firmware update, that need went out of the window.
I won’t speak to the Pi 5’s need for active cooling – not only do I not have one yet, but it’s still early days for it. Perhaps it’ll need it, perhaps in the long term it won’t.
I do wonder what the Pi foundation was thinking when they got rid of ALL hw decode/encode except for hvec decoding.
I think it is to save money. The early models of the Pi needed them to decode video aka watch YouTube. Now, I think, only h.265 is outside of what a core on the Pi can do in software.
At the cost of huge power drain doing it in sw significantly limiting mobile applications.
Still waiting for a SBC with useful hibernation, would do wonders for deep sleep battery life for solar-powered devices that don’t need to be on 24/7. Went with ESP32 instead but it certainly lacks a powerful processor.
Use the Esp to wake up and shut down the pi if extra processing power is needed.
“supply chain issues that dogged the Pi 4 and Compute Module 4 during and after the pandemic”
so what was the excuse before the pandemic? Everone know you have never been able to get more than a couple before stock runs out.
Namespace collision error 0xdeadbeef
Well, damn, I guess with a name like ‘Mike’, I should have figured the chances of running into someone else posting under the same name were going to be fairly high.
I’m gonna change my name to this.
I just hope nobody has the same idea :(
Well, crap, there goes my backup Identity.
Tell me about it.
B^)
0xdeadbeef is the name of a successful artist whom uses plain C code to make entirely algorithmic/generative audiovisual works of electronic music.
Really not the case for a Pi product, except for the month or two right after launch most of the time – they do always seem to underestimate how deep the at launch stockpile needs to be. But once that initial rush of orders dies down Pi tend to just be in stock at least baring a global silicon disruption event.
Any board with mainline distro and device tree support AND with sensible power management would be Raspberry Pi killer for me
I don’t see a problem with using just PI OS. Linux … is, well, Linux. I run PI OS on my RPIs, Ubuntu and Fedora on desktops/servers and … Yep … Linux is Linux.
Have an RPI-5 now and it is a nice stable platform. Hooked up a USB 2TB SSD as the OS drive, and a USB 3.0 4TB HDD and no reboots or problems. Did a 4 hour backup to the 4TB drive. No problem. USB power problem appears to have been solved. The RPI-4 would have croaked on this simple setup. So far really happy with the RPI-5. Not sure what I’ll use it for yet…. as the Pico and Pico-W has taken over most of the GPIO projects. Maybe time to attach a camera and try out some CV stuff.
What it takes is, is to come up with a non-ripoff business, that charges reasonable prices for its gear and designs innovative productive educational products as well. This encourages the effort from a huge number of essentially volunteers because the business has ideals above the pure profit motive. It appeals to the notion of “good for human kind” notion, not a listing on the stock exchange.
All the rest who hope to rip off great young thinkers and turn them to a personal profit can go and wither on the stump of stupidity.
It is not just a technical comparison of so many nefa-fxxks per second, but a philosophy about what the technology and who gets to use it.
I can’t stand youtuber. I respect Jeff but he shat on his talent/hardwork by becoming this youtube tv host. :/
Rather strange take considering his text based web resources are still worked on – you don’t need to do the youtubing stuff at all if you don’t want to and you can still get that detailed documentation – actually way more complete than the videos in most cases and being updated after the video comes out – its not just the transcript, and you won’t even get that from many youtubers!
So if you don’t like that side just don’t watch it, not like he is holding a gun to your head…
Indeed, I usually spend about 8x as much time now on any given project, and *thoroughly* document my testing and work not only on GitHub, but in a more detailed writeup on the blog (at least for any substantial work).
I hate video too (I subscribe to over 300 blogs via RSS, and prefer text for information), but if I didn’t have the YouTube channel I would not be able to work on the projects I work on. Despite attempts at VCs dumping money into Medium and Substack, text publication as an individual isn’t a viable career option long-term, for like 99% of us.
Also useful resource to have the more pop-sci accessible/inspirational video for the folks that didn’t even know what questions to ask or that some of the stuff you are demonstrating is possible (or in the case of GPU on Pi almost possible) – text is in many ways superior, but doesn’t help you at all if you don’t know what search terms to use, which book to reference to find out those search terms etc – good technical documents tend to be rather more technical in language.
There’s always that One Guy….
Thanks for your Youtube and “print” info Jeff
For me, a MCU/MPU/SOC with proper peripherals would quickly seal the deal. Like come on, two I2C, a single SPI with two chip enable, and a single uart? At this point I see the RPi-s as glorified smart TV PCs. Sure, they pack computaion power, they have nice GPU, ethernet, wifi, HDMI, PCI, but it’s more PC than an embedded computer.
Just have a look at ST Micro’s STM32MP15x, TI’s OMAP-L13x, or for the fancier stuff, the Intel Aria 10 SX or AMD ZYNQ. I’ve listed only these chips because these are the only ones that I tested, and each of these is more versatile than any of the RPIs ever.
Of course, if one only knows the software part then even the RPi is more than enough.
I think the Pi generally is in the sweetspot myself as ultimately if you need lots of low level IO its generally not a real computer, its a microprocessor board you want – may be tethered to a general purpose CPU burdened by a full operating system for compute power if required though.
More compute power (Like a Pi) heavily implies running a full operating system which adds all that variability in response time etc. So functionally it will often be less performant as a whole system than having that high performance system (like the Pi) consuming data and issuing instructions to something like a Pico at fast as its task scheduling will allow while the Pico cycles predictably keeping the data fed to the powerhouse up to date and the outputs doing whatever they are supposed to be. And if performance isn’t that important you can pretty trivially add IO expanders as required.
I suspect however once the Pi5 is a bit more understood that 2040 like chip that handles the GPIO onboard will give you what you want anyway though – but I really don’t know what its limits will be (doubt anybody including the Pi folks themselves really know either).
Put a raspberry pie in front of me, I will “kill it” in short order!
B^)
Most pies don’t fight back, unless they’ve been left out too long.
It’s not the pie, but the pie’s steam that you have to watch out for. Just ask Peter Griffin.
Sometimes, I like to imagine a world where the low power SOC world adopts enough of the ethos of the PC world to where it would be possible to select a board that best fits the application I have in mind, select a distro from someone else and have everything just work together. But it seems like this will have to remain just a dream.
Good documentation, mainstream sensible OS with no weird wrinkles, and guarantees about lifespan of boards, support, and updates.
Outright performance or price doesn’t matter so much, I’m not going to spend 100k designing it into a product if I don’t think the thing will still be manufactured / supported in 18 months time.
I clicked the link in Google discovery expecting the building for some destructive device with a usb connector and a supercapacitor for frying raspberries LOL.
@Jenny List said: “The folks at Raspberry Pi are riding on a bit of a wave at the moment…”
Maybe things are looking up for you Jenny, but it’s still thin-pickins’ here in the U.S. for anything Raspberry Pi except for the RP2040. PiShop in the US has some PiZero 2 Ws at a high-ish price, but not too bad – very low stock though.
I think Raspberry has lost their way. They’ve left their community hanging so long by not focusing on manufacturing their older designs, that it’s natural to want to find alternatives.
Two things have made them successful, but they seem to ignore those and are only pushing, more expensive, complex and, albeit it, powerful systems.
First, it was cheap with more power than arduinos. For some, not an issue, but for the average “maker” who buys as a hobby, it’s foremost. I’m afraid to use their latest offerings for hobby projects (except RP2040) since I can’t justify a >$100 USD device to build robot to play around with.
Secondly, the community of software developers (and thanks to you all) has made it easy for those that don’t have expertise in writing code to take their work and roll it into a quick project without having to learn the nuances of software development. (Same reasons apply to arduino). It’s these casual makers who account for most of the R Pi volume and they are looking for alternatives.
I’m a hardware engineer and I count myself as being in the casual maker category when I’m doing stuff at home. I’ve found alternatives that work that are much less expensive but don’t have developer support yet. If I need something powerful enough to cost me more, I’ll get a PC.
They keep making and keep supporting the old ones for a very very long time (the only discontinued Pi’s that really count are the very first Pi 1’s with tiny RAM quantity – as putting 1G on a Pi4 to help ease the shortage and then going back to 2G doesn’t count IMO), and then put the brains of a Pi into the zero lineup for extra low cost to performance ratio. The highest end price has gone up along with skyrocketing performance, but so too have their cheapest offerings got cheaper and improved.
That hardly seems like loosing their way, yes the price to performance of the still available Pi 1 B is now not great compared to the still available Pi 2 and 3’s but its still possible to get one if you want to. Heck one of my usual suppliers for electronics will even sell you 150 of them at once for a better per unit price!
Availably has been poor (though mostly only of the more performant Pi’s it seems to me) but is recovering, and that has been true of everything around the pandemic its not a uniquely Pi problem.
no blob, open firmweare/bios, power on board (lipo, AA or 18650)
nothing more