- Conveniently naming their chip "5" just to one up the pi 4. They're actually were four raspberry pi's; I'm not sure the orange can say the same thing
- There's a notice in there that 'orange pi' is registered trademark of their company but clearly stealing the pi thing from raspberry
- Finally they say that the pi is aging. I think they're missing the point of the raspberry pi.
This just kind of feels like a cash grab. The pi came out with the noble purpose of educating children in the UK with a computer that was so simple and cheap all you needed to do was grab a keyboard and a TV and you could learn to program.
They've accomplished so much more than what they started out to do.
It's just insulting to see something like this. If I wanted more power I could just get a NUC.
That said I have wanted to look into something like the pi but in a laptop form factor.
> - Conveniently naming their chip "5" just to one up the pi 4. They're actually were four raspberry pi's; I'm not sure the orange can say the same thing
It'd have taken you two seconds to check that. No, it's not the first Orange Pi product.
> The pi came out with the noble purpose of educating children in the UK with a computer that was so simple and cheap all you needed to do was grab a keyboard and a TV and you could learn to program.
That's great and all but as a hobbyist who can't buy the Raspberry Pi in part because of production allotments to commercial integrations, it rings a little hollow.
There's quite a bit of activity getting the rk3588 working in the mainline kernel. People having current mainline kernel+patches working today, and getting those patches mainstreamed has been making steady progress.
the orange zero
orange pi plus 2e
orange pi 3 lts
orange pi 4 lts
orange pi 5
In any case the rk3588 is available with 4x the PCIe, 2x the ram, 2-8x the performance, and ethernet and storage are over PCIe (not USB). Seems like a decent generational upgrade.
I'm so glad to see something accessible from ARM that is vaguely performant & somewhat affordable. I bought one, pre-ordered it! Still untouched, still not running Debian, but delighted to see, excited to have this. Phones have sick incredible ARM cores but general availability of ARM chips is crazy shit, A72 (2015) or A73 processors have been the best one could hope for on a reasonable priced SBC, and the RK3588 here's A76 (2018) is vastly more modern, albeit still half a decade old.
Still. This is not a fair comparison, but the used/refurb/old-stock PC market is just so ridiculously more capable. A HP EliteDesk G2 has i5-6500T (2015), 8GB ram, and 250GB SATA drive is $140 on amazon. It's a higher powered machine, 9W versus ~3W idle (shame on this review for having no power consumption figures at all), and much higher full-open power (I think my last kernel compile was <35W which is the TDP of the chip alone). And the OrangePi is significantly smaller form factor. There's cases where the OrangePi is a better fit. But for most general use, the x86 market has far far better value, does so much more, has so much more bandwidth (often 8x USB3.0 hosts, plus significant internal pcie).
There's periodic articles on repurposing thin-client computer hardware. But again there's just such a mismatched value proposition. Not the same shakeout: they have hot, old, low-end chips, but sometimes very cheap on secondary markets. But there's typically much better mini-PCs with vastly more performance for not that much more, that will idle at least as well if not greatly better. https://news.ycombinator.com/item?id=33791069
It's just so weird how few ARM chips are available. There had been a little more variety in who one could buy an SBC from. But now... Rockchip is the main company making purchaseable chips. Allwinner had some pretty fine offerings but has gotten sparse. MediaTek is doing excellent things in many spaces, but has almost no presence in accessible SBCs. There's tons of other chips but most of the market is invisible, is utterly uninterested in selling SBC type boards. There's a parallel x86 market, but Ryzen embedded systems tend to start at $400 for very sparse models. It's frustrating how deliberately unserved the low end is. That commodity x86 hardware dominates so savagely, by so many country-miles keeps feeling unfortunate. That ARM has such small niches it plays competitively in, and everywhere else it's broadly unavailable & unaffordable is a weird market dynamic, while these normally very expensive x86 cpus become available & affordable systems is such a strong & visible counter-trend.
I was actually thinking of this a few years ago. My workplace was getting rid of a bunch of old Win7 thin clients, they had VIA x86 processors and 4GB of ram. Never acted on it though.
I wouldn't bother. You can get things like a used Lenovo M70q which gives you the tiny form factor, upgradable memory, NVMe storage, 10G Usb-C and with a 10th, or 11th if you are patient enough, Gen i5 for £200.
That will blow all these SBCs and thin clients out of the water.
If you don't need 10th gen Intel, and are okay with 7th, you can drop below £100.
It will be really interesting at this low end once they have to start competing with used Framework boards.
One twist is that phone/tablet SoCs use different I/O protocols than PCs. For example, phones use MIPI-DSI for display and UFS for storage but computers use HDMI/DisplayPort and SATA/NVMe. So you can't take e.g. an awesome Dimensity SoC and put it on a usable SBC; the ports would be all wrong.
There are a few SoCs being targeted at Chromebooks that have PC I/O, including the RK3588 used in the Orange Pi, which makes them also suitable for SBCs.
> One twist is that phone/tablet SoCs use different I/O protocols than PCs. For example, phones use MIPI-DSI for display and UFS for storage but computers use HDMI/DisplayPort and SATA/NVMe.
Going from MIPI DSI to DisplayPort is as simple as adding a common bridge chip and updating the device tree.
Many of the new SoCs have PCIe lanes, so NVMe storage is easy, too. Some of the RK3388 boards even have a nice M.2 slot for exactly this purpose.
The interfaces aren’t a problem, though they are more limited than a full PC chipset.
I not so secretly wonder what a 2027 looks like when SBC all have to have USB4, when USB4 hub chips (actually, routers) are semi-commodity parts and not uber-botique Intel crazy-pants expensive units, and when Thunderbolt/PCIe is a logical royalty-free step away.
We're only just seeing Thunderbolt show up in mobile, via the IPad Pro. The percolation down of the universal, good, flexible, kick-ass connectivity standards is semi-inevitable. Right now it's a ton of add-on extra chips, but that will go away, will get integrated. These discrepancies of PC chips having some things and mobile parts having totally different specs feels like it can't last forever, at least in every category. There will be some uptake of common capabilities.
At the risk of a duplicative "This" comment thank you for a very well thought out, holistic comment.
It's strange. One of the many things I "harp" on here and elsewhere are all of the "Do $THIS on a Raspberry Pi" stories and articles around. What "bothers" me the most are people doing simple GPIO with a Raspberry Pi 4 when an ESP32 (or equivalent) would suffice. Or people trying to stretch a Raspberry Pi to do something an old, used, $100 x86 something would do significantly better in it's sleep. I understand the Raspberry Pi is a common denominator with a rich ecosystem but it still irks me in an irrational way. Don't even get me started on the people trying to build a NAS with a Raspberry Pi.. It's just wrong.
Getting on to to your comment - I have an entire graveyard of ARM SBCs (and more) that are incredibly poorly supported either by the vendor, Armbian, mainline, etc. I'm over getting excited by the latest "More performant, newer, better, faster" ARM SBC of the moment - only to burn incredible amounts of time getting it to do what a Pi or x86 something would do at a cost/performance/power trade-off.
To talk the graveyard, that's an interesting situation. I went heavy in on the Marvell Kirkwood chips (ARMv5), which hearkened to a very hackable nice early PogoPlug device that lead as a very hackable Linux system, that paved the way for hacking/repurposing many many other affordable consumer devices. Glorious circuit bending days!
I ended up with quite a collection. Small 2-drive NAS-ish DLink DNS-320's were great. IoMega IConnect had a solid (and user-replaceable) wifi chip & numberous USB2 ports, & was very afforadble on ebay. These kind of came latter-ish, but there were great other botique SBC or simple-device options pre-ceding GuruPlug: SheevaPlug, GuruPlug, OpenRD.
Many years have elapsed. But the upstreaming work on all these machines was very great. They required very little customization. Really hope I haven't messed up the attribution here, but I believe Marvell paid Free Electrons (named Bootlin at the time) to do a ton of the driver work. It's in kernel. Each device took tended to require only a little DeviceTree configuration to work great. UBoot got upstream support, & I'd replace the onboard uboot with a modern far more featureful uboot, greatly opening boot capabilities versus the minimal/older uboot devices shipped with (usb boot? yes. ethernet boot? yes.). Devices would slowly gain a couple modest not-super-important-to-me capabilities a year, or better tuning.
I've given away, sold, & (sadly) thrown out a bunch of these devices over the years. I still have two NAS running in friends' houses as off-site connected backup devices, humming along, but mostly ignored. Still, I fully expect, if I wanted to, I could completely freshen nearly any of these systems up. Brand new kernel, totally new Debian (actually many do run modern Debians, just on old kernels). My vibe has been very good, that old hardware doesn't die, doesn't get dropped, that the graveyard is merely of my interest to give a shit. Not all SBC are such: the Marvell Kirkwood was an early 2010's vibe of hope & promise, Linux uber-alles, where some good hackers got paid to Do The Right Thing, and I think this post is a testament to how obvious, how clear, how simple that path was. These devices are still fully supported. They still work great, kernels still support them, because Marvell worked with people who knew what the heck they were doing to get support upstreamed. These devices may well never ever die. They may be able to last forever. Because someone didn't just dial it in, because it wasn't just an internal team with some heavily forked kernel shipping shit: Marvell did the right thing, they upstreamed their work. The value of that, right now, seems like it may be ever-lasting.
The graveyard isn't necessary. Good vendors upstream their stuff & it will keep working.
It's also notable & exciting to see, as I've mentioned else-where on HN recently, Qualcomm chips have gotten much more upstreaming recently. The Chromebook folk really lead the charge with do-the-right-thing-duh are-you-stupid-of-course-we-will path. But after getting a bunch of slightly upscale tablet-grade Qualcomm chips supported upstream, we're starting to see much broader classes of Qualcomm chips getting supported, & seeing growing device-trees for a variety of phones. Turns out it kind of mostly is the same hardware, no surprise, with different cores, and the decades of shitty unsupportable heavily-forked-and-useless Qualcomm-SDK horse-crap they've shovelled at people is & always has been a farce & once the ball started rolling towards upstreaming it's just kept snowballing, mostly with little thanks to Qualcomm themselves, but just just just recently, the company seems to be a adopting a not so universally useless sad pathetic view & actually stepping up every now & then to actually do a little bit of the upstreaming themselves. For a company famously reputed to require lawyers to sign off on literally every ...
Regarding the graveyard - just looking around my office I see a variety of Khadas products that are essentially useless. I have a RockPi 4-something running Armbian that is one `apt-get full-upgrade` away from not booting (has happened before) that already has extremely temperamental video and NVMe (the selling point). A few Odroid boards that fall somewhere in between... Plus a few random generics that are long forgotten and not even worth mentioning.
Yes, ideally support hits mainline. However, in practice between Linux mainline, things like Mesa for video, etc it just doesn't happen and trying to sell/tell people otherwise is disingenuous at best.
I agree there is significant value in the comfort of Linux but WRT to ESP32 I would argue the Arduino/PlatformIO ecosystem around them is actually second to the Raspberry Pi in terms of community support. There are "ready to go plug these variables in" examples for nearly anything you want to do that the hardware is capable of floating around everywhere on the internet. You can buy "ESP32 Cams" with TF Lite object detection from Amazon[0].
Also, not to be argumentative but the Kirkwood SheevaPlug line (I was early on that too) is the ARM equivalent of a truly embedded Intel Core something that no one dare touches if it's still running (it's USB 2).
Somewhat strange thing is I've worked on "commercial" embedded projects with things like the iMX line and Yocto - and the support there is pretty incredible.
> Also, not to be argumentative but the Kirkwood SheevaPlug line (I was early on that too) is the ARM equivalent of a truly embedded Intel Core something that no one dare touches if it's still running (it's USB 2).
But unlike everything else you've complained about, it has good upstream support & is reliable & all it's peripherals are quite un-flakey. I have upgraded my kernels numerous times. I do run Debian/testing. I haven't in a bit, but they would still run, and better than before.
You absolutely are correct about most of your complaints. There's a huge world of Ambian & other SBC that are off-tree semi-supported worlds to try to make shitty vendor crap keep going for a bit. And it's flakey, unreliable, & unmaintainable. But you are dead wrong about Kirkwood. Upstreaming means something & it has kept a lot of decade old devices running & stable & happy & it's just obvious. The negative experiences you cite are real, but they're a owed to a variety of over-aggressive under-budget hardware builds and more-so to never ever having actual real support, to never having had drivers good enough to get upstreamed, to being a bunch of shitty off-tree vendor code that is just poorly done. The network of enthusiasts trying to hack access to these sytems, to make off-tree Ambian or OpenWRT support at any cost is admirable, but ultimately, the actual cost of doing the right thing, like Free Electron does, is tiny, especially versus the pain incurred from not doing the right thing. I think you're missing that distinction, by lumping these all together.
USB2 is also really not that bad for many remote-storage applications. It's also a bit comedic to, in one sentance, blast it as bad, then go on to talk about ESP32 systems & other microcontrollers. Which can almost never saturate even USB2, & almost never have any kind of USB3 what-so-ever. Sure they may run TensorFlow Lite. But they are markedly smaller systems.
Just company community support seems arbitrary. Maybe Arduino has the best support community on the planet. It's still a vastly less capable platform with incredibly less software available to help you build systems/solutions out of. And showing up to Arduino will be a unique bespoke different experience, whereas running a Linux system will have a common frame of reference & transferable experience into the rest of computing, as well as having the world's best library of software & tools available for you to use (which Arduino/platformio have not even the faintest most shallow shadow of).
Yes, it may seem like I'm complaining - but my observations (complaints, as you say) come from the perspective of someone who has experience in this field and someone who wanders in the space expecting something even resembling a Raspberry Pi experience.
Kirkwood is 2009, that's where my "USB 2" comment came from. Only in the completely fringe grey-beard ARM ecosystem would it even come up in discussion (this thread). For the rest of the world it's completely irrelevant and relegated to history. Especially for HN - where a substantial portion of top rated stories are ML model developments from 30 minutes ago. Kirkwood is in The Smithsonian by comparison. Frankly, even referencing it or using it as a point of discussion only further proves my point of how far/long mainline ARM support is.
No one thinks, tries, or is under the impression that ESP32 is a storage controller to try to saturate anything. I think we're talking past each other on that point.
The Kirkwoods were great for the time. I had a PogoPlug and a Seagate DockStar.
Thermals were the Achilles heel though - very easy to overheat either one of those little boxes doing simple things like uploading a lot of files to its storage.
In the olden times, when computers were slow and dumb, GPIO or “Parallel Printer Port” as it was sometimes called, was no novelty. Some of us in this dark but innocent age, would use our computers for all manner of strange experiments. Driving foolish robots with long tethers, or experimenting with incoherent audio effects. But that was the before time, and todays computers make such play impossible.
A PI is an old style computer, just smaller and faster. I can use it as a desktop, then put wires out the back of it to play with an inverse pendulum. Can you do that with a real pc and an Arduino? Sure, but it’s how a little hobby computer ought to be.
I was there for when ECP and EPP were a new thing. Same for 50 ohm resistors on thicknet and plenty of other random things I could rattle off to establish bonafides.
I understand what you're saying but the world has evolved - there are plenty of tools for a variety of problems.
I have Pis going back to 1B, and they have their place. However, the supply shortage of Pis doesn't come from the use case you're describing. It comes from an entire world of people defaulting to them for anything and everything.
There are plenty of kids who want to play with GPIO where a Pi is the best approach who can't get one because there are people trying to build a NAS, a commercial product, drive multiple 4k displays, etc - even though there are plenty of better ways to go about that. At this point Raspberry Pi sucks the oxygen out of the room for so many more appropriate solutions.
In many ways the Raspberry Pi is a victim of it's own broad capabilities and success.
I agree with you. I had been assuming the difficulty in buying them was just another “global chip shortage” casualty.
There is no denying the impact of a strong community + cheap + multi-capable. Hopefully the shortage will end before long and the PI can once again be a cool little computer for hobbyists.
Latency is a killer for real-time purposes. Maybe things have improved, but years ago these were all more trouble than they were worth. May as well just use the usb direct.
Most of these stories are probably from when the RPi cost <40$. For most of its existence it was significantly cheaper, used significantly less power and space than comparable x86 boxes.
I run a Futro S740 thinclient as Server. J4105 from Q4 2017, so while older, not exactly ancient. Super low power, easily reaching <3W idle in many tests, purely passive cooled. And I only paid $50 used for one.
This is a category of machine that I didn't even know to look for, so thanks for sharing your experience - the bit about the low idle power is particularly good news.
It became so popular amongst selfhoster in Germany, that the companies selling them used almost doubled the prices :D There is a github with information about the S740 [0] including power measurements [1]. It’s in German, but should be partially understandable because it’s just data.
I think I might settle for a previous-generation S920 or S720, as I'm also in Germany, and those things are still super cheap - also, they usually have true 9-pin serial ports, which are an asset for what I'm planning to do with them.
The hardware does not seem to be mainline supported, which means that you are at the mercy of a vendor.
In other words, when they say "Ubuntu, Debian, Android, and Armbian" they don't actually mean you can run any future version of it, like you could for Raspberry Pi or x86 hardware. Instead, what they mean is: "We took a latest version of Debian/Ubuntu/Android/etc... and applied some patches to it to make it work on our board. But hey, it works today, and would maybe work next year too! And if you complain super hard we can fix some kernel bugs but no promises!"
This lack of support is a big reason to stay away from less famous SBCs.. Sometimes all you want is to make a project and forget about it, and that maybe fine. But if you want something to base multiple projects on, keep to stuff with mainlined kernels.
It looks like they are working on getting the Rockchip RK3588 mainlined.
CPU, PCI support is all there thanks to the QuartzPro64. At least with this one it looks like there are efforts rather than some of the other chipsets which don't have a chance.
Do any Rockchip chips have good upstream? Number of boards is not at all what I'd look for. Plenty of terrible sbc get made.
The track record of the chipmaker is the important thing, to me. It's not always clear cut. I think now there may be a more formal relationship, but Allwinner used to have good support from purely volunteer Bay Libre.
I'm far from certain but part of me wants to believe that's in sizable part also Allwinner doing a less ship, less ultra-forked job in the first place when they dump their own ugly gross hacked out vendor tree fork, where-as many of these vendor trees are just wildly awful.
Again, just to reiterate, I would never ever use board count as an indicator. Especially these days where there are only 2-4 (at best) chips available for use. Of course there's lots of sbc with this chip: the specs are great, the performance appears impressive, & there's literally no other chips to buy. That the vendor kernel is garbage & maybe there are oodles of little design bugs is like, not a big impediment to releasing hardware. Alas.
I think the biggest effort is being made by Pine64, the makers of PinePhone. Their PinePhone Pro used Rockchip's RK3399 which now has good mainline support (part of the selling point of the PinePhone Pro). Pine have released a SBC using the RK3588, and are working on mainlining that: https://wiki.pine64.org/wiki/QuartzPro64_Development#Upstrea...
That was my original point, I think this board might get some support but only by virtue of Pine also using this chip. Orange Pi and Banana Pi have never got anything upstreamed as far as I'm aware, and Rockchip seem uninterested as well.
I've played with these boards and they seem cool. I have a couple of NanoPi NEO3's because I really liked the form factor, and they perform really well, but I would never buy one of these and expect support. I would stick with Intel/AMD since they perform better, have better support, and the modern versions use little power and used is similar pricing.
Yeah, I saw you've mentioned the QuartzPro64, but I like linking to that mainlining status page. I've got a few RK3588 boards myself, the Rock5 and OPi5, but the use I wanted them for, better desktop OpenGL support with Panfrost than a Pi, isn't quite there yet. They're writing some new drivers for that:
The TuringPi RK1 modules should be available soonish too, they're prototyping boards right now. I'm hoping by the end of summer I can pick up a few for my cluster board.
It's always good to hear about support, as more options are never a bad thing.
I've been a fan of ultra-SFF computers since the original Mac Mini, and SBCs are just an extension of that. If you're shrewd and smart about how you use the hardware, it's a tremendous amount of real computing power that can work done cheaply and for little power.
For certain, even if I don't have own an Orange Pi, I feel I owe at least a hat tip to software maintainers for putting in the effort to ensure some other ecosystem has a future.
Apparently it looks like DietPi — the primary lightweight distribution I use on my Pi Cluster — looks to be regularly creating images for the Orange Pi now, as well as tons of other Radxa boards, and RockChip SBC boards in general.
Definitely looks like there’s efforts to long-term support a lot of popular SBCs out there: https://dietpi.com/#download
I bought rk3588 nano pi r6s a few months ago, running ubuntu 22.04 with a custom kernel. Been watching progress, and it's been quite active. Out of tree patches work with newer kernels, various upstreaming efforts are underway, and various people have it working. Device trees, patches, drivers and the like are still a bit hodge podge, but getting better. OpenWRT has an active effort, the GigE+Dual 2.5Gbe make fora nice small router.
For now I'm happy with the 5.10.x supplied with ubuntu 22.04. Ethernet works, it's stable, and crazy faster than a Pi 4. But I do look forward to using a mainline kernel. Here's two URLs to track the progress:
In my experience they are very well supported. I think the old USB controller lacks some of the performance optimizations mainly. Not sure how fast the new board versions get support, though.
Every time somebody tells me something is a great alternative to a raspberry pi, it doesn't work or idles at 90c or comes unsoldered quickly or is missing critical software or all kinds of nonsense that makes it not a close alternative, especially not at a near price point.
I'm willing to try stuff, but I'm skeptical of every new board because of previous bad experiences.
I've had similar issues, but bought a NanoPi R6S with 8GB ram. Got ubuntu 22.04 going on the first day, and it's a nice little server. It's crazy faster than a Pi4, stable, and runs cool. I paid for the $20 CNC machines case with passive cooling. Under normal use it's quite cool to the touch, and plan to make it into a small home router after I'm done tinkering with it.
I have this board, you better add some cooling but I had no issue setting up docker containers with node red, prometheus, grafana, Emby and Home assistant.
Too bad I currently can't succeed in using a raspberry pi camera on it.
Also the RockPi 4 is pretty cool, with a full fledged m.2 socket and stronger GPU. It has mainline support although I had to chroot install void linux since there was no official image
Orange Pi (Xunlong) has a record about not caring at all about up streaming their products. Mentioning the OPI 5 as an alternative to RPI4 is very abstruse.
At least I could buy an OPI 5. RPI4 is basically RIP, let me pun about it. Who cares if it is mainlined if there is no way to buy it at all or at a reasonable price.
I've my doubts about Arm's future with the reports of intending to up the price of licensing. That strategy seems insane with the Risc-v progress and increased uptake by big tech companies.
It'll be a spike in income, which will look good for IPO purposes.
Then a sharp drop to nothing, as everybody has moved on to RISC-V. But by then the company will be long sold to suckers through IPO. So it doesn't matter.
73 comments
[ 2.0 ms ] story [ 155 ms ] thread- There's a notice in there that 'orange pi' is registered trademark of their company but clearly stealing the pi thing from raspberry
- Finally they say that the pi is aging. I think they're missing the point of the raspberry pi.
This just kind of feels like a cash grab. The pi came out with the noble purpose of educating children in the UK with a computer that was so simple and cheap all you needed to do was grab a keyboard and a TV and you could learn to program.
They've accomplished so much more than what they started out to do.
It's just insulting to see something like this. If I wanted more power I could just get a NUC.
That said I have wanted to look into something like the pi but in a laptop form factor.
It'd have taken you two seconds to check that. No, it's not the first Orange Pi product.
> The pi came out with the noble purpose of educating children in the UK with a computer that was so simple and cheap all you needed to do was grab a keyboard and a TV and you could learn to program.
That's great and all but as a hobbyist who can't buy the Raspberry Pi in part because of production allotments to commercial integrations, it rings a little hollow.
Some places to watch the activity if interested:
My nano Pi R6S (another SBC with RK3588) has dual 2.5GBE and 1GBE, is pretty fast, and has 8GB ram. I expect it to be useful for many years to come.Still. This is not a fair comparison, but the used/refurb/old-stock PC market is just so ridiculously more capable. A HP EliteDesk G2 has i5-6500T (2015), 8GB ram, and 250GB SATA drive is $140 on amazon. It's a higher powered machine, 9W versus ~3W idle (shame on this review for having no power consumption figures at all), and much higher full-open power (I think my last kernel compile was <35W which is the TDP of the chip alone). And the OrangePi is significantly smaller form factor. There's cases where the OrangePi is a better fit. But for most general use, the x86 market has far far better value, does so much more, has so much more bandwidth (often 8x USB3.0 hosts, plus significant internal pcie).
There's periodic articles on repurposing thin-client computer hardware. But again there's just such a mismatched value proposition. Not the same shakeout: they have hot, old, low-end chips, but sometimes very cheap on secondary markets. But there's typically much better mini-PCs with vastly more performance for not that much more, that will idle at least as well if not greatly better. https://news.ycombinator.com/item?id=33791069
It's just so weird how few ARM chips are available. There had been a little more variety in who one could buy an SBC from. But now... Rockchip is the main company making purchaseable chips. Allwinner had some pretty fine offerings but has gotten sparse. MediaTek is doing excellent things in many spaces, but has almost no presence in accessible SBCs. There's tons of other chips but most of the market is invisible, is utterly uninterested in selling SBC type boards. There's a parallel x86 market, but Ryzen embedded systems tend to start at $400 for very sparse models. It's frustrating how deliberately unserved the low end is. That commodity x86 hardware dominates so savagely, by so many country-miles keeps feeling unfortunate. That ARM has such small niches it plays competitively in, and everywhere else it's broadly unavailable & unaffordable is a weird market dynamic, while these normally very expensive x86 cpus become available & affordable systems is such a strong & visible counter-trend.
Also see: "Unpopular Opinion: Don't use an RPI for that" on these mini-PCs: https://news.ycombinator.com/item?id=35260322
I was actually thinking of this a few years ago. My workplace was getting rid of a bunch of old Win7 thin clients, they had VIA x86 processors and 4GB of ram. Never acted on it though.
That will blow all these SBCs and thin clients out of the water.
If you don't need 10th gen Intel, and are okay with 7th, you can drop below £100.
It will be really interesting at this low end once they have to start competing with used Framework boards.
There are a few SoCs being targeted at Chromebooks that have PC I/O, including the RK3588 used in the Orange Pi, which makes them also suitable for SBCs.
Going from MIPI DSI to DisplayPort is as simple as adding a common bridge chip and updating the device tree.
Many of the new SoCs have PCIe lanes, so NVMe storage is easy, too. Some of the RK3388 boards even have a nice M.2 slot for exactly this purpose.
The interfaces aren’t a problem, though they are more limited than a full PC chipset.
We're only just seeing Thunderbolt show up in mobile, via the IPad Pro. The percolation down of the universal, good, flexible, kick-ass connectivity standards is semi-inevitable. Right now it's a ton of add-on extra chips, but that will go away, will get integrated. These discrepancies of PC chips having some things and mobile parts having totally different specs feels like it can't last forever, at least in every category. There will be some uptake of common capabilities.
It's strange. One of the many things I "harp" on here and elsewhere are all of the "Do $THIS on a Raspberry Pi" stories and articles around. What "bothers" me the most are people doing simple GPIO with a Raspberry Pi 4 when an ESP32 (or equivalent) would suffice. Or people trying to stretch a Raspberry Pi to do something an old, used, $100 x86 something would do significantly better in it's sleep. I understand the Raspberry Pi is a common denominator with a rich ecosystem but it still irks me in an irrational way. Don't even get me started on the people trying to build a NAS with a Raspberry Pi.. It's just wrong.
Getting on to to your comment - I have an entire graveyard of ARM SBCs (and more) that are incredibly poorly supported either by the vendor, Armbian, mainline, etc. I'm over getting excited by the latest "More performant, newer, better, faster" ARM SBC of the moment - only to burn incredible amounts of time getting it to do what a Pi or x86 something would do at a cost/performance/power trade-off.
I ended up with quite a collection. Small 2-drive NAS-ish DLink DNS-320's were great. IoMega IConnect had a solid (and user-replaceable) wifi chip & numberous USB2 ports, & was very afforadble on ebay. These kind of came latter-ish, but there were great other botique SBC or simple-device options pre-ceding GuruPlug: SheevaPlug, GuruPlug, OpenRD.
Many years have elapsed. But the upstreaming work on all these machines was very great. They required very little customization. Really hope I haven't messed up the attribution here, but I believe Marvell paid Free Electrons (named Bootlin at the time) to do a ton of the driver work. It's in kernel. Each device took tended to require only a little DeviceTree configuration to work great. UBoot got upstream support, & I'd replace the onboard uboot with a modern far more featureful uboot, greatly opening boot capabilities versus the minimal/older uboot devices shipped with (usb boot? yes. ethernet boot? yes.). Devices would slowly gain a couple modest not-super-important-to-me capabilities a year, or better tuning.
I've given away, sold, & (sadly) thrown out a bunch of these devices over the years. I still have two NAS running in friends' houses as off-site connected backup devices, humming along, but mostly ignored. Still, I fully expect, if I wanted to, I could completely freshen nearly any of these systems up. Brand new kernel, totally new Debian (actually many do run modern Debians, just on old kernels). My vibe has been very good, that old hardware doesn't die, doesn't get dropped, that the graveyard is merely of my interest to give a shit. Not all SBC are such: the Marvell Kirkwood was an early 2010's vibe of hope & promise, Linux uber-alles, where some good hackers got paid to Do The Right Thing, and I think this post is a testament to how obvious, how clear, how simple that path was. These devices are still fully supported. They still work great, kernels still support them, because Marvell worked with people who knew what the heck they were doing to get support upstreamed. These devices may well never ever die. They may be able to last forever. Because someone didn't just dial it in, because it wasn't just an internal team with some heavily forked kernel shipping shit: Marvell did the right thing, they upstreamed their work. The value of that, right now, seems like it may be ever-lasting.
The graveyard isn't necessary. Good vendors upstream their stuff & it will keep working.
It's also notable & exciting to see, as I've mentioned else-where on HN recently, Qualcomm chips have gotten much more upstreaming recently. The Chromebook folk really lead the charge with do-the-right-thing-duh are-you-stupid-of-course-we-will path. But after getting a bunch of slightly upscale tablet-grade Qualcomm chips supported upstream, we're starting to see much broader classes of Qualcomm chips getting supported, & seeing growing device-trees for a variety of phones. Turns out it kind of mostly is the same hardware, no surprise, with different cores, and the decades of shitty unsupportable heavily-forked-and-useless Qualcomm-SDK horse-crap they've shovelled at people is & always has been a farce & once the ball started rolling towards upstreaming it's just kept snowballing, mostly with little thanks to Qualcomm themselves, but just just just recently, the company seems to be a adopting a not so universally useless sad pathetic view & actually stepping up every now & then to actually do a little bit of the upstreaming themselves. For a company famously reputed to require lawyers to sign off on literally every ...
Yes, ideally support hits mainline. However, in practice between Linux mainline, things like Mesa for video, etc it just doesn't happen and trying to sell/tell people otherwise is disingenuous at best.
I agree there is significant value in the comfort of Linux but WRT to ESP32 I would argue the Arduino/PlatformIO ecosystem around them is actually second to the Raspberry Pi in terms of community support. There are "ready to go plug these variables in" examples for nearly anything you want to do that the hardware is capable of floating around everywhere on the internet. You can buy "ESP32 Cams" with TF Lite object detection from Amazon[0].
Also, not to be argumentative but the Kirkwood SheevaPlug line (I was early on that too) is the ARM equivalent of a truly embedded Intel Core something that no one dare touches if it's still running (it's USB 2).
Somewhat strange thing is I've worked on "commercial" embedded projects with things like the iMX line and Yocto - and the support there is pretty incredible.
[0] - https://www.amazon.com/s?k=esp32+cam&crid=R6C4OGY5Y5D3&spref...
But unlike everything else you've complained about, it has good upstream support & is reliable & all it's peripherals are quite un-flakey. I have upgraded my kernels numerous times. I do run Debian/testing. I haven't in a bit, but they would still run, and better than before.
You absolutely are correct about most of your complaints. There's a huge world of Ambian & other SBC that are off-tree semi-supported worlds to try to make shitty vendor crap keep going for a bit. And it's flakey, unreliable, & unmaintainable. But you are dead wrong about Kirkwood. Upstreaming means something & it has kept a lot of decade old devices running & stable & happy & it's just obvious. The negative experiences you cite are real, but they're a owed to a variety of over-aggressive under-budget hardware builds and more-so to never ever having actual real support, to never having had drivers good enough to get upstreamed, to being a bunch of shitty off-tree vendor code that is just poorly done. The network of enthusiasts trying to hack access to these sytems, to make off-tree Ambian or OpenWRT support at any cost is admirable, but ultimately, the actual cost of doing the right thing, like Free Electron does, is tiny, especially versus the pain incurred from not doing the right thing. I think you're missing that distinction, by lumping these all together.
USB2 is also really not that bad for many remote-storage applications. It's also a bit comedic to, in one sentance, blast it as bad, then go on to talk about ESP32 systems & other microcontrollers. Which can almost never saturate even USB2, & almost never have any kind of USB3 what-so-ever. Sure they may run TensorFlow Lite. But they are markedly smaller systems.
Just company community support seems arbitrary. Maybe Arduino has the best support community on the planet. It's still a vastly less capable platform with incredibly less software available to help you build systems/solutions out of. And showing up to Arduino will be a unique bespoke different experience, whereas running a Linux system will have a common frame of reference & transferable experience into the rest of computing, as well as having the world's best library of software & tools available for you to use (which Arduino/platformio have not even the faintest most shallow shadow of).
Kirkwood is 2009, that's where my "USB 2" comment came from. Only in the completely fringe grey-beard ARM ecosystem would it even come up in discussion (this thread). For the rest of the world it's completely irrelevant and relegated to history. Especially for HN - where a substantial portion of top rated stories are ML model developments from 30 minutes ago. Kirkwood is in The Smithsonian by comparison. Frankly, even referencing it or using it as a point of discussion only further proves my point of how far/long mainline ARM support is.
No one thinks, tries, or is under the impression that ESP32 is a storage controller to try to saturate anything. I think we're talking past each other on that point.
Thermals were the Achilles heel though - very easy to overheat either one of those little boxes doing simple things like uploading a lot of files to its storage.
A PI is an old style computer, just smaller and faster. I can use it as a desktop, then put wires out the back of it to play with an inverse pendulum. Can you do that with a real pc and an Arduino? Sure, but it’s how a little hobby computer ought to be.
I understand what you're saying but the world has evolved - there are plenty of tools for a variety of problems.
I have Pis going back to 1B, and they have their place. However, the supply shortage of Pis doesn't come from the use case you're describing. It comes from an entire world of people defaulting to them for anything and everything.
There are plenty of kids who want to play with GPIO where a Pi is the best approach who can't get one because there are people trying to build a NAS, a commercial product, drive multiple 4k displays, etc - even though there are plenty of better ways to go about that. At this point Raspberry Pi sucks the oxygen out of the room for so many more appropriate solutions.
In many ways the Raspberry Pi is a victim of it's own broad capabilities and success.
There is no denying the impact of a strong community + cheap + multi-capable. Hopefully the shortage will end before long and the PI can once again be a cool little computer for hobbyists.
[0]: https://github.com/R3NE07/Futro-S740/blob/main/README.md
[1]: https://github.com/R3NE07/Futro-S740/blob/main/power_consump...
https://hackaday.com/2023/03/07/laptop-motherboard-lets-boot...
Don’t let me down, HN!
In other words, when they say "Ubuntu, Debian, Android, and Armbian" they don't actually mean you can run any future version of it, like you could for Raspberry Pi or x86 hardware. Instead, what they mean is: "We took a latest version of Debian/Ubuntu/Android/etc... and applied some patches to it to make it work on our board. But hey, it works today, and would maybe work next year too! And if you complain super hard we can fix some kernel bugs but no promises!"
This lack of support is a big reason to stay away from less famous SBCs.. Sometimes all you want is to make a project and forget about it, and that maybe fine. But if you want something to base multiple projects on, keep to stuff with mainlined kernels.
CPU, PCI support is all there thanks to the QuartzPro64. At least with this one it looks like there are efforts rather than some of the other chipsets which don't have a chance.
The track record of the chipmaker is the important thing, to me. It's not always clear cut. I think now there may be a more formal relationship, but Allwinner used to have good support from purely volunteer Bay Libre.
I'm far from certain but part of me wants to believe that's in sizable part also Allwinner doing a less ship, less ultra-forked job in the first place when they dump their own ugly gross hacked out vendor tree fork, where-as many of these vendor trees are just wildly awful.
Again, just to reiterate, I would never ever use board count as an indicator. Especially these days where there are only 2-4 (at best) chips available for use. Of course there's lots of sbc with this chip: the specs are great, the performance appears impressive, & there's literally no other chips to buy. That the vendor kernel is garbage & maybe there are oodles of little design bugs is like, not a big impediment to releasing hardware. Alas.
I've played with these boards and they seem cool. I have a couple of NanoPi NEO3's because I really liked the form factor, and they perform really well, but I would never buy one of these and expect support. I would stick with Intel/AMD since they perform better, have better support, and the modern versions use little power and used is similar pricing.
https://www.collabora.com/news-and-blog/news-and-events/panc...
I've been a fan of ultra-SFF computers since the original Mac Mini, and SBCs are just an extension of that. If you're shrewd and smart about how you use the hardware, it's a tremendous amount of real computing power that can work done cheaply and for little power.
For certain, even if I don't have own an Orange Pi, I feel I owe at least a hat tip to software maintainers for putting in the effort to ensure some other ecosystem has a future.
Definitely looks like there’s efforts to long-term support a lot of popular SBCs out there: https://dietpi.com/#download
For now I'm happy with the 5.10.x supplied with ubuntu 22.04. Ethernet works, it's stable, and crazy faster than a Pi 4. But I do look forward to using a mainline kernel. Here's two URLs to track the progress:
https://gitlab.collabora.com/hardware-enablement/rockchip-35...
https://gitlab.collabora.com/hardware-enablement/rockchip-35...
https://patchwork.kernel.org/project/linux-rockchip/list/
I'm willing to try stuff, but I'm skeptical of every new board because of previous bad experiences.
Too bad I currently can't succeed in using a raspberry pi camera on it.
ill try to reproduce/document for you.
I have the original 5 board running for some weeks now with some docker containers and it's running great.
I love my VisionFive 2. It is no Orange Pi 5, but it is fast enough, and it is RISC-V proper.
Interesting statement. Care to elaborate?
Then a sharp drop to nothing, as everybody has moved on to RISC-V. But by then the company will be long sold to suckers through IPO. So it doesn't matter.