Hacker News Reader

> I remember the Asus Tinkerboard being the first clone board that I thought could _really_ make a dent... but then Asus basically put it on life support.

Right -- it had potential but it also had problems they were seemingly unwilling to fix.

> Pine64, BananaPi, Orange Pi, Radxa, et all often make some pretty decent hardware, but the fact that it's often as hard today to get my projects working on those boards as it was with all SBCs back in 2012-2014 means I still stick to Raspberry Pi-branded boards for most things.

This is something the Raspberry Pi Foundation people will stress when people snark in comments about specs when they launch a new machine. The total picture matters much more than specific, minor, things.

Even when it comes down to something like pi-gen, which is obscure but enormously useful for tinkering with distribution builders. Or the Raspberry Pi Imager app for desktops.

Eben Upton credits Eric Schmidt with providing the motivation to make computers even cheaper, rather than pursuing better and better Raspberry Pi machines, and in the grand scheme of things it's hard to ignore that Raspberry Pi is better when they do focus on choosing their own constraints with stuff like the 2040 and the Pi Zero 2W.

> This is something the Raspberry Pi Foundation people will stress when people snark in comments about specs when they launch a new machine. The total picture matters much more than specific, minor, things.

Total picture is that I can't still run mainline Linux on Rapsberry Pi 2B without losing a bunch of basic features, while I can run it on almost all my Orange Pis. This makes Raspberry Pi very annoying to use. I wanted to use Rpi 2B as a networked sound card for a room. Nope. No audio support mainline after like 7 years after release.

I'm talking about the foundation's total picture, obviously.

The Pi isn't mainstream hardware, is it?

Their total picture includes OS distribution and support that eclipses all the other mini-SBC-type ARM machines.

Ubuntu has a mainstream distribution supported by the Pi Imager, and it should support sound, if you have the right boot parameters set?

(edit: not sure this supports as far back as the 2B; I might have to test that)

The Pi is a lot less "weird" than a lot of the hardware the mainline kernel does support

Wanna run Linux on a Sega Dreamcast? (game console from 1998)

Well there's a dreamcast_defconfig in the mainline kernel still

https://github.com/torvalds/linux/blob/master/arch/sh/config...

How about a Nintendo 64 from 1996? Someone sat down and hacked that out too!

https://lore.kernel.org/linux-mips/20201225190503.1235321881...

Or for more recent examples. How about some ancient routers?

https://www.phoronix.com/scan.php?page=news_item&px=Linux-5....

These are bits of hardware added by hobbyists or small teams. The Raspberry Pi foundation is a company with paid developers supporting a set of devices you could count on one hand (or two if you include every variant separately)

The later ARM64 Raspberry Pi hardware is more standard, and you definitely can get most of the hardware working with a mainline Linux kernel. It looks like sound may be an outlier - but bcm2835_audio is in staging, and I2S is in the tree proper but that does require additional hardware.

Sure, but how many Orange Pi-like sbc's bother to advertise mainline kernel compatibility? That matters too if you want ease of support. The Raspberry Pi folks have a long history of working with the kernel dev community; their support is not just hacked together after-the-fact.

Weird question. Do you want an actual count?

> Weird question.

Rhetorical question, I think.

>I wanted to use Rpi 2B as a networked sound card for a room.

The analog sound output is a hack through 10bit pwm. It's better than nothing, but I wouldn't recommend building an audio related project around it.

I built a moderately successful audio project around it. We've got a few hundred Pi's as audio player appliances for our streaming platform in a hospitality niche in my country.

Yeah, Pi audio card leaves a lot to be desired, but mostly the power level of the output. The signal quality can be worked around somewhat with processing the audio beforehand (standard compression and EQ stuff, so that the signal doesn't need something that the card can't do).

However, there are sound cards you can put on top of Pi, such as https://www.hifiberry.com (which we use), without breaking the budget, and they sound great.

Edit to add: what I love about Pis is how reliable they are. We literally treat them as appliances and they end up in places that are really not optimial from either ventilation, dust or temperature aspect, and they very rarely break. Even SD cards are not as fragile as we've initially feared (we've minimized write cycles but there is some of that).

Before the Pi I'd bought a number of BeagleBoards over the years.

I still have a few BeagleBoard xMs and original BeagleBones sitting around. The bone specifically was super neat in that it had a ton of IO (PWM, A2D, encoder inputs, SPI, UART, etc.) that could be controlled by either the CPU or microcontrollers embedded in the SoC.

Problem is at this point I have no idea what to do with any of them. ARM CPUs got so much faster in such a short span of time that it feels like arm boards from a few years ago are kind of just junk today. There are many dirt cheap boards that are way faster than a 1 GHz Cortex-A8 (or the 700 MHz armv6 core in the OG Pi) available today.

I've never owned a Beagle. I discovered earlier this year that they were open hardware and even had real-time capabilities. It struck me that this is what the Pi should have been.

Then I saw the price. So I'll be sticking with Pis.

It would be a master-stroke if the Pi gained real-time capabilities.

The BeagleBone Black was $45 at launch (10 years ago, ouch), which was $10 more than the Pi B. At this point they certainly seem to have abandoned that price point.

The first BeagleBone (the 'White') was $85, but it had a JTAG debugger embedded in it so you could actually single step through the Linux kernel if you wanted.

I share that sentiment. I'm so sad that the Beagle project doesn't come out with a CPU refresh.

The BBAI doesn't count?

Sort of, but they tripled the price to take the CPU cores from 2010 (single A8) to 2012 (dual A15). I know the whole point of the BB is the real-time capabilities and the two DSP cores, but I do wish the application CPUs were faster. Now that there's compute support for the GPU in the Pi4, I wonder what the performance difference is (in AI stuff).

That's all fair. I love that the BB platform is so stable, I can design around it with reasonable confidence that my gizmo will still work in _another_ decade, which is a perspective only recently coming to the Pi world. But there are certainly sacrifices made in that approach.

Beaglebone AI was released: April 23, 2013.

Considering we're in early 2022, no. I wouldn't call that a refresh.

That's a typo, where did you find it? The BBAI announcement was February of 2019, and shipped in September of that year.

You are right, I was wrong. My apologies.

It looks like the date I quoted was the release date for an older variant of the beagle.

Pine is in a league of its own by now, with the RK3399 and some of the other chips used in their boards having pretty fleshed out mainline support. I would personally consider them better options than a Raspberry Pi just based on software options. You can slap postmarketOS on a handful of Pine stuff now and get very up-to-date stuff.

I like PINE64's stuff, too.

Armbian is another good choice, for people who prefer Debian based OS.

I was worried early on about availability. They started with the goal of getting into classrooms, being cheap so kids or low-income schools can afford them, but then you also had for-profit businesses buying hundreds of these to power kiosks and digital signage. I'd almost describe it as "taking advantage" but perhaps that's too harsh. I was a little put off at first that these educational tools were being snatched up by businesses trying to save money.

I suspect that was an initial concern too, until they came to terms with the idea that e.g. Cisco's requirements in a Raspberry Pi were both achievable and also help underwrite production of more imaginative things.

> I was a little put off at first that these educational tools were being snatched up by businesses trying to save money.

I think the RPi's attraction for commercial users is less the price point and more the size of the community around it. It's by far the most approachable system to work with when you need a small linux device with cameras and gpio.

> I'd almost describe it as "taking advantage" but perhaps that's too harsh.

It is, especially when you consider economies of scale with things like computer production.

Building 50 boards is expensive. Building 500 is cheaper. Building 5 million? That's now custom equipment dedicated to the project, and costs just plummet per unit.

I also like the Pi4 model in which the "higher end" versions seem to be more profitable, subsidizing the lower cost/RAM models.

Yeah the fact is that the more demand there is for it the cheaper they can make it, so everybody wins in the end.

Well except the people living around the place where the raw materials are extracted, but that's another story.

The RPi would be viable in my work at any price lower than the cheapest Intel NUC. We’ve bought 800 since November. We’ve decimated our Windows licensing costs, power use, MTBI, and shrunk our attack surface considerably.

There is really no competitor that has even been a consideration for us. Software stack is the hardest to get right with them at scale for fleet management IMO.

Do you mean that your company has switched 800 employees from Windows machines to RPi? Are the 800 people non-engineering? Are the RPis thin clients? Everything running in the cloud?

I was thinking servers.

If they'd taken the same path OLPC did there would not be a Ten Years of Raspberry Pi.