A Raspberry Pi 5 is better than two Pi 4S
hackaday.com
hackaday.com
Isn't this mostly false? The Pi4 uses 15W at 5V3A. The standard says, to exceed 15W you should increase the voltage, but Pi5 increased the amperage instead. 5V5A USB-PD supplies are almost unheard of.
I wonder how much a 9V3A -> 5V5A buck converter dongle would cost.
https://www.anandtech.com/show/16712/usbc-power-delivery-hit...
> USB-PD R3.1 supports three charging models:
> - Fixed voltage
> - Programmable power supply (PPS), and
> - Adjustable voltage supply (AVS).
> In the fixed voltage scheme, the Standard Power Range (SPR) mode supports 3A and 5A at 5V, 9V, 15V, and 20V.
Also keep in mind the existence of PPS as another data point. I have an Anker charger sitting in front of me that offers 3.3V-11V at 5A. 5V at 5A falls squarely in that range, as long as you have an e-marked cable that can support 5A. (I have no idea if the Pi 5 supports negotiating PPS, but it would increase compatibility with chargers if it does.)
I agree the Pi 5 should have included a buck converter or something, but I don’t think it’s correct to say they’re not standards compliant. It’s just not a common use of the standard.
https://en.wikipedia.org/wiki/USB_hardware#USB_Power_Deliver...
5v3A is the limit of USB PD SPR.
Here is another industry source which has a table that also says 5V at 5A: https://www.graniteriverlabs.com/en-us/technical-blog/usb-po...
In fact, using Wikipedia’s sources, I found the actual specs and here is another piece of supporting evidence for these other articles:
> The Fixed PDOs Maximum Current field Shall advertise at least 3A, but May advertise up to RoundUp (PDP/Voltage) to the nearest 10mA. Requires a 5A cable if over 3A is advertised.[0]
This is a footnote attached to 5V3A, indicating to me that chargers are allowed to offer more than 3A at 5V.
At a minimum, the spec seems to be ambiguously written, but I’ve only spent a few minutes skimming it. Multiple industry sources (previously linked) believe that 5V@5A is within the spec, even without using PPS.
[0]: “USB_PD_R3_1 V1.8 2023-04”
The first revision, rev 1.0, had six fixed power profiles: 5V/2A, 12V/1.5A, 12V/3.0A, 12V/5.0A, 20V/3.0A and 20V/5.0A. These were deprecated by rev 2.0, version 1.2, which instead introduced power rules at four fixed voltages, supporting power supplies with different output powers. The maximum supported currents by the specification were 5V/3A, 9V/3A, 15V/3A and 20V/5A. These power rules were retroactively named the Standard Power Range (SPR) in rev 3.1. That revision also added the Extended Power Range (EPR), which raised all the current limits up to 5A, but only when used with electronically marked cables. EPR also added power rules at 28V, 36V and 48V, again requiring (differently) marked cables and up to 5A. So 5V/5A is a valid option, but only when used with a USB PD rev 3.1 compatible power supply and cable.
Orthogonally to the fixed power rules, rev 3.0 introduced the Programmable Power Supply, which allows a configurable voltage between 3.3V and 21V in steps of 20 mV. This was extended by rev 3.1 with voltages between 15V and 48V in steps of 100 mV, called the Adjustable Voltage Supply. As far as I'm aware this is not commonly used, and most products advertising Power Delivery support use the predefined power rules.
The reality is that with variable and programmable power supplies, it’s now up to the device to draw power at a balanced rate from the supply, instead of the “supply” pushing power in. Modern tech has allowed us to have supplies between 3V-48V, 500ma to 10A. All with a usbc head.
What good is a universal port if it's not actually universal? I would actually prefer different shapes, because then you could be pretty confident that if everything fit it was going to work. Like how USB used to be.
It’s such a mess that the EU took one look at it and said “let’s make that a universal standard and force it upon everyone!” Modern USB truly is the government bureaucracy of standards.
I believe there were discussions about potentially standardizing on Lighting if Apple relinquished it's tight control over the technology but they were not interested in that path.
The EU decided that Apple cannot dictate what cables are used to charge their phones to the degree of at one point charging manufacturers $4 per cable (to be fair it seems like Apple was providing the hardware so it wasn't a pure licensing charge).
If you're curious, my main objection to the EU is that I don't think USB-C is a good standard and I don't like the prospect of the whole industry stuck with USB-C longer than appropriate because there's no incentive to experiment and possibly come up with something dramatically better.
Having a plug on say a laptop that can be a charger, a HID, a drive, a monitor, or all of the above with a docking station is vastly superior to the old day IMHO.
Unfortunately the cable to do that isn't cheap and is massively overkill for say a phone charger. Thus cable compatibility becomes a thing but looking at HDMI cable compatibility is always a thing.
I think the problem is what used to be complaints about niche connectors are complaints about USB-C and it feels different and more impactful even if it isn't necessarily.
After all I know back in the Mini-USB days you had manufacturers skipping the data pins to save some copper leading to frustration when you couldn't transfer files.
Besides Thunderbolt cables, there is not a standard nor universal way to identity them. I had to test each cable by plug in an SSD then speed test the SSD.
Flash back 15 years, a cable with MicroUSB either transfer data, or not. The difference between high speed / super speed only lives in the device.
A USB-C docking station will plug into many of my laptops, but only two will display anything on the screen.
Being able to pick and choose features like that is not user friendly. It will do everything except when they didn't feel like implementing it is not a good standard.
That's actually not a bad solution. The speed of a cable doesn't matter unless you have a device that needs to be fast.
At least you get minimal functionality (why are you hoarding data less cables?) Better than having to go buy one because you have a billion micro USB but no FireWire to hook up that drive.
The baseline profiles for power delivery (i.e. anything up to 20V@5A) have been all but done for two, three years now - you can readily buy combinations of PD and MUX chipsets that handle everything you can throw at them, and Anker has high-quality chargers to supply the juice.
The thing where USB-C still has issues is alternate modes. USBx usually works on all ports of a laptop or PC and virtually all cables, but anything involving TB, display, debug or audio is a hit-and-miss given how incredibly difficult it is to route all these high-frequency signals and properly mux them.
The simple fact that we can finally share charging cables across phones is a god send, not having a single cable for all phones was the stupidest thing people had ever done.
They go into the trade-off in one of the launch videos. Apparently if you've only got a 3A supply then the downstream USB power gets limited but otherwise it works.
Even if the power adapter supplies it, without a high quality cable there's enough voltage drop to trigger the low voltage warning.
I recommend the official supply or eventually the official new PoE+ HAT for now.
As you say, there are a ton of "minipcs" on the market that directly compete with the Raspberry Pi on cost and power usage. They're typically slightly larger but the expansion options (bring your own RAM/storage) plus real I/O (with real PCIe), disk, etc IMO significantly outweighs this. They're also typically more performant and while aarch64 platform support is increasing dramatically there are still the occasions where there's a project, docker container, etc that doesn't support it.
Taking it a step further, there are a TON of decommissioned/recycled corporate/enterprise SFF desktops on the market. They don't compete in terms of size (13" x 15" or so) but they can actually get close in power usage. Many of them have multiple SATA ports, real NVMe, multiple real half-height PCIe slots, significantly better USB and PCIe bandwidth, etc.
With my project Willow and Willow Inference Server[0] we're trying to drive this approach in the self-hosting community with an initial emphasis on Home Assistant. They're generally sick of Raspberry PI supply shortages, very limited performance, poor I/O, flaky SD cards, etc. The Raspberry Pi is still pretty popular for "my first Home Assistant" but generally once people get bitten by the self-hosting bug they end up looking more like homelab very quickly.
For Willow particularly we emphasize use of GPUs because a voice assistant can't be waiting > 10 seconds to do speech recognition and speech synthesis. There are approaches out there trying to kind of get something working using Whisper tiny but in our ample internal testing and community feedback we feel that Whisper small is the bare minimum for voice assistant tasks, with many users going all out and using Whisper large-v2 at beam size 5. With GPU it's still so fast it doesn't really matter.
The Raspberry Pi is especially poorly suited for this use case (and even amd64). We have some benchmarks here[1]. TLDR a ~seven year old Tesla P4 (single slot, slot power only, half-height, used for $70) does speech recognition 87x faster, with the multiple increasing for more complex models and longer speech segments. A 3.8 second voice command takes 586ms on the Tesla P4 and 51 seconds on the Raspberry Pi 4. Even with the Pi 5 being twice as fast that's still 25 seconds, which is completely unusable. Not fair to compare GPU to Raspberry Pi but consider the economics and practicality...
You can get an SFF desktop and Tesla P4 from eBay for $200 shipped to your door. It will idle (with GPU and models loaded) at ~30 watts. The CPU, RAM, disk (NVMe), I/O, etc will walk all over a Raspberry Pi anything. Add the GPU and obviously it's not even close - you end up with a machine that can easily do 10x-100x what a Raspberry Pi can do for 2x the cost and power usage. You can even throw a 2.5gb Ethernet card in another slot for $20 and replace your router if you want to go really dense.
Even factoring in power usage (10-15w vs 30, 2-3x) the cost difference comes down to nearly nothing and for many users this configuration is essentially future-proof to anything they may want to do for many years (my system with everything running maxes out around 50% of one core). Many also gradually grew their self-hosted situation over the years with people ending up with three or more Raspberry Pis for different tasks (PiHole, Home Assistant, Plex, etc). At this point the SFF configuration starts to pull far head in every way including power usage.
Users were initially very skeptical to GPU use, likely from taking their experience in the desktop market and assuming things like "300 watt power usage with a huge > $500 card". Now they love having a GPU around for Willow and miscellaneous other CUDA tasks like encoding/decoding/transcoding with Plex/Jellyfin, accelerated Frigate, and all kinds of other applications. Willow Inference Server (depending on configuration) uses somewhere between 1-4GB of VRAM so with an 8GB VRAM card that leaves for plenty of additional tasks. We even have users who started with the Tesla P4 and then got the LLM bug and figured out how to get an RTX 3090 working with their setup which also of course leads to absurd performance with Willow - my local RTX 3090 goes from end of speech to command completion in HA to TTS feedback in ~250ms. It's "speak, blink, done" fast.
[0] - https://heywillow.io/
[1] - https://heywillow.io/components/willow-inference-server/#ben...
Another instance of using the right tool for the job.
That said, the link you provided is potentially a great add-on for someone who wants/needs the logic available on a real Linux host!
Thanks!
The initial release of the BOX-3 was essentially a pre-production run with ESP-BOX similar 3D printed plastics.
The full production run of the BOX-3 from Espressif with proper injection molded plastics should become available from a retailer/distributor near you within the next couple of weeks.
The issue (among others) is we achieve the speech recognition performance we do largely thanks to ctranslate2[0]. They've gone on the record saying that they essentially have no interest in ROCm[1].
Of course with open source anything is possible but we see this as being one of several fundamental issues in supporting AMD GPGPU hardware.
But there were other times where ARM were an issue too. I just don't want to (and sometimes can't) compile things.
I do like the pi for the long term form factor as bad as I think it is.
I had mine running for a bit with long SATA cables snaked out of the case to a couple 3.5 drives in a makeshift enclosure, but SATA/NVMe drives have gotten so cheap, it calls into question the need for all the power.
The Federal minimum wage has not changed since 2009, but the CPI captures effects like per-state minimums increasing, less people working minimum wage jobs, etc. No "adjust for inflation" calculation will capture the "pain" that every individual experiences from making a purchase, but this index is pretty close.
For more context:
Raspberry pi: April 2014
Pi Zero: Nov 2015 (1x ARM1176JZF-S @ 1 GHz, 512 MB RAM)
Pi Zero 1.3: May 2016 (now you can use cameras)
Pi Zero W: Feb 2017 (Wifi and bluetooth 4.1)
Pi Zero WH: Jan 2018 (omg, soldering the gpio pins? Much wow)
Pi Zero 2 W: Oct 2021 (4x ARM Cortex-A53 @ 1Ghz, still 512 MB, now bluetooth 4.2)
I'm not at all convinced they are caring about this market. Realistically there have only been 3 models and there really hasn't been much push into this area. The Zero 2 upgrade wasn't anywhere near the leap that the normal pis are making. I know there is more limitations, but they also have more competitors and it isn't like the zeros are sitting on shelves. There's till a good market for <$20 computers (and especially for a $5 one)
They are already kind of doing this by still producing and selling the old models, on the page for the 3b there is this obsolescence statement:
> Raspberry Pi 3 Model B will remain in production until at least January 2028
https://www.raspberrypi.com/products/raspberry-pi-3-model-b/
I can find the 3b for around 40€ new, under 30€ used, which for me is kind of ideal, I don’t have a lot of interest in the more expensive models.
That just means that you're seeing closer to the real price instead of subsidized price.
I'm holding out for the RISC-V boards. At least I'll be able to get real documentation unlike the RPi boards.
the main point for the price was to make it more accessible for kids, so that parents can buy one without thinking too much about the costs. the 1/2 GB may not support the desktop use cases that might be expected from the performance of the new pi.
i think the pi zero is now their main go-to device for the price-consious audience at this rate.
But for anything involving raw GPIO, you know, the reason you might want a true embedded board instead of a commodity PC, the pi5 has completely changed the architecture. The GPIO pins used to come straight off the BCM CPU with direct register access, now they're peripherals of the RP1 southbridge IC which is connected over PCIe.
Chief in my mind is the rpitx/pifm projects, which (ab)use the DMA pipeline to cram baseband-rate samples into the PWM registers and make a GPIO pin oscillate at radio frequencies. Connect a (bandpass filter, please! and a) piece of wire, and voila, the board becomes a radio transmitter.
This simply won't work on the pi5 without a complete rewrite of the core. Someone will have to figure out if the RP1 even offers similar capability, figure out if PCIe packet jitter will screw it up or if there's some buffering possible in the RP1, and rearchitect the whole project around it.
However, neither of those projects has seen much activity lately. In other words, it ain't likely to happen. Just keep using 3's and 4's for this.
For some stuff, pifmrds for instance, stereo processing really chews through CPU, and precludes running a station-automation package on the same iron. It would be really nice to have a faster CPU wrapped around the same transmit capability, but that's not what the pi5 is. If someone builds a station-in-a-box around this, it'll have a pi5 for the front-end but a pi4 as the actual exciter.
I’m doing the same with Proxmox Backup Server, Proxmox itself runs on a more powerful thin client ;)
For a general purpose computer though, the 5 is getting to the point where it is competitive for a lot of use cases where the 4 was a bit of a struggle. It sounds like I could definitely consider using a 5 as a mini-workstation for web browsing, some light dev work and the like, in those cases where the 4 felt a little clunky.
It's 12W under high load, but at equivalent load (in terms of output performance) to a 4B draws less power.
A Pi4 is still a perfectly fine microcomputer.
Pi 4s require cooling for many workloads (they run notoriously hot).
You're more likely to run into the needs-cooling envelope of a Pi 4.
I'm sure those who put a premium on thermals or quietness will be able to run RPi5 with passive cooling if they really want to. The RPi4 already suffered from thermal issues and that didn't pose a major problem. Anyone could simply drop by Amazon and pick one of the many passive cooling cases without any issue.
You heard wrong, then.
"For normal usage of your Raspberry Pi, adding cooling is entirely optional. The idle performance of a Raspberry Pi 4 and a Raspberry Pi 5 is about the same, and under typical loads Raspberry Pi 5 will run cooler than a similarly loaded Raspberry Pi 4."
"even when fully throttled, a Raspberry Pi 5 is still going to run faster than a Raspberry Pi 4!"
https://www.raspberrypi.com/news/heating-and-cooling-raspber...
A Pi 5 has a higher performance ceiling than the Pi 4 when passively cooled, and an even higher one with active cooling. The Pi 5 does not need cooling for most tasks.
https://www.youtube.com/watch?v=35_5wRKi_TI&t=12m20s
With a big passive cooler, temperatures were only marginally higher than with the stock cooler. (58.7 vs. 60.9℃)
But the R.Pi 4x series is peak R.Pi for me. In the last couple of years, NUC-class SFF devices with Intel-based chipsets have appeared [1] and they offer a much better performance profile (speed, power consumption) than the R.Pi 5. Some of these devices offer dual gigabit Ethernet.
[1] Liliputing is one site that I follow for related information.
I have a few NUC-style very small form factor Intel machines around too, of course, but there are a lot of good options out there and ARM shouldn't be discounted.
The real challenge is though that for most people, there's a huge market of secondary/resold business mini-pcs. These systems offer exceptional value. They're bulkier & idle close to 10w often, but if you're not doing battery powered or mobile robotics it's almost always a great trade-off.
My old Acer Chromebox units were even smaller and idled around 4w, and had two pretty beefy cores (alebit with low clock and small caches) and gobs of io. 6 USB3 ports, which it could easily saturate & barely break a sweat. Internal sata and mini-pcie wifi. Incredibly great value, and I was picking them up for under $100 years ago. Stuck 2*16gb ram in em. Amazing value.
If you need gpio, there are really really cheap USB based addon peripherals that offer whatever you might want and more.
It bothers me a lot how many people seem to support a market for junky piddling slow & limited celerons. The chips you cite have barely changed in almost a decade & were low end to start. I used a Chuwi 12.3 laptop with n3450 for a while & it was quite survivable sure. But there's so much more value going to resale markets or spending a little more & buying a Lenovo business 1L mini PC when it goes on sale. I wish Intel wasn't able to get away with selling such bad performance for so long.
Thankfully the new Alder Lake N replacement is worlds better. It's a great chip, and we can buy units at reasonable prices. I wish AMD competed here too but they have remained far up market.
Alder Lake stuff is showing up now, and that's pretty great. It is not, however, particularly fast hardware, and the GPUs are still really pokey.
I would love to see an RK3588S board for < $50, because the chip is incredible in terms of efficiency and feature set... but it's just so expensive still.
I've had bad luck with the Pi 4 with respect to video decoding. 4k HEVC decoding is an advertised feature of the Pi 4, but it chokes on anything with moderate bitrates in my experience. I'd like my next board to support HDR and HEVC so reliably that I never have to think about it, and I'm not sure if that's going to be the Pi 5 or not.
My contention is for things that actually need relatively-modern performance (taking into account that the RK3588 is still a little old, albeit Very Fresh for a SBC), I think that $20 is probably worth it for most of them.
Not that I don't have an Orange Pi 5 for other uses, but that's gathering dust in a drawer because I'm waiting for good GPU drivers. The Raspberry Pi 5 seems like it'll have Vulkan 1.2 out of the box. I'm still waiting for Panfrost's new PanCSF drivers for the RK3588 Malis.
A 30$ Le Potato is still more than the 25$ Raspberry Pi 3A+. Which I like for it's smaller size. A single USB port isn't too big a deal if you get a CSI Pi Camera for monitoring instead of USB, or use the GPIO for serial.
I see a lot of people who overspent on RPi4's, even, when they just could've used something with an H616 or whatever on it.
I am in that group now. Basically, my hobby development has moved to Raspi machines. I use a 400 as my primary station and various 3 and 4 models to do projects with.
One very nice thing is to package everything up with a Pi and have that project be self-contained. Dev tools, I/O whatever all nice and compact.
Start another project? Just setup another Pi and go.
Having a higher performance model means moving more of what I do onto a Pi. Things like VS Code run reasonably, but not great. On the 4, many things are close to great.
The 5 appears capable enough to change that equation.
Nice!
What I have seen happen is people basically get going on Raspian and start to like it. Myself, I have used UNIX flavors since SGI IRIX and I find Raspian simple, well supported and easy to use.
A fast Pi, plus a solid load of OSS packs quite a punch!
Dev tools for many languages.
Open, Libre Office and friends suddenly makes for a very cheap, effective desktop one can use to get a whole lot done.
Anyone inclined to work with OSS can grab one of these little Pi computers can do a lot. I use them that way frequently.
On the less snarky side... I'm sure it's a better system, but just because a core can run at 2.4GHz instead of 1.8Ghz, you have to take power dissipation into account. You have the same problem with overclocking: you can up the clock speed, but there's still only so much thermal energy you can remove from the packaging before you have to throttle the core to avoid melting the solder off the board or delaminating the PCB underneath the package. I'm sure the good people at Broadcom have taken this into account, I'm just commenting you can't just say "oh. it has a 33% faster clock, it's going to chunk 33% more data."
Ultimately though, at least IMO, the Pi isn't about being cheap, it's about being capable while staying a great value. I bet there's a lot of hidden complexity in the pricing, and it's very possible that they'd have to compromise significantly more than 12.5% of the functionality/performance to get a 12.5% price reduction (to $35).
Wonderful, this is long overdue.
I think as long as there are both 4 and 5, it is good, otherwise this is fantastic they are experimenting as they should, but there is value in being low-power and versatile in my view.
It does not. They’ve made a point of noting that it’s faster that a Pi 4 even if you run it without a heat sink and let it throttle the cpu back when it hits the temperature limit.
Really cool stuff in any case, glad to see this product continuing to get new features and spec bumps.
Or the Compute Module which can be even more compact than the standard Pi format.
The small form factor is really convenient for turning on and off.
Pis are designed for the edge case and that allows them to sit in drawers. If you design for the median only, then you’ll never get to median because people won’t be able to run and test the edge cases.
Of course as you go up costs rise and eventually you might as well buy a Mac mini or similar.
But pis are easy to build and manage and that’s the platform my kid likes.
I use a Raspberry Pi 4 to host a Nextcloud instance. It's sitting idle most of the time, but the weak CPU of the Pi 4 (combined with a fast connection and the fact that Nextcloud is a terrible pile of slow PHP) means that syncing lots of small files is actually CPU limited.
(Seriously, I set it up on the Pi 4 thinking that Nextcloud would be mostly IO limited. I wasn't considering the fact that its file syncing is based on WebDAV, and every single WebDAV operation is a separate HTTP request which goes to a freshly spun up PHP interpreter which has to go through all the work of connecting to databases and handle authentication and what not in order to write a 100 byte file to disk. Still, it works well enough for the small size and quiet of the Pi 4 is worth the trade-off.)
I don't plan to upgrade my rpi4 running homeassistant, a Plex video streaming server, a tailscale VPN exit node, and a few other minor services.
Plex on the rpi4 can't keep up transcoding 4k video to 1080p, maybe the rpi5 can, but I don't personally need that. Many Plex users do need 4k transcoding though.
Also, I have tried to use the pi4 as a desktop replacement, and I just found it too slow for the modern bloated world of webapps. All the little things just added up to too much and I gave up. The pi5 might be good enough though, but with all the peripherals and pains of arm64, a used old laptop is still probably better on the price/performance ratio.
What I've been trying to do is get it to chew through whisper.cpp at anything like a reasonable rate. It can so nearly do it with the right sized model, the latency isn't good enough yet, but the reason I'm doing this on a pi zero 2 is because of the USB-OTG port. I can take the whisper.cpp output, and have the pi pretend to be a USB keyboard, typing what it hears into a second machine over USB. This gives you dictation capabilities in environments where upgrading the PC it's plugged into isn't an option.
Like I say, it's not quite there yet in terms of usability, but it's close enough that I think the problem is just a matter of software (and arguably model) optimisation.
Moving to the Raspberry Pi 5 just isn't an option, because I'm already memory limited at 8 gigs, so 4 gigs would be unusable.
I bought my Orange Pi 5 with 16 gigs, so I'm quite looking forward to doing the upgrade! Granted, there might not be too many other people putting this much work on Pis, but when you're power constrained, they're an excellent option.
The Pi firmware comes as a binary blob, but has there been any effort to reverse engineer it?
[0]: https://datasheets.raspberrypi.com/bcm2711/bcm2711-periphera...
But in this interview https://youtu.be/vXYzJ1os4NA at 11:50 Eben says that the chipset does not support 16 GB.
This seems to defeat the whole purpose of a Raspberry Pi: cheap, almost disposable computer to tinker around with.
A 8GB RPi4 board is already more expensive than some used desktop computers, and until recently some major hardware manufacturers (ahem apple) were still selling high-end laptops with less than 16GB of RAM.
I understand asking for features such as PoE and SATA support, but 16GB of RAM feels like an attempt to make the RPi something that's not supposed to be.
They're awesome for prototyping, but there are so many places a Pi isn't the right fit or scale. They're common, not the only choice!
I've been running multiple 2GB Pi's and I can't think of any reason why I'd ever need more than 4GB. If I need that also want more IO (PCIe/NVME) and more CPU to make use of it.
The VideoCore VI has 8 QPUs at 500MHz producing 32 Gflop/s according to https://github.com/Idein/py-videocore6
So, VideoCore VII should produce about 86 Gflop/s
https://www.armbian.com/ is also pretty awesome and also supports bunch of other RaspberryPi clones for cheaper price.
But it’s weird enough that I would only use it when being jokey.
The s stands for sarcasm. /s
So pi^five clearly beats two pi^four. Three pi^4 would be close, though. If only we could just define pi as a nice simple integer...
I just always deprioritized doing it
I've built desktops in the past before, its more about disinterest in doing it at all for anything
the services that will do it move the pricing of raspberry pi's to an equally as unattractive territory. might as well just get an old android phone and turn it into a server at that point.