Raspberry Pi 4 announced

Just received a 4GB Pi4 and thought I'd offer another rationale for its existence that isn't often talked about: The Pi's relatively miniscule power consumption makes it a great [backup] desktop option during power outages (i.e., when forced to run for extended periods on limited battery power).

I'm currently running it with a non-standard power supply (p/s) that isn't providing the recommended dose of current, but an OEM p/s's scheduled to arrive tomorrow and I'll have a better take once it's hooked-up good 'n' proper.

But, so far at least, this gizmo would easily suffice when the lights're out and one needs to be mindful of every milliamp. And geting it to [properly and reliably] run XP or 'BSD would ice the cake.
 
Actually, that's a good argument. Say it uses 250mA at 5V; add to that 50mA each for keyboard and mouse, that's less than 2W for the computer itself. The problem is actually the screen; just looked it up: The official 7" RPi display uses 0.5A at 5V (that's another 2.5W); while it is a small and crappy display, that's better than nothing. SO let's assume the total is 5W. A car-size 12V battery contains about 30Ah usable (don't want to drain it much further); at 12V that's 360Wh, which can provide the 5W for the RPi for 72 hours. Even if you slap a few penalties for voltage conversion on, that's still a heck of a long time.

On the other hand, a good laptop keeps you going for 12 hours, and has WAY better display.
 
I wish I had a laptop that ran for 12 hours, but I don't. Still, I reckon that any laptop will draw more current than the Pi, so I 'spect the Pi'll still run longer.

In my case, when the lights go out: I've got a 19" LED TV that runs on ~30w (which I'll be mating to a small Blu-Ray player that runs on ~10w) that I'll be running the Pi through. My least power-hungry laptop—a dated but fully functional Dell running BSD 12—eats ~70w and the Pi's got more RAM (4gb vs .5gb in the laptop), a faster cpu (1.5g vs 1.0g) and runs longer....lots longer. And since I already had the TV, mouse and keyboard, the Pi cost me less than I'd spend on a coupla books. I kinda think everyone outta get a Pi and some type of battery back-up to use when the unexpected occurs. And, aside from its many nits, I think that more than justifies the price of admission.
 
I agree this is a very low-cost version of getting compute power for when the power goes out. Your suggested setup requires lots of tinkering, and you'll have to deal with several voltage converters, but that's all doable.

We have a "similar" solution, just with different components. In spite of living within 20 minutes of Silicon Valley (the technologically most advanced place on the planet), our electrical power at home is quite unreliable, with frequent outages due to rain and storms in the winter, and due to fires in the summer. Unfortunately, our house needs electricity to have water (we have our own well and pump system), and water is particularly important when there is a fire nearby. So the way the way I get nearly uninterrupted computing is this: Small computer in the basement, together with a wireless AP and the DSL modem. Both are fed by a small office-grade UPS. All user interface computers are laptops, which have their own batteries. The whole house electrical system is on a large automatic transfer switch, which is fed from our normal electrical company, and from a generator. The generator is a 16kW model, supplied from a 500 gallon (2000L) tank of propane. When the power goes out, all we notice is that the lights go out for about 15 seconds, which is how long it takes for the generator to decide that the electrical grid has really failed, start the engine, and operate the transfer switch.
 
The whole house electrical system is on a large automatic transfer switch, which is fed from our normal electrical company, and from a generator. The generator is a 16kW model, supplied from a 500 gallon (2000L) tank of propane. When the power goes out, all we notice is that the lights go out for about 15 seconds, which is how long it takes for the generator to decide that the electrical grid has really failed, start the engine, and operate the transfer switch.

Not to deviate from the thread theme (too far).....
Although I'm on the other side of the country, it sounds like we share the same [unreliable] electrical "pump".

I've been without power—sitting in the dark—for > five consecutive days, four times in the last seven years. I learned my lesson after the third time. Unfortunately, I'm unable to install a 16kW generator (though my boss's got something similar at his pad), so I went with a 1.1kW solar generator, which, although it lacks the testicular fortitude to run high current devices (e.g., house A/C, heat, electric stoves, etc.....tho' it will run the likes of an upright Hoover and microwave oven), it's silent, small/portable, produces no noxious exhaust (i.e., usable indoors) and able to run small- to mid-sized devices for as long as there's sunlight to keep its battery "refueled".

And since switching all the 75-to-100w incandescent lights over to 6-7w LED light bulbs, purchasing a coupla lithium/D-cell-powered smart fans and having the tiny-watt Pi to use in place of 70+ watt laptops/desktops, I'm finding that I can weather the coming storms with relative ease. And all for a relatively reasonable cost that doesn't keep the neighborhood up all night..
 
Your suggested setup requires lots of tinkering, and you'll have to deal with several voltage converters, but that's all doable.

Yeah, it used to. I built a diy generator a few years ago that did require a good bit of messing about. But after it got stolen by landscapers, I abandoned that for one of these new plug 'n' play generators and some PV panels (which just plug into the generator). Easy-peasy. While not perfect, it gets the job done and I never even have to warm up the Weller or pester any HAM guys to check my (often faulty) electrical computations ("e=cm2" or something like that)..
 
UPDATE: Now that I've got the proper p/s, shadowed the cpu/ram with an oscillating propeller contraption (i.e., cpu fan from the Pi guy, which's keeping the board close to room temperature!), tinkered with Raspian's innards and placated the LAN config abacus, this thing swings! I've been running it for several hours today--watching youtube videos (via Chromium), taking care of a bit o' correspondence (via Claws) and watching a coupla .mp4s (via VLC)--and it hasn't missed a beat or bottle-necked my movements in any way. Golly, I dunno how I did without a Pi all this time. Everyone really outta get one of these gizmos, if for no other reason than as a backup desktop because it runs on fumes (i.e., 15 watts)! Kudos to the Pi guys--they finally got it right. I guess the next version'll be a Pi-supercomputer running Plan9 or Maruti.
 

Attachments

  • pi4.jpg
    pi4.jpg
    135.5 KB · Views: 444
My experience with 12 mirrors yours. I've tried three ARM versions of 12 and none booted. Until/unless there's an ARM version made for the Pi, I'll stick with Raspian, which I think is a very good OS for the platform.
 
It won't because uboot has not been updated yet.
Yes but that is one small part of it. The DeviceTreeBinary is like the BIOS in that it describes every piece of hardware to the O/S.
Uboot is like a bootstrap loader.
Now imagine the differences between Pi3 and Pi4.
Do you really think another DTB is going to work. No it will not. It is describing totally different hardware.
The DTB describes the harware and address locations to the OS loader.
 
Regardless, no uboot, no boot. It's only just been finished for Linux, so expect to wait a year for FreeBSD. Use Linux instead. (I differentiate Linux v Raspbian as obviously Rasbian has binary blobs for their boot code - via the GPU)
 
Raspbian very good OS for that platform. You are right. And I am really impressed with the capabilities of this micro computer.
 
It won't because uboot has not been updated yet.

There is a functioning u-boot, written by Andrei Gherzan, for the purposes of booting 64 bit Linux. He is working on upstreaming his patches. I have some patches on top of his on my github aimed at FreeBSD. It’s entirely experimental. I don’t think I can post a link to it as a newbie here but you can google it.

However, u-boot passes a bad pointer to the http functions which crashes the freebsd loader; a work around is to compile the freebsd loader without http support.

There are a few other changes you need to make in the bcm device drivers to get a bootable kernel. You also need to patch armstubs.bin if you want multiprocessor support (also on my github but needs a further update to initialize the interrupt controller).

At the moment I am having a lot of trouble with the DMA controller. Disabling DMA support in the sdhci driver has enabled me to mount root. Yesterday I got as far as starting /sbin/init, but I have another panic to debug.

Even when this is done, there is no support yet for ethernet or USB. USB is tricky because it comes over PCI-E for which we don’t have a driver. Trying to smuggle in a generic driver crashed during hardware enumeration, but I admit I didn’t try very hard.

The ethernet may work with some tweaks to the mii driver. I haven’t tried it yet.

The framebuffer driver also needs work.
 
I have some patches on top of his on my github aimed at FreeBSD. I don’t think I can post a link to it as a newbie but you can google it.
We don't have a problem with links like that. We only frown at links to sites that have nothing to do with FreeBSD (mainly to combat SEO spam).
 
I am pleased to report I have reached the login prompt on the raspberry pi 4.

Unfortunately I can’t login yet as shortly after /sbin/init starts something trashes the serial com link and I only get garbage over serial (in either direction). Probably just a misconfiguration.
 
My original RPi 1 is still nailed to the wall and chugging away serving web pages. Just keeps on ticking. :)
 
You really can't go wrong with any of the rpi models. I have the rpi4 coming and am looking forward to tooling with 64bits and 4gb ram. I do nearly everything important on my network via my headless rpi and this will be a much needed upgrade for me.
 
I am pleased to report I have reached the login prompt on the raspberry pi 4.

Unfortunately I can’t login yet as shortly after /sbin/init starts something trashes the serial com link and I only get garbage over serial (in either direction). Probably just a misconfiguration.
Almost!
 
I tried FreeBSD 12/13/release then neither of them can boot on RPI4B.

There are a number of code and configuration changes you need to make in the armstub firmware file, u-boot, and the FreeBSD kernel.

I'll try to find some time to package up what I have so far. I want to make sure SMP works properly first (I did most of the legwork for it but then switched back to testing single-threaded for ease of debugging).

Note, I do not have USB or ethernet working, these require new drivers, so there's not a lot you can do with the Pi4 even when it is running FreeBSD.
 
For a few weeks, I had a RPi3B on my workbench, while I was prototyping with it and setting it up (it now does full-time duty as an industrial control data acquisition system, which is a lot of big words for saying it measures and records a few water pressures and tank water levels). During the setup time, I had monitor, mouse and keyboard connected, and for fun I started X on it. Works fine, you can run a web browser, you can edit text in multiple terminal emulator windows. Boring. Just a normal Unix desktop. A little slow at time, but at $35 and 3W power consumption, I can handle that. So if the RPi3B was already a nice little desktop, the model 4 would definitely do it.


With the USB 3.0, is that really necessary? And given the form factor, and the size of a SATA connector, I'm not sure that would work well, without making it physically incompatible with the established ecosystem.


Which brings up a question: For what workload and application would the CPU and other interfaces be able to actually serve the 100 MByte/s that true gigabit can transport? Given the CPU speed and number of bytes, there are only 60 instructions one could execute per byte. That's not very much.


That might be feasible, by putting the M.2 "socket" on the bottom side of the board, parallel to the board; sort of in the same fashion that the SD card slot is mounted today. It might barely fit. Given that all the engineering information (circuit diagram, PC board layout) for the Pi is public, I think it should be possible for a vendor (either the Pi foundation or someone else) to try to build such a device, although I'm not 100% sure what the IP situation is.
You can get a msata hat
 
u-boot works fine on the rpi-4 and has done for a while

From a FreeBSD perspective I have a bootable system (running arm64 13-CURRENT smp) but the problem is now the lack of any drivers. No ethernet, no PCI-E, no USB. (Well, the OTG port works out of the box but that’s most typically used for power.)
 
Back
Top