I'm not a cheap person, but one that specifies system setups to the proper scale.
I'm the proud new owner of an Ender3 v.2, along with an Ender3 Pro, and a (very sick) Monoprice Mini v.1. I've been happily running Octoprint on the same 2 Raspberry Pi devices for 2+ years.
I'm a retired Systems & Network Engineer (of ~27 years). I use simple Pi 3A+'s on both my Ender3 printers (the MP Mini is doa right now). I use a 64GB High-Speed ScanDISK or Samsung SD cards. You can really notice the speed!
I use Cura 4.8 as a slicer. I've installed a variety of editing tools. Blender, Libre-CAD, FreeCAD, Fusion360 (free version) & some others. I use Tinkercad for 1st pass object design. Then I use FreeCAD or F360, Blender still scares me! LOL
I have an Ubuntu 20.10 HP desktop with 32MB RAM and a 1TB SSD and a 2TB HHD, 750MB USB3 HDD. Space isn't an issue.
I have an HP "Rizen 5" with 32GB RAM, a 128MB SSD, and a 2TB HHD. The 128 is not enough.
The newest version of (stable) Octoprint, works great! It DOES NOT tax my system. I use a single Pi-Camera. Right now I'm printing 15+ hours a day & my systems are NEVER taxed.
You don't need a "monster 8GB Pi 4". No, a Pi Zero W won't cut it. But the 3A+ is plenty of machine. See the attached system monitor screenshot.
While in normal use I would agree, where I found I needed to go up to the pi4 was I often end up with multiple machines watching the video stream. That made the pi3 quite warm, versus the 4 (admittedly with a cooler) never seems to be working hard.
On printers with a small buffer this is not true. Anycubic Mega S for example suffers print quality when it does numerous unwanted small pauses when printing with a slower Rpi. This is if course only true when the gcode consists of a lot of short commands, so it depends a lot on what you are printing.
With boards with a small buffer and prints with a lot of short gcode prints a slower Rpi will have trouble with the printer board buffer running empty at times. It will not have a steady enough data transfer. The max speed over USB is by far enough. But the small buffer on the printer board will put a strain on it responding fast enough when the printer at a high rate again and again asks for stopping for more data, and more crucially asks for more data. And at times when the Rpi doesn't respond fast enough the printer will print very short gcode items so fast that the buffer goes empty. Then the printer pauses for a fraction of time, waiting for more data to arrive to the empty buffer. When the printer stands still this very short time it will make a small blob due to the pressure in the print head since the filament isn't retracted. There is a long thread on this somewhere here, I did a lot of testing with different Rpis, printing from the memory card and running Octoprint on faster hardware.
Note: it's only a problem with prints that have a lot of short gcode prints in fast succession and boards with very small buffers.
Never tried a Rpi 4. But a low end Intel pc solved all blob problems on prints with multiple short gcode prints like small cylinders. You don't have to believe me
On a normal PC you don't have your printer, webcam and network connected to a single usb port which is limited to around 6 MB/s because otherwise it will shred the sd card and you have to reinstall your OS after a while.
I had OctoPi on a Zero with a Wifi Dongle and webcam attached through a USB hub.
It did work, but took an awful toll on the throughput of the video feed, and rendering time of any timelapse (that would take usually almost as long as the print itself).
Now I moved to a Pi 3B+, with the same wifi dongle and webcam. And that works satisfactory indeed. Still would like to see the rendering of the timelapses go a bit faster, but that alone is for me not worth investing in a Pi4.
The only reason for using a Pi4 is when you use the pandaPi board.
Then the marlin compilation is much faster.
But even when Marlin, Octoprint and mjpeg streamer run on a Pi3 there are no issues.
What I found is that using a small SSD (64 or 128 GB) on the USB of a Pi 3 (which can boot from USB directly) made a lot of difference, both in terms of speed and durability. I'm running both my octopi and a home control server from (two different) Pi 3 and they're much more responsive than from SD card. The random access speeds for SSDs are a lot better even over USB2.
A Pi 3B+ is doing his job here quite good. I also thinking about using one of the older 3B (without +) but they do not have WiFi. However, I'll test them with OctoPrint and a USB WiPi-Dongle laying around when I've time. I'm printing directly from the Pi to my Ender 3v2, not using the SD card anymore. No issues with the prints so far. No pauses etc. Pi's are running on SD cards here (but not the slowest ones, mine are EVO+)
Oh, and I'm happy user of the OctoPod iPhone app to print over Octoprint and also the corresponding Cura Plugin. Also used Astroprint together with Octoprint do to some "emergency" printing remotely from work.
I am using a Raspberry Pi 3 with a single rPi Camera and it works great. For a second camera, I am using a Wyze Cam V2 internet camera. For $25 this is a fully self-contained WEB cam hires color with infra-red cut filter. I can view it from anywhere on or off my local network. Everything WiFi, Infrared light source, USB power brick included. No monthly fees for basic services.
I think if you are dedicating the PI to one printer, yes , a 3 would be sufficient but I am not always 3d printing so I tend to use my 4, 8gb for multiple uses and just swap out the SD card where needed...( that makes me really cheap since a PI's are so cheap!). I've had no issues with PI and Octoprint. Thanks team!!!
It isn't just the clock speed difference of 0.1 GHz. The memory speed is significantly faster (fewer clocks needed to read and write). The Cortex-a72 can decode 3 instructions on each clock instead of just 2 on the Cortex-a53. The superscalar pipeline is 15 deep instead of 8 to support out of order execution so that 5 instructions can execute simultaneously even out of order if it is safe (a very high percentage of time when code is written for it as in the Pi 4). There is s much larger on core cache to reduce the number of memory accesses.
The I/O speed is improved with faster access to all of the USB ports. It has a more efficient network stack. Most of all, the will do the same work at a higher speed and consume less power.
People need to not look at just the GHz of a processor but how fast it can accomplish a task at what cost of power.