Geez for the prices listed in the article (153 EUR and 301 EUR), get a TMM (tiny mini micro - https://www.servethehome.com/introducing-project-tinyminimic...) computer. These are the SFF/USFF off-lease older enterprise computers like Dell Optiplex. Note that ServeTheHome.com does power consumption for every model it reviews.
These ~1L sized computers idle in the 7-13W range. Even with EU electricity prices, it would take you a very, very long time to make up the difference from a 5W SBC.
You also get standard x86 support, normal expand-ability, M.2 slots, PCIe ports, etc. You lose GPIO support.
For $100-$120 on eBay, a search for 'dell optiplex 3070' typically results in getting either a i3-9100 or i5-9500, 8-16GB memory, and 120-256GB NVMe. I bought one a couple months ago for Blue Iris and it uses 8W at idle. Here's an example of a USFF (ultra small form factor) for $110 - https://www.ebay.com/itm/325826958401
I think RPis have pretty much outgrown (outpriced?) their usefulness, except in very specialized applications:
If you want to run a server-task, listen to the common chorus and buy an 3-7 year old SFF business PC. Slap Proxmox on it, and you can virtualize a bunch of small 'servers' as LXCs with plenty of CPU and memory. Easy to add storage, dependable, and well-supported. Power draw is highly overstated for typical ancillary-server tasks, and if you were planning on 2+ Pis, you're now in the same ballpark.
If you want to hack together something with sensors/electronics, use a $5 ESP32, which likely has way more power than you need. If you need more processing power, move that from the edge device to something like the cheap SFF above. Plenty of sample ESP32/8266 Arduino libraries to expose/control the GPIO via a simple API, MQTT, or UDP/TCP payloads. Then, you can write your processing logic in whatever language you want, on a powerful server, and the edge device doesn't need Linux distro updates to blink some LED strips or whatever.
The more I mess with SBCs the more I lean towards this method.
The M720q's or M710q's are two of my favorites.
As an added bonus the chargers even work with my lenovo laptop(s).
This sort of underscores the weakness of RPi as a complete computer:
A complete computer is at least somewhat hardened against ESD and what not, has onboard storage of some description, and isn't severely throttled from a power standpoint.
I like RPis and similar for rapid prototyping if I know I need GPIO control, but if you don't, there are better machines out there, at similar price points.
If you need GPIO you might be able to use a uc like an rp2040 or esp32 and connect that to your dell machine. The uc run as cheap as $5 for something that is 32bits with wifi.
That $4 dev board looks like it might be an esp32 killer, or at least good competition. I was thinking of the esp32-c3 which is also risc-v with wifi and ble. It currently has really great rust support. You can even build the stdlib for it so you don't have to mess with no_std. The bl602 doesn't seem to have rust support yet, but I'm looking forward to more options.
Thanks! It wasn't on the awsome rust embedded github yet, but when I dig in I found it. The bl602 does support wifi, but only 2.4 and I presume at lower rates. It definitely doesn't run linux though. But I wouldn't put it past someone to figure out a way.
Try github[0]. There's a community around this chip, with repos dedicated to collecting the documentation and code the vendor itself wouldn't provide, or would require filling request forms to obtain.
It tends to be like this with Chinese chips, and often times western ones too, unfortunately.
That's where I've landed for the most part. Last real gasp for dev-board-style SBCs, for me, is when I'm severely space constrained, but can afford to locate a USB hub somewhere else. Then its just simple serial comms with a microcontroller that does the timing-sensitive pin twiddling.
> You also get standard x86 support, normal expand-ability, M.2 slots, PCIe ports, etc.
Let's acknowledge that there are good reasons for avoiding x86 for those who care. The security problems for one, and the long shadow of Microsoft over the ecosystem. Perhaps an additional concern about the supply chain. The closer we can get to an open-source system without blobs, the better.
But the above concerns apart, it seems clear that R.Pi 4 and 400 were peak R.Pi. The performance and low-cost of Intel's recent x86 NUC mainboards is impressive. AMD is also offering strong value in the SFF market.
As I long-time R.Pi/Odroid user, I continue to enjoy these devices. But their lunch has been eaten by competitors.
To be fair, you also need to account for consumption when you make the machine do something. The TDPs for these machines are in the 25-35W range, and hitting 50W does not seem unreasonable. You of course won't be running that the entire time. The larger SFF machines with 65W TDPs will go way above that.
An orange pi 5 uses 7.5W under full load, 3.3W idle in comparison.
The orange pi 5 Geekbench score is ~75% of the i5-8500T used as example, so no - the sff won't be done in a fraction of the time. These CPUs are old and crusty after all...
I do agree that modern CPUs are required for high-load situations.
If you agree, than relative measures of TDP/benchmark at the redline are pointless, ignoring that TDP number are often inaccurate anyway.
What you probably want to compare for most home uses is power at some low idle vs performance at that CPU throttle level. The difference is probably a Watt or two max, which never would justify a brand new SBC over repurposing something headed to ewaste for environmental reasons.
Is orange pi 5 as fast as 35W TDP intel CPU ? How many times slower it is than the 65W one ?
Average consumption matters, not peak. And you can pin CPU to not go on higher frequencies if you don't want to go into inefficient max frequency/max voltage region.
You will also be recycling already produced device.
According to Geekbench, an i5-8500T scores ~1100 single core, ~4000 multicore with it's 35W TDP, while the Orange Pi 5 gets ~840 single core and ~2900 multicore. About 75% of the performance at less than 20% of the power. Not too shabby, despite the orange pi being no efficiency star.
Recycling is good, but only to a point. At some point the cost of operation (and resource pressure from its use) will exceed the cost of a more efficient device - at least until we get abundant green power...
Or you can get a brand new N100 mini pc from Amazon for < $150 with 16Gb RAM and a 512Gb SSD (probably cheaper on AliExpress). It's much newer (10nm instead of 13nm+) and can idle as low as 6W.
And for GPIO (if really needed) you can always add something like an USB Arduino micro (clone) for a couple of bucks.
I got one of these recently. It was ridiculously easy to just install stock debs on.
Every time I look at the wall wart plug pack it comes with I laugh to myself. 8W.
In terms of price, speed and compatibility the Raspberry Pi looks laughable.
If I want GPIO I just use ESP32s with wifi. I am then unconstrained in terms of other stuff hanging off them. Using the new nano ESP32 boards I can just hot glue them to whatever I want to talk to.
Even ~2000, that might still have squeezed you into the bottom of the list? Note that power consumption of above system will have been in the 10s if not 100s of kW. To say nothing of size or purchase price.
That's only 1 human generation ago. What on Earth are you doing with these things? Running physics simulations of nuclear explosions, or what?
These look really cool. The biggest downside is having 100 brands to choose from. It makes me feel like there's a good chance I could pick something that ends up being total e-waste.
I've also compared many 1L PCs from a performance/price/power consumption perspective.
I decided on the HP prodesk/ elitedesk 400/800 G5 with the i5-9500T. It idles at ~2-3W and has considerably more computing and storage than these embedded solutions. I only see something like a raspberry pi be handy when relying on GPIO ports.
I don't know what you did to get yours idling that low. My 800 i5-9500t idles at 9w. I took out the flex io module & wifi card, and make sure nvme has "spin down" but still nowhere near that low.
I had some Acer chromeboxes that would do 4.5w easy without any tuning or setup, and they used pretty low end dyal-core i3's from just barely before the i5-9500t. Generally I don't think servethehome finds many tiny mini-pcs below 7w.
I can attest to coffeeri's consumption, albeit on pure server use-case only. I own a OptiPlex 3080 with an i5-10500T and under default BIOS settings (power states and whatnot), I was very surprised to see a 1.7-2.5w power draw on idle, but under typical desktop use, this becomes 6-8w. If this is your draw under a server deployment as well, unfortunately I don't know of any methods to rectify this personally.
When it has to roar however, the draw is substantial compared to my ODroid, but since it idles the majority of the time, I'll take it.
Have you enabled the deep sleep state in the BIOS?
It always depends what you are running on it. For me it's very light with only paperless-ngx and a smb share on a nixos.
I also only have one RAM stick installed and set up powertop on startup.
Not quite the same machine, but the Wyse are plenty fast enough for everything I've needed a home server for. It's quad core and has Intel Quicksync for Plex, and can hold a lot more memory than they state in the official documents.
The Serve The Home articles inspired me to go this route. I want to improve my understanding of Kubernetes for work. I picked up an HP EliteDesk 705 G4 mini desktop and an HP EliteDesk 800 G4 mini desktop. Each was less than $200, and included 16 GB RAM, 256 GB SSD, HDMI and DP outputs, and a power brick. One from Amazon ("renewed") and one from eBay. My 705 came with a Ryzen 2400G, and the 800 came with a i5-8500. I'm not planning to stack them so I was happy with the 65W chips although it means keeping the top of the case clear for cooling. Serve The Home recommends the 35W chips if you are going to stack them.
You don't need to buy the OEM power adapter for $46.99
The great thing about these Optiplexes is that they use a standard 20v 4.5x3mm DC jack for power. I run mine off a UPS with a usb-c to barrel jack cable.
No they don't. It is a barrel jack you can obtain an adapter for elsewhere, but your adapter is going to be missing the third pin. This type of system uses the 3rd line to communicate with the power supply and during startup the computer asks it how much power it is capable of supplying. If the power supply doesn't answer, it drastically throttles your CPU. Sure, you can boot the system, though it would be throwing away the majority of your cpu performance.
Furthermore, I have several of these systems and tried sourcing 3rd party power adapters online. For this aspect, not one of them worked correctly.
The author got almost everything correct except some minor parts. Using his charger as an example, it can be broken down as:
DELL 00 AC090 195 046 CN09T2157161543835EAL 03 .
046 is the output current 4.6A instead of his partitioned as 46(there are power supplies output more than 10 amps).
The 22 bytes is the Dell PPID, which should be also present on the charger's label or the box:
CN: Manufactured in China, some of them are TH which means made in Thailand.
09T215: Dell Part Number.
71615: Manufacturing factory code.
438: 2004 or 2014, March, 8th (Dell use YMD 3 bytes thus it is a lossy conversion always).
35EAL: This should be the serial number.
The last part 03 is a CRC16 (x16 + x15 + x2 + 1) checksum.
My router is an Optiplex fx160. Been running for years. I had to upgrade the storage and add another NIC. Has a built in power supply. I've never managed to run a Pi for this long, even with a decent power supply.
They might be good performance wise, but if you watch the Jeff Geerling video he mentions in the post, it will show you how much trouble he has with it. Custom distros with broken repos, non mainlined outdated kernels.
Having owned many raspberry pis, a handful of rock pis (4A, 4B, Zero and S), Mango Pis, Khadas VIM4 and other SBCs (I have a problem haha)... I can say without a doubt the two shining stars the Raspberry Pi has are their distro/software and general ecosystem.
Just upgrading and dealing with downloading/upgrading the Raspberry Pi is a dream compared to the Rock Pis. The Khadas is a little better, but still I have a hard time upgrading it and sometimes certain images just won't even work from the company. Sometimes when trying to upgrade Rock Pis with eMMC you will find 2 or three different ways to install the freaking image on their own wikis.
There's also the "security" aspect for me, I really don't trust some of these distros (maybe I shouldn't worry??). A random image someone put on their website with broken instructions and messed up repos/kernels doesn't exactly instill confidence to me.
I will say the one GREAT thing from Rock Pi is their rock pi x. But it has an x86 processor and I can just install any distro. I wish they were easier to purchase and had more memory!!!
Yeah, I wouldn't recommend one that doesn't have mainline kernels. But Pine64 does good work on getting stuff upstreamed. The RK3399 used in the PinePhone Pro, the Orange Pi 4, Rock 4, and other boards has proper mainline support. I'm also a fan of the open sourced GPU drivers for that SoC done by the panfrost team that has desktop OpenGL support (not just the ES stuff).
Especially when dealing with low-level hardware stuff, like SPI and PWM and other pin I/O, design can be much easier when sticking to Raspberries only, because there are inevitably a lot of quirks that are different between companies.
Yup, can confirm. Was trying to setup an Octoprint machine for my 3d printer, none of Raspberry Pi’s were in stock. After trying to catch one for weeks, I ended up buying the second-newest revision of OrangePi that was pretty powerful (not the barebones version).
To say it was trouble would be an understatement. Even after I successfully completed the setup (which took forever, as the documentation was contradicting, confusing, and felt like it was worse than machine-translated to english). Stability wasn’t there either at all. Occasionally it would just lock up and freeze, requiring a hard reboot. It would heat up pretty bad sometimes, despite an extra heat dissipator/cooler (that i triply made sure was installed properly). Webcam would randomly just refuse working (and I had to manually edit linux configs to get it to work in the first place). Sometimes SSH connection would just die and I become unable to connect to it again without either a reboot or restarting networking services on it (which would require me physically being next to it, thus defeating the point of Octoprint being remote). And mind you, I didn’t run it as some heavy production-tier 3d printing controller, I would print something a couple times a week tops, and nothing too complex or too high res (in terms of the model being printed).
All while Raspberry Pi has great documentation, and Octoprint setup on it is pretty much plug and play with just a few clicks of installing dependencies (or even simpler than that, if you install a dedicated Octoprint distro made specifically for Raspberry Pi).
Custom and outdated distros is a big problem with ARM boards unfortunately, unless you are a big and popular player like the Raspberry Pi. But even then a lot is not supported in mainline
That's my experience with many alternatives, basically nothing (INCLUDING rPi) is just supported in mainline, but rPi's patched version of Debian is well supported unlike the alternatives.
If just for playing around paying extra for pi just to have something that works makes a lot of sense.
Where do you buy an Orange Pi? When I was researching these devices I found it to be extremely weird/sketchy. These are three different websites all apparently from different companies:
Welcome to Chinese fruit clone vendors. They're all copying RPi and now copying each other. Good luck spotting the good ones.
Then when you get the device it won't do anything new after a while because the vendor never maintains their distro and the GPU or device tree are is proprietary so nobody else can either. If you're lucky maybe there's an Armbian build which marries new userspace with the older kernel and its unmaintained drivers.
This is why Raspberry Pi still dominates this market.
Ebay or Aliexpress are good options with buyer's protection. I bought quite a few OrangePis in my time, although that was at a time when the Orange Pi Zero was 9 EUR a piece. At current prices, I don't see the point.
The OPis themselves were quite decent devices, Raspbian worked very well on them.
So the third link is the official one? Its footer has some very sketchy looking links, including to two different telegram channels, three aliexpress stores, and some guy's skype username...
Then again, we might not even be seeing the same site, since it forces http:// and could be modified in transit by anyone on the path of either of our ISPs and the web host.
In my experience on Aliexpress, 3 stores aren't unusual. Most of the stores I buy from on Aliexpress (mostly 3d printer parts) have at least 2 stores. Even the reputable ones like BigTreeTech which is a pretty popular for 3d printer control boards have at least five stores by my last count. According to them, Aliexpress don't allow multiple accounts to manage a store, so they make a new one for each salesperson.
Orange Pi / Xunlong is a stable brand that's been primarily on Aliexpress for 7-8 years IIRC, producing various boards, all of which work fine for me. Very reliable seller, and the product + it's cheap. In the past they had Orange Pi PC, which was half the price of Raspberry Pi 2 or whatnot and much better in connectivity (4x real USB host interfaces, ...).
Their software I would not touch, but I never had to. Most Allwinner and Rockchip boards are well supported upstream. I just upstreamed support for Orange Pi 5 Plus a week ago or so.
All you ever need with these random boards is to build your own Linux kernel and bootloader for them, which is a few fairly standard steps, and then you can use any normal aarch64 distro on them, even if the board is not "supported" by the distro.
I just have a script to cross-build the recent kernel for the boards I own, and update it remotely. Not a huge issue.
Unlike standardized PCs, embedded systems are not just about specs and prices, but much more about stability and the whole ecosystem. What helps a higher spec if you can not find a stable driver or the documents are so bad, you can not make heads or tails from that? What helps a cheaper price tag if you have to spend days trying to figure out why sometimes a hardware connection works but sometimes not?
Ultimately, time is money. If I manufacture something en-masse, obviously a few dollars difference is important. But for my one-off hobby project? I'll stick with what works.
No they don't. I have a Dell Optiplex 3070 SFF (not even ultra small form factor, it is standard desktop sized) that idles at 8W. It has an i5-9500, 8GB memory, and a 128GB NVMe.
My Rpi Zero idles at 0.2W - it's nor in the same computing class as SFF/NUCs. I don't run a homelab in it; and I find it ideal for ambient, always on compute/networking/monitoring tasks
That's not totally true. I have a Dell Optiplex mini-PC version (and not that new) and it draws on average 10W, which is surely more than a RPi4 but not 10 times.
There are multiple uses for the Pi. The mini PC is good are being computer, which Pi really isn't good at. But the mini PC can't do the other uses.
For example, the mini PC doesn't run on PoE. It won't fit in enclosure and can't put it on pole outside. But that is what I'm planning to do with Pi for ADS-B receiver.
The mini PC doesn't have PPS support or ability to install GPS receiver. That is what I need for Stratum 1 time server.
A lot of things I want to do just need USB. The question is if one PC or multiple Pis is better. I like idea of Pi per task since are independent units.
You need to attach the GPS and its PPS pin (Pulse per second) to the gpio on the rpi. Gives you microsecond NTP, pretty cool if you need that. It can be used to sync network, radios and audio over long distances.
every person I know who has had problems with the Pis has had problems because they've cheaped out on either storage or power supply.
the official power supplies (and others with the same ratings) are only sufficient if you don't plug anything into the USB port, don't have any hardware drawing power from the GPIO, and don't utilize Bluetooth and wifi 100% of the time at full throughput or anything.
boot to a fast thumb drive, power the board via a 30W or more power supply, and get an onboard battery so in the rare event that the onboard power regulation can't keep up, a battery powered supply can deliver power through the GPIO pins.
Raspberry Pis are built down to a price point, not up to a quality level. they are extremely good per unit of money spent on them, but they are not perfect.
calling them "shit" shows a general lack of understanding, to me.
All solid advice, though I mostly prefer to use the Zero W-s, which need less power, and to boot off a decent sd card into the 'overlay filesystem'[1] which is RAM-only, so writing logfiles etc. doesn't wear out the sd card (I also use 'high endurance' sd cards intended for dashcams etc. which are a little slower but theoretically last longer). I have a couple of Pis that have been running for years this way, no probs.
they aren't industrial embedded computers, so comparing them to some $1500 hardened industrial embedded platform is not only unfair, it's not a comparison that's respectful of the other participants in the conversation. you're relying on people not knowing about embedded hardware so that it appears your comparison is fair.
The Raspberry Pi is not an embedded platform in the ways that you are using for comparison. The Raspberry Pi is an educational platform.
it is very hard to know if your raspberry pi issue is power related if you don't do what you can to make sure that power isn't an issue.
prior Pis didn't log when they had power issues, and current ones do, if they are able to.
on a Pi 2 A (I think) I had to hook up an oscilloscope to catch all the tiny power problems I was having, and only then realized what was happening. I was 100% sure I was delivering enough power. I was not. those symptoms were only visible without the scope as weird errors in the OS and running applications, and corrupted SD cards.
After my 3b+ died and the 4 was nowhere to be found, my enthusiasm kind of shifted towards something even crazier - Ryzen 7840 based machine for 5x the price.
Since these SBCs keep getting pricier and need active cooling while software is still far from ideal, it's not that big of a jump.
You might be able to run on a 5W supply, but my experience is that booting on a less than 15W supply is a recipe for corrupted SD cards and hard-to-trace bugs.
My biggest frustration with hobbyist boards is how little power the can source to e.g. the USB ports. Its super frustrating to need a powered USB hub to add e.g. a USB indicator light. The fact that you can power the RPi from the powered USB hub seems hacky, but maybe that was always the intent.
Try to run even a Pi 3 on 5W and let me know how it goes. You'll get undervolt warnings like crazy. Not sure if you looked at the article but these are the power adapters listed. These SBCs are not exactly low power anymore:
You'd need to limit the CPU operating points or offline most of the CPU cores to get lower max power consumption under load. 5W is too strict for Orange Pi 5 Plus.
And of course forget nvme if you want predictable low power consumption. nvme alone is 5-10W load. The board itself can run under 5W, but you'd have to carefully consider what peripherals you can use and how.
M600 idles around 5-8 watts. If you spank it hard it gets up to 12W. That has a 500Gb SATA SSD and 8Gb of RAM and ethernet.
Note that the M600 does not have any active cooling. The entire unit is a big heatsink. The thing will run as a headless network computer with just the power supply (a proper Lenovo SMPS brick) and an ethernet port connected.
power usage on a computer is not constant by any measure.
power usage spikes with CPU activity (among many other things) and if your power supply can't deliver the current that is required, voltage will drop, and things start failing, but only for 1ms or maybe even less. maybe even a single clock cycle in some cases.
if you can't keep the board powered fully for every clock cycle, then you are going to have problems, and that's true for any computer, not just the Pi.
If you are hobbling with a PI it's all about the community support, form factor, low power and the header pins. Hopefully if you are buying a Pi you already have a specific project in mind.
For example: I use a few zero 2 w's with shairport-sync to make Airplay stereos. I use the header pins to control a relay to turn on and off an audio amp. Pi 4+ would actually work a lot better for this especially when playing audio on multiple of these setups at once. A Lenovo mini pc wouldn't be as easy to hide.
I wouldn't go too far down the "community support" thing with the RPi foundation. Careful if you say anything negative, even in a constructive tone, in case Liz shuts your threads down. I had some problems with brownouts on the RPi 2 and posted a technical thread. It was deleted and my account terminated. There are various reports of this from community refugees who switched to other platforms.
I guess by community support is if you want to build something you'll find 100 tutorials on how to do it or it will even be part of the GitHub project. I never visit the official forums. These community mods are crazy with the little bit of power given. I had similar issues with arstechnica.com before it became an echo chamber.
ESP32s. So easy. As well supported as any RaspberryPi.
Proper real time on the metal. Interrupts. Native i2c, i2s, spi, serial with a cross-plane so you can re-route most pins around, depending on the board and a few minor caveats.
Any non-real-time operating system will eventually get a little bit buggy around timing intensive operations like I2C or (especially) SPI. Better to have your GPIO operations performed by a dedicated micro that only handles your serial traffic and whatever pin-twiddling operations you need to do. Ideally, your serial traffic is in a binary protocol like MODBUS-RTU, so your traffic parser is as fast as possible.
Bitbanging I2C or SPI is tedious but generally pretty achievable in a bare-metal application. But unless you're hacking your kernel to force that kind of timing fidelity, bit banging on an RPi running Debian is just impossible.
Sounds like this would be very annoying compared to a SBC. You have typically several SPI, UART, I2C, gpio interfaces on the pin header on usual ARM SBCs. Ready to use from Linux via standard API, using DMA, etc.
I once had a very simple I2C operation performed on an Orange Pi running project nerves. I had to add a try-again loop to the I2C operation, because the underlying OS kept disrupting the I2C timing. I2C is generally run between 100-400 KHz, so you'd think it wouldn't be an issue, but it definitely was. You can get around this by patching the kernel to carve out timing protection for those operations, but speaking as a firmware programmer, if I'm patching the kernel to avoid writing a simple arduino sketch, I've made some pretty deep mistakes.
100 khz is still very fast, with sub-millisecond I2C tranfers possible.
You'll have the same issue on any SBC running Linux. You'd need to carefully use realtime process scheduling, if you want more predictable timing on a general purpose OS for such stringent timing requirements.
What I see a lot of is people trying to cram all sorts of shit into a Pi and expecting everything to work properly. I actually wrote a fairly small event-driven kernel for AVR parts a few years ago called XOC "exec-or-communicate" which could do small real-time tasks easily (combinational logic, state machines, interrupt handling) and delegated complicated ones to a host machine over virtual serial port. It was used for a metering system I developed. The AVR side could withstand a complete host failure and reboot.
> GPIO Zero supports a number of different pin implementations (low-level pin libraries which deal with the GPIO pins directly). By default, the RPi.GPIO library is used (assuming it is installed on your system), but you can optionally specify one to use. For more information, see the API - Pins documentation page.
> One of the pin libraries supported, pigpio, provides the ability to control GPIO pins remotely over the network, which means you can use GPIO Zero to control devices connected to a Raspberry Pi on the network. You can do this from another Raspberry Pi, or even from a PC.
Presumably, e.g. TLS to secure that control channel is your responsibility.
> TTL serial refers to single-ended serial communication using raw transistor voltage levels: "low" for 0 and "high" for 1. [31] UART over TTL serial is a common debug interface for embedded devices.
I've been using one for my media centre for many years, only swapping it out for newer versions for the performance upgrades. No issues here, and things like CEC are plug'n'play.
The Pi 5 looks neat and I hear it's finally got a power button built-in, which was always the sorest point of the rpi series for me - the fact that it had no good story for safely idling was always absurd. The default configuration was "if you lose power there's a good chance the filesystem will be corrupt".
Still, does the new one do better beyond having an off-button? Like does it wake via USB properly? Does it have some way to safely recover the filesystem after power-loss?
Because that has always been the gigantic asterisk on the otherwise-cool RPI. "It's a ticking time-bomb until its memory gets corrupted, and if it crashes you have to roll the dice by power-cycling it" was never a good story. That and its unbearable prima-donna pickyness about voltage.
> The default configuration was "if you lose power there's a good chance the filesystem will be corrupt".
"Good chance"? We used to do kernel module development directly on Pi3/4s at this one client of mine, and I would OOPS the kernel from time-to-time (couple of times a day on average in the beginning stages) and never had a loss that ext4 couldn't recover from (and as the Pis lacked a "reset" button, resets were done via power-cycling).
We used standard PNY/Samsung Micro-SDs, too (i.e., not industrial-grade varieties), but we did pick the faster-IO variants when possible.
FWIW once I started deploying my Pis with automotive-grade SD cards bought from Arrow and Canakit 5.1v PSUs, all the stability & integrity issues completely disappeared. Costs a bit more, bringing the price of a setup closer to a used SFF HP or whatever, but it’s super stable.
Whether we should feel bound to go to such lengths is another topic…
Yeah, that's what kills me with rpis. Yes they're cheap... until you add a chassis, a safe SD card, a 1st party power supply that won't give you the dreaded voltage warning icon, etc.
I don't regret owning some since they basically taught me to do Linux adminning, but the shopping experience is a bit like buying an airline ticket.
Yep, I often wonder how the story would've gone if they'd put a little cap and some circuitry on to make sure the SD card would come down gracefully, and configured the power circuitry so the Pi is comfortable running on a little less than 5v.
When they have a recommended OS with their brand name on it (Raspbian/Raspberry Pi OS), it's a platform feature. Either I'd expect the hardware to provide some minimal battery or something for safe shutdown, or I'd expect the software to handle sudden power-loss gracefully. They provide both the hardware and the software, there's no way for blame to leave their ecosystem.
Raspberry Pi can't redirect blame for filesystem problems if the user is using an OS that's called "Raspberry Pi Operating System".
I applaud your user-first attitude to responsible branding, but we all know open source software and economics of mass production cheap boards don't work that way.
> One thing worth pointing out is also the compatibility of software and the breadth of the community.
>Obviously, the Raspberry Pi ecosystem and community is much stronger than the Orange Pi or Rock communities. However, I can basically promise you that the Orange Pi community and Rock community will both grow in coming years.
This seems way more important than the differences in computing power, and even the (massive) differences in price. Until I can seamlessly use raspberry pi OS and all the RPi software, those other boards are just non-starters for me.
I've been waiting patiently to shut off my old power-hungry NAS server and move the management of the 8 disk zfs array on to something like a raspberry pi.
It looks like I might be a step closer with the RPI5's pcie expansion options.
I'm really not sure that you'll get that much power savings. Given the idle consumption of a drive is ~6W I'm actually surprised my 10 year old server/NAS only consumes about 110W for a system with 10 drives, 10GBase-T, 8 DIMMs, Intel 3930k and fans/some USB peripherals.
there hasn't been a shortage for a while now, but you probably didn't hear about it because shortages ending isn't considered newsworthy.
it's the same with sriracha, you can pretty easily find it for msrp now, but lots of people still think it is hard to find because the last thing they heard on the topic was a news article about bottles selling on the gray market for $40-$80 back in may.
I check every so often, and checked before I posted that too. First hits I got from the Raspberry Pi website were all sold out, so I presumed it was business as usual. Cycling through a few more shops, I see it is available in some places now.
Those RPi4 supply problems seem to be mostly fixed. And RPi5 may come from different production line(s) such that it's available in quantity even if the 4s were not.
It's a lot better in the last few months. If you're still looking for an RPI 4 then sign up on a waitlist at place like adafruit and you'll get one soon.
I've stopped shopping for ARM boards when I realized not all docker images support ARM architecture making it hard to install apps that I might want to.
Otoh, Odroid H3 is a quite capable x86 machine and these days parts are so cheap, I bought a quad core Odroid with 32GB mem and 1TB nvme for around $300 and figured I'd retire my cloud server that costs $50+/mo (pitiful 4GB memory) for a much better spec.
Ignoring that the article is just an ad for his store, the thing that shocked me was the Orange Pi 5 Plus which includes "onboard microphone". I would NEVER want an onboard microphone unless it was a device I had far more than average trust in. What a great way to reward security breaches!
Now that the rockchip cpu is very close to full upstream kernel support, how does the raspberry pi compare? Do they maintain their own kernel branch (considering they use a pretty uncommon/restricted Broadcom chip) or is it mostly upstreamed?
Do we know how easy it is to remove the microphone from the orange pi's motherboard? A microphone mixed with proprietary firmware is sort of a deal killer for me.
These ~1L sized computers idle in the 7-13W range. Even with EU electricity prices, it would take you a very, very long time to make up the difference from a 5W SBC.
You also get standard x86 support, normal expand-ability, M.2 slots, PCIe ports, etc. You lose GPIO support.
For $100-$120 on eBay, a search for 'dell optiplex 3070' typically results in getting either a i3-9100 or i5-9500, 8-16GB memory, and 120-256GB NVMe. I bought one a couple months ago for Blue Iris and it uses 8W at idle. Here's an example of a USFF (ultra small form factor) for $110 - https://www.ebay.com/itm/325826958401
Lenovo variant (M720q tiny) for $121 w/ i5-8400T, 16GB, and 256GB SSD (probably NVMe) - https://www.ebay.com/itm/266450621632