logoalt Hacker News

throwaway_new_1 • yesterday at 3:53 PM • 10 replies • view on HN

The goal of raspberry pi's was to have a standard GPIO connected to a moderately powerful and cheap computer for educational purposes. Then it slowly evolved into many use cases because of it's vast community and standardization and availability, not to mention the large number of boards that connect to it. The standardization did erode a little but it still maintains that reputation because of the community and availability and now because its turned into a big brand. Many industrial projects depend on it.

If you use the raspberry pi for homeservers or general compute (which I did/do sometimes) then the value proposition is quite poor regardless of pre AI prices unless you have very strict power requirements (in which case it probably fits the above use case already). The comparison with NUCs or mini PCs or laptop boards feels like it's a wrong comparison.

The chips used in PIs are repurposed Broadcom SOCs that have inferior bus speeds compared to a general purpose processor from intel or amd, even few generations old. For any general purpose computing these chips will always be slower. its like comparing apples and oranges. I never really understood why people did that.


Replies

deng • yesterday at 5:02 PM

> its like comparing apples and oranges.

Yeah, I don't disagree, but I would argue the Raspberry was always an applorange, and that's exactly what made it popular (together with the cheap price in the beginning, which made it basically an impulse buy). It brought people into microcontroller/embedded systems territory without having to actually deal with all of the stuff that made this so daunting. As you said, it is basically a normal Linux PC, just that it runs Arm and has GPIO. People who are really into electronics/embedded systems will use a proper microcontroller nowadays anyway, these have gotten so powerful and easy to use that there's really no reason for a Pi anymore. And on the other side, people who want a little server, a pi-hole, tailscale or octoprint or whatever, can just use an old used laptop/miniPC or similar. Since the Pi has gotten so expensive, I see basically no reason to get one anymore.

➕ show 1 reply
derefr • yesterday at 4:56 PM

I would argue that even if you need "a cheap SBC with GPIOs", at this point you'd be better served by connecting a tiny $5 microcontroller-powered board with a USB PHY + its own GPIO headers, to any cheap "computer" (i.e. a GPIO-less SBC, or an old laptop, or a phone like in the article, etc.) over USB; where the MCU board essentially serves the host a "GPIO gadget."

You can make such a dongle yourself from ~any MCU dev board, but there are also productized "USB GPIO breakout" adapter boards, e.g. the Adafruit MCP2221A.

The "Pi as an SBC with GPIOs" case now only makes sense if you really, really require the Pi's compactness and/or monolithic-ness (as might be the case for e.g. ruggedized embedded systems like robots.)

➕ show 1 reply
criemen • yesterday at 5:05 PM

I use a pi (4th? generation) to run homeassistant, calendar sync target and a few other random small services on. It's great. It works, it doesn't make noise, it's small, it plugs into USB-C, the power brick is small, too. The SSD I'm storing everything on because I had a few failed SD cards has a larger footprint than the rest of the computer. What not to like? Nothing I run on it requires faster CPU, and the small footprint is worth something to me. Sure, if you have a big house with a dedicated server room/hobby space, you might as well have a NUC/old laptop/whatever.

Rohansi • yesterday at 3:59 PM

> The comparison with NUCs or mini PCs or laptop boards feels like it's a wrong comparison.

For many (maybe even most) people they do the same thing. Some just want a HTPC, basic desktop, or small server, no GPIO needed. Raspberry Pis were definitely cheaper early on because you could actually buy boards at or near MSRP, but over time the price rose and mini PCs became cheaper.

➕ show 1 reply
Aurornis • yesterday at 6:17 PM

> The goal of raspberry pi's was to have a standard GPIO connected to a moderately powerful and cheap computer for educational purposes.

At first it was a blessing that the Raspberry Pi gained popularity as a general purpose computer, not just for education and embedded applications. The additional demand and popularity brought more money and development attention to the ecosystem.

At some point, which I could never really put my finger on, this general purpose computer popularity grew so large that the embedded systems and educational aspects of Raspberry Pi seemed almost forgotten by the mainstream users. At a local in-person meetup group it was a running theme for people to brag about how many Raspberry Pis they owned for different purposes and to boast about using them for applications where they didn’t really fit. In person I’ve heard more about Raspberry Pi NAS systems (where people are always disappointed with the performance and the price of forcing a Pi into this role) or Raspberry Pi Kubernetes clusters (which mostly sat idle and could have been replaced by a single PC with more RAM).

This reached such a critical mass that it almost seemed like the mainstream discussion forgot the original purpose of the Pi and thought it was literally just another computer, but small. Most people who tried to use it for this purpose were left slightly frustrated by how it didn’t perform like they wanted a computer to perform.

So now every discussion about Raspberry Pi attracts comments about how nobody should buy a Pi when you can buy a NUC or laptop or some other used piece of hardware and it’s so much “better”. This is the right advice for a lot of people who wanted a powerful general computer but bought a Pi instead, but it’s also furthering the misunderstanding of what the Pi is for.

The real advice is that everyone should be shopping for the right tool for the job. There are a lot of cool applications for the Pi where a NUC or old laptop or phone doesn’t make sense because you actually need the compact size, the low-level I/O, or some other feature. However anyone who can put a NUC or laptop side by side with a Raspberry Pi and consider them both viable options for their project probably shouldn’t have been looking at a single board computer to begin with.

There is a parallel situation happening with the “just buy an ESP32” arguments. If an ESP32 solves your problem, you also probably shouldn’t have been in the market for a Raspberry Pi.

➕ show 1 reply
dapperdrake • yesterday at 5:04 PM

It combined cheap GPIO for $35 instead of $500 with HDMI, tiny size, full Python instead of MicroPython, passive cooling and Linux.

Suddenly the Kodi and the GPIO crowd (Arduino, ESP8266, ESP32, etc.) could work with the same board.

A new era, like the IBM PC or the iPhone. Changed computing forever.

EDIT: like other commenters noted, it seems like it made the intel NUC a reality.

numpad0 • yesterday at 8:04 PM

The GPIO in Pi is almost useless. Nobody runs servos off of those. Pis are mostly used as just Linux PCs with USB ports that reliably power itself on.

IMO, the problem with the Pi GPIO is that it's a collection of GPIO pins and low pin count buses for controlling peripherals, rather than a singular bus. The fact that it runs Linux without an RTOS monitor program or a sidecar of sorts adds to the problem.

mathgeek • yesterday at 4:32 PM

Much like the goal of the PS3 was to have a powerful gaming console that also let you run Linux. The market quickly deals with any hardware that is cheaper than its value.

stackghost • yesterday at 6:06 PM

For me the value prop is a low-power device that can be passively cooled but has a full Linux and tcp/ip stack, at a price point where I don’t care if I accidentally wreck it.

Fantastic for one-off projects.

Nowadays they require cooling fans and cost absurd amounts considering the performance so there seems little point.

Most of my one off projects these days are now on microcontrollers.

➕ show 1 reply
TheRealPomax • yesterday at 4:43 PM

writing a USB driver takes a fraction of the effort it used to these days. GPIO no longer holds quite as much value as it used to.