The Raspberry Pi 5 is the new wunderkind single-board computer on the block, so new in fact that users are still finding out its quirks. One of those quirks is a surprisingly high power consumption when powered down, despite halting the SoC, it leaves the power on and consumes over a watt even in standby. [Jeff Geerling] has a solution, and it’s a simple config change.
It’s useful to know how to fix this, and we’re indebted to him for finding it, but it’s hardly the most complex of hacks. Where the interest lies is in why the board leaves the lights turned on when nobody’s at home in the first place. It seems that some HATs have an issue when the 3V3 rail shuts down, but the 5V rail doesn’t. The Raspberry Pi foundation took the most compatible route and kept the rails on all the time. Perhaps future OS releases will come up with something more elegant, but at least there *is* a fix.
If you’re new to the Pi 5, you can take a look at our review of a preview model, and see why it’s the closest yet to a usable everyday PC that they’ve produced.
Is this also a problem with Pi 4? If so, does the same fix apply? I’d fancy the LED off at very least when powered off.
How can a voltage regulator consume a watt with no load ?
Hopefully not the regulator itself, but it is likely powering some peripherals.
Not the regulator but SoC itself!
It doesn’t, the SoC remains on but in a halt state, it doesn’t actually switch off the SoC.
It seems quite ridiculous that the whole SoC needs powered just to keep the 3.3 V rail on. This is obviously a problem that raspberry pi knew about and how hard would it have been to separate the 3.3 V for the SoC and 3.3 V power pin of the GPIO header?
Also, if the 5v rail never turns off that can cause many problems by forcing USB devices to run forever. Unless they’ve got PM for USB and not the headers.
VUSB is not connected directly to 5V rail since forever (something like 3B).
And just how many separate power rails or switches do you want to add to every device? As by that sort of argument you start getting into does every 3.3v pin on the GPIO need its own, and because the load you are switching in and out could be a pretty huge swing more caps are needed to smooth out the power rail etc.
Ultimately I expect the Pi folks didn’t even know in advance they would need to keep that rail powered to keep the HAT happy – afterall how many HAT are there compared to how many would care at all? Even if they did lots of research it would be easy to conclude it didn’t matter while in the early enough phase of the design you could do something about it, and now they have to take the only way out that keeps all their users happy enough as everything works – and as that worked well enough for the Pi4 why put additional costs per unit on the Pi5? (The power off flag IIRC was also on Pi4 and really great saving as it also cut that power hog of a USB chip off giving you lots of control of how much power it draws in its ‘off’ state.)
Is it possible that this is some elaborate backward way of fixing SDCard data corruption so prevalent on Piss?
Is this verified? I tried it and it didnt work. My gpio pins are still pumping out juice and powering hats while the pi5 is “off”.
1 W consumption shen powered off? Man that sucks big time (sucks a lot of power). Sorry, but no excuses there are acceptable. The default should be the power saving mode and a boot time warning that some (a small minority) could have issues and how to fix them. Shame! Shame! Shame!
It’s a brand new release and you’re expecting perfection. If you’re so bright why don’t you crank one out from scratch and let’s see how yours turns out.
I guess the point is there’s a big team and big money behind the design of the raspberry pi. Whether they knew about this problem or not when they released the board, and config to the public is overlooking something relatively simple as the “issue” is out in plain sight. 1W+ of power consumption when the device is expected to be “off” is something they should’ve noticed in at least one of the phases of producing the pi’s. Unless of course they “designed” it to be that way…
How do people already have these? I preordered mine on launch day and I was quoted sometime in January.
Depends where you ordered it from. Some places seem to give first shipments if you order their kit (with power adapter, hdmi cord etc.) and not just the bare pi5 by itself. I ordered a kit from CanaKit a few days ago and it says estimates late November, but I wouldn’t be surprised if it arrives in December or January.
Haven’t got my pre-order yet either… I think it is like an amoeba … slowly spreads out from the center in all directions, … and crossing the ocean isn’t easy ;) .
As for the above ‘problem’, for me it isn’t … Mine are either on 24×7 or disconnected from power. I really like the cords with a switch on them. I wish the official 5V-5A cord and one built in to it. Would be worth the extra couple bucks!
magic settings for closed source firmware that will change with the next OS release! once again, user contributions are limited to helping people poke blindly at the closed source firmware! and in fact you can see there is a rich history of people complaining that POWER_OFF_ON_HALT works differently than documented, works differently than other users have reported, because *it’s just poking at a closed source blob that keeps changing*!!!
people just told me this was the past but it’s also the future, at raspberry
I got mine from coolcomponents.co.uk about a fortnight ago. They seemed to be the only one of the usual suspects who had them in stock. As of time of posting this [13 Nov] they’re out of the 8GB model but still have 100+ of the 4GB available.