I don't know why people love Pi-s so much. They filled a niche, once, years ago, and were quickly outcompeted not even a year and a half after.
Get some old i7 or Ryzen, get a big case, put 12-18 HDDs, spend a little extra on quality cooling solution if you have the server in your bedroom / living room, install modern Linux, tinker to your heart's content.
> I don't know why people love Pi-s so much. They filled a niche, once, years ago, and were quickly outcompeted not even a year and a half after.
They still fill a niche for me, just not a server niche. The easy-to-access GPIO in a close-to-vanilla Linux system really doesn't have a competitor at its price point. For a fourth grade science project last winter, I had a pi 4 already (but it'd have been about $40 at my local microcenter if I hadn't). We were able to source a few $2 sensors off Amazon. I showed her how to look up the pinouts, figure out which GPIO pin to connect the dupont connectors to, and helped her write a python program to log the data from the sensors to a spreadsheet. She had fun with it, learned some stuff, and it really sparked her interest.
I don't think anyone has outcompeted them in accessibility for that kind of tinkering and learning. Or, if they have, they haven't caught my attention yet, and I've usually got my eyes open for that kind of thing.
Ah, education, right. I never had interest in the whole GPIO thing but I'll admit life has been pulling me in very different directions, hence this dropped off my radar. Thanks for the reminder.
Thing is, I was aiming at servers. I've read many HN comments where people adore a Pi for some reason that I just can't see; they have to install custom kernels, get Pi hats, do some extra cabling, 3D-print cases, mount small (or big) fans, and all that.
And don't get me wrong, I love tinkering myself but after reading people's experiences for a while I just thought to myself "Why all this trouble? Get a $250 - $400 mini PC off of Amazon / eBay / AliExpress and put a 2-4 TB NVMe SSD and you have something 20x more powerful and with 100x the storage space".
Again, I love me some tinkering. But nowadays I want to get something out of it in the end. Like the mini PC I bought that I want to dedicate only to a PiHole even if it's a 50x overkill for it. Might add some firewalling / VLAN management capabilities to it down the line.
So yep, for education RPi and Arduino (+ its derivatives) seem mostly unbeaten.
On a RPi I can control more aspects than I can a mini pc ITX board. I can boot straight to my program. I can write directly to frame buffers. I don’t need Linux. I don’t really need a kernel…
Here are some examples of where an RPi outshines a mini-PC (though one can still achieve the same results, just putting the box outside the box):
Coffee table Digital Touch map.
Weather Station powered by a solar panel and a LiPo battery.
ADSB receiver also powered by solar and a battery.
Arcade Cabinet that sits on a bar top with a bill reader.
Mini JukeBox at the local hacker space.
Sailing autopilot using NMEA2000 connectors.
Wearables.
Playing with high density distributed computing. (More than 5 machines)
As I understand it, that is incorrect. It's more like a hypervisor still running in the background. That may have rolled back more and more with recent versions, but in principle the hypervisor can show shit like temperature, voltages and frequency in wrong ways, or delayed to the 'guest', whatever that may be. Actually that was the case, for some time, as some low-level tinkerers and/or overclockers discovered.
Similar to anything running in SMM (System Management Mode) on X86/AMD64 since the times of the 386SL. Be it BIOS, APM, ACPI, UEFI, or whatever.
I repeat: "It's just an illoooshn..." (There is now raw iron/silicon for consumers)
For homelabs, yer you can get something much better for much less.
For use cases where consistency and future support is key (education and industry) you really can't beat a Raspberry Pi. Their hardware and software support is top class. The first Raspberry Pi is still supported by the latest version of their OS over a decade later and it's even still being manufactured.
For all their products they commit to long term availability. For example, the Pi 5 will be in active production until at least January 2036 (assuming the company itself exists of course).
For anyone with a fleet of these, that's an amazing commitment. It means that when a piece of hardware breaks you can buy a band new but identical piece of hardware to replace it.
For most other companies you'd need to buy a different piece of hardware. Yes, the specs would be better, but now you have a fleet with mixed hardware which _you_ need to support and maintain going forwards.
Oh, I see. It's about fleets of easy-to-manage / predictable-to-support machines. That's valid, thanks for making me aware.
And indeed I was wondering about homelabs. RPis were never good there, not even when they got out for the first time. The form factor is what won over people back then. Feature- and speed-wise they were always mostly substandard. Not to mention Linux kernel support and driver issues (that might have been fixed since the last time I looked, admittedly).
And I agree on the fleet thing. Best if you can flash an SD card, drive to the spot in meatspace, pluck away the broken RPi, plug the new one in, wait for boot, test, drive away. Heard people doing that with RPis and others.
Userspace-accessible GPIOs, I2C, SPI, PCM, and UART on a system that runs Linux. My employer uses them for a bunch of our hardware-in-the-loop test automation, with the GPIOs used for CAN, relays for switching various signals, vibration table control, etc. The USB gets used for SCPI device control (power supply, multimeter, etc.) and DuT connection. It's a lot cheaper to use a Pi for this than it is to use a small form factor x86 machine with a bunch of USB-<protocol> dongles.
power draw. running 24/7 it makes a huge difference in overall power usage, and by switching from a repurposed desktop mobo & cpu, to a dedicated low power saved me thousands in electricity costs every year.
that fact that in EU power isn't cheap... is also a main reason for keeping total power draw as low as possible.
Valid, thanks. But to what degree? The light bulb that runs 18h a day in the kitchen likely draws the same power that my mini form factor Optiplex 3060 does.
To me arguments like "2W vs 10W" are fairly meaningless.
I am much more concerned about data center power usages, especially in the age of LLMs.
Like that ancient German teacher I had that kept preaching we should stop using electric kettles because it's bad for the planet. While the 3 plants in her hometown amounted to ~83% of all power usage and ~92% or all pollution. Boy, was she unhappy when I did that research and pointed it out to her.
Pi-s / SBCs are I suppose very good for computing out there in the meatspace, where you might need a battery because sometimes power stops for 6 hours? Could be that.
She did not offer any alternatives. That was also a very funny element to her preaching. She saw a class of students and thought she can signal her virtues.
She was, shall we say, disappointed with the response.
Because for $40 I have a system that runs at a decent speed.
For $300 I could get an ITX to run.
So for the cost of an ITX, I could run a dozen RPIs. Who wants to have a server running in their bedroom? Have you heard the noise those things make? Sorry, no.
A “server” doesn’t need to mean a pizza box with 15k rpm jet engine fans.
My server is repurposed desktop hardware in a desktop tower case and is nearly silent except for the subtle hard drive noises. The hardware cost next to nothing and is far faster and more capable than any pi (except the pio of course which wouldn’t be used anyway).
You’re running the pi and drives in a plastic take away container off usb power for that price.
At the very least you want the case and psu. At which point the question is which cpu+motherboard+ram combo do you want in that case. The rpi is one of many such options and is actually quite expensive for the amount of cpu+ram you get for the price.
An ITX isn’t the competitor for a Pi. I’d suggest a USFF prebuilt. I use an HP Elitedesk and Dell and Lenovo each have similar tiny PCs. They’re nearly silent or completely silent, and half the size of a Mac Mini. Cost is about $150 for hardware that is more than enough for me, plus they can have 1-2 SSDs and a hard drive inside the case.
Clarification: They're about half the height of the OLD Mac Mini. Better comparison: They're the size of a typical hardcover book if you chopped it to be square.
I'm uncertain of why $40 vs $300 is even a point of debate on HN. The latter is a one-time investment and you likely can expand it a bit i.e. add a 2.5" or M.2 drive later.
What's the gain of running 12 RPi, exactly? Do you do research work requiring distributed low-cost computing?
They are if the GPU can be attached. I avoid virtual machines in favor of container workloads from containerd for this reason. It’s easier to attach Mali GPU and do my work than it is to find cash in this economy for a dozen RTX’s.
It doesn't even have to be a Pi though, just look at competing NAS solutions that have hit the market since Synology peaked in popularity.
Why am I spending more on a Synology versus something like a UGREEN NAS and just flashing a wide selection of NAS/home cloud operating systems on it? Synology's customer base certainly has the technical know-how to accomplish that.
I've got an RPi 4 with a Samsung 990 EVO Plus 1 TB NVME SSD in an external USB-C enclosure connected to one of the Pi's USB 3.0 ports, and get 280 MB/s.
I would have expected going to an RPi 4 with an NVME SSD not going through USB to do a lot more than just boost storage speed by 80%. I had been thinking of getting an RPi 5 and moving my RPi 4 stuff to the 5, freeing the 4 to replace the 3 that is current running Home Assistant, but for what I'm doing on the 4 I'm no longer sure the 5 would actually give much noticeable performance improvement. It may be better to simply get another 4 to replace the 3.
I guess this is a side note personally don’t think any of the Raspberry Pi hardware is worth it unless you are using the GPIO pins or any of those not-NAS not-PC type of functionality the Pi offers. I think for general compute it’s hard to make it make sense.
I think there are a whole lot of mini PC type of solutions that just make more overall sense.
Around 200-270MiB/s is what has been publicly benched. I’m sure there’s someone squeezing 300 out of one.
The PCIe bus in an RPI is Gen 2 so it’s not that fast. The point isn’t whether an RPI is a Synology device. The point is there are other ways of having a cheap NAS other than Synology.
Hell, a Beelink with an external USB 3.0 HDD rack would also do just fine.
VScode has a bit of a history now of quickly deprecating competitors who innovate in this space. It already has good options for code completion, AI chat bots, and more features on the horizon. I'm not sure what cursors moat is. Seems to me like Microsoft could easily implement any new feature cursor comes up with.
That is why most forks lose in the end, be it egcs, deno, NoSQL or cursor, the leaders eventually integrate the key features that have made them a differentiatior, thus people don't have to change, lose the investement they already had on the existing tool, and get those features as well.
For me, the best kind of "moat" (tbh I hate that word, since it specifically implies needing to design (...scheme...) and engineer some kind of user lock-in, which is inherently user-hostile) would be staying aggressively on the forefront of DX. More important than feature churn, making it polished and seamless and keeping a smile on my face as I work is the best kind of "moat."
It requires constant attention and vigilance, but that's better for everyone than having some kind of "moat" that lets them start coasting or worse— lets them start diverting focus to features that are relevant for their enterprise sales team but not for developers using the software.
Companies really should have to stay competitive on features and developer happiness. A moat by definition is anti-competitive.
Wow. What an intelligent take. I would have never expected this from Ben Affleck. He seems extremely familiar with the technology and it's capabilities and limits.
And what exactly makes "Robert Graham" such an expert in this particular domain? I don't know who this person is or why I should trust their personal blog over the NYT. The article itself is rather hand-wavy in it's assessment of the report. The thesis is essentially "bot farms use lots of sims & this is an example of using lots of sims, therefore this is a bot farm and not espionage."
BlackICE was a big personal firewall 20 or so years ago - you can read all the CNet/ZDNet reviews if you search for it. You can also look at his code (for a port scanner that can scan the entire Internet in 5 minutes, whew) on GitHub:
Thank you for sharing. I recall blackice. I'm not seeing anything here would lead me to believe he's an expert in this particular domain though, which is more about nation state intelligence operations than it is anything technical.
I think his point is that it's not about nation-state intelligence operations, and that the capabilities claimed here are garden-variety cybercriminal operations. You or I could set up something very similar, if we were willing to participate in a dodgy business.
And by some basic napkin math and a few Google searches, he appears to be right. Prepaid sim cards are about $5/each [1]. A 16-port SimBerry server is $499 [2]; their full-fledged servers are "contact us" for pricing, but support up to 18,000 SIM cards [3]. Assuming their enterprise solutions are cheaper on a per-SIM basis than retail, that's about $35/SIM in hardware costs. For $100K in startup capital, you can run a 3000-SIM farm. And then, like this article suggests, once you get started you reinvest the profits: if you assume each SIM card gives you 1000 txts, then if you charge 2c/txt your $5 investment becomes $20 and you can expand your operations accordingly.
I wonder sometimes if, when it comes to cybercrime, "[Russia/North Korea/China/Iran] did it!" is actually code for "The FBI has no idea who did it, but if we said that it would encourage all sorts of script kiddies to do this for profit, so we might as well blame it on our nation-state level adversaries." Many of the hacks in question (eg. ransomware) are not out of reach of a lone malcontent in their 20s with some tech skills.
I would gladly accept a bit of a slow down on progress if it meant my contributions to society were more meaningful. Additionally, I strongly believe this continuous dwindling of small organizations has resulted in an overall loss of community and a sense of belonging. In my opinion, this is what's causing the overall decline in health that we're seeing in developed nations.
For many, life seems aimless. Your future is to simply contribute what you're told to some faceless multinational for which after 20 years your only recognition will be a small piece of canvas with a mass produced screen printed design.
reply