RasPi power usage measurements ALL Models


150 posts   Page 3 of 6   1, 2, 3, 4, 5, 6
by Forcie » Sun Jun 17, 2012 8:14 pm
I'm gonna add my 2-cents about RasPi power usage/drawn current measurements.
At idle I get 1.64 Watt (322mA at 5.09V at login prompt on Arch, idle CPU, nothing plugged but power).
Below are some tests I made:

- 322mA ~ 323mA - only power, no IO, (peaks at 350-390mA when CPU solicited),
- 335mA - only a couple PS2 keyboard & optical mouse plugged together on one USB port with an adapter (+30mA when mouse in use, +7mA when num lock LED on!),
- 375mA - ethernet only,
=> CPU stress test with ethernet only with 'stress --cpu 8 --io 4 --vm 2 --vm-bytes 4M --timeout 20s':
current jumps from 375 at idle to 435-465mA,
- 386mA - Mouse&KB + ethernet and idle CPU.

(Another test showed that the couple mouse and keyboard alone draws 28mA when idle, to 80mA when mouse in use + 7 to 8mA per keyboard LED when on.)

- 377mA - ethernet and composite video (compared to 375mA without video composite),
- 389mA - Mouse&KB + composite video + ethernet at idle CPU (not sure how KB+Mouse only adds 12mA, when it's supposed to be 28mA, oh well),
=> CPU stress with this setup:
current jumps from 389mA to 440mA (peaks @ 480mA),
- 394mA - CPU idle, LXDE/X11 on!, composite video, ethernet, PS2 KB&Mouse together on one USB port,
=> Peaks at 475mA ~ 515mA when: mouse in use + CPU stress, composite video, LXDE/X11 on, ethernet,
That was still not enough to blow up my 500mA (fast melting) fuse. I'm sure I can managed to do that by adding a 70mA flashdrive on the remaining USB port though. Didn't try it.

Note that the values above are with mouse idle unless specified (the optical red thingy goes a bit down and draws significant less power when the mouse is not moved: Logitech).
Also note that the peaks are fast displayed values on the multimeter and are probably quite higher for very very short whiles since the multimeter displays short averages. The other values are constant/fixed values.

Also it's interesting to note that after a "sudo halt", the raspberry pi displayed "system halted" as expected, but still draws 435mA constantly with ethernet, Mouse&KB, composite video, 424mA without the Mouse&KB, and then 372 without ethernet (just video)! Those are higher values than when the PI normally operating with corresponding setups.
During the tests, my HTC charger went from 5.15V to 5.01V. From now on I'll use a Y-USB cable that can provide 1A on my computer (2x 500mA): http://goo.gl/Z4IR9. It seems to be totally enough for current but drops to 4.70V at times though. But that's just my computer, and the PI seems to be okay with that.

Anyway, was just taking a look to have an idea what a headless RPi server would need, and it's 375mA to 500mA for the use I'll make of it (to be somewhat sure that it's enough).
So we're still quite below 700mA but you guys knew that before my tests.
Hope all this helps a bit.
Posts: 14
Joined: Sun Jun 10, 2012 9:31 am
by alexeames » Sun Jun 17, 2012 8:32 pm
Forcie wrote:Also it's interesting to note that after a "sudo halt", the raspberry pi displayed "system halted" as expected, but still draws 435mA constantly with ethernet, Mouse&KB, composite video, 424mA without the Mouse&KB, and then 372 without ethernet (just video)! Those are higher values than when the PI normally operating with corresponding setups.


Excellent - thanks for posting. It's great to have independent confirmation of observations - particularly this one about more current draw when it's supposedly "off". Were your measurements made between power supply and Pi? (i.e. After the regulator?)
Alex Eames RasPi.TV HDMIPi.com RasP.iO
User avatar
Posts: 2070
Joined: Sat Mar 03, 2012 11:57 am
Location: UK
by Forcie » Sun Jun 17, 2012 8:41 pm
alexeames wrote:Excellent - thanks for posting. It's great to have independent confirmation of observations - particularly this one about more current draw when it's supposedly "off". Were your measurements made between power supply and Pi? (i.e. After the regulator?)


You're very welcome. Yes indeed, measurements made between power supply and Raspberry PI thanks to a little device like this: http://www.instructables.com/id/USB-Vol ... version-2/.
Posts: 14
Joined: Sun Jun 10, 2012 9:31 am
by AndrewS » Thu Jun 28, 2012 1:53 am
AndrewS wrote:Another quick question for one of the Broadcom guys: presumably undervolting doesn't blow the "there goes your warranty" fuse? ;) Or does any voltage change trigger it?

Now got a reply from Dom viewtopic.php?p=110276#p110276 who says that undervolting doesn't void the warranty :D
User avatar
Posts: 3626
Joined: Sun Apr 22, 2012 4:50 pm
Location: Cambridge, UK
by alexeames » Thu Jun 28, 2012 6:34 am
Interesting - seems Dom agrees with me about the likely small gain from undervolting.

I'll predict the power saving will be small. You'll get most of the benefit from lowering the frequencies.


Still, looks like something else to try :D
Alex Eames RasPi.TV HDMIPi.com RasP.iO
User avatar
Posts: 2070
Joined: Sat Mar 03, 2012 11:57 am
Location: UK
by rew » Thu Jun 28, 2012 8:31 am
In theory, you're charging and discharging millions of capacitors millions of times every second. This costs energy.

There are several things you can do. You can reduce the number of capacitors: idle the CPU as much as possible.

Next you can reduce the number of times every second that you charge and discharge them. Reduce the frequency. The problem with this is that if you need XXX calculations for a task you're not going to reduce the energy spent. you're spreading out the energy use over a longer time.

And lastly, and most importantly, you can reduce the voltage. For each charge-discharge cycle, you need 0.5 * C * U ^2 , where U is the voltage. Reducing the voltage is super-effective in reducing power-use. The drawback is that if you reduce it, the speed of the electronics drops. So signals travelling from one unit to the next will be "too late". So the CPU will start making errors. Usually it'll crash.

So when you reduce voltage, you can also reduce the frequency to compensate for the slower electronics. But at a certain moment there will not be much to gain. If say 0.9V is the limit, you'll be able to run at 70MHz at 0.9V, but at 0.95V (about 10% more power), you'll be able to run at 200MHz: 3x faster. I'm just guessing at the numbers. Don't take my word for it. The moral of the story is that as you drop the voltage, the max-faultless-frequency will start to drop, but at a certain point it will sharply drop below useful values....
Check out our raspberry pi addons: http://www.bitwizard.nl/catalog/
User avatar
Posts: 396
Joined: Fri Aug 26, 2011 3:25 pm
by Alligator » Wed Jul 04, 2012 2:03 pm
Slightly different than the previous discussion, but has anyone tried unsoldering all of the LEDs on the board?

With five LEDs, each pulling (at the most) 20 mA, we could easily save 50 - 100 mA (hopefully).

I need my Pi to consume as little power as possible since it will be a solar powered web server.
User avatar
Posts: 49
Joined: Tue Jun 26, 2012 6:18 pm
by AndrewS » Wed Jul 04, 2012 2:23 pm
Alligator wrote:Slightly different than the previous discussion, but has anyone tried unsoldering all of the LEDs on the board?

With five LEDs, each pulling (at the most) 20 mA, we could easily save 50 - 100 mA (hopefully).

According to the schematics, each LED is connected to 3V3 via a 1K0 resistor, so each LED will only draw a max of 3.3mA, for a total of 16.5mA if they're all on at once.
User avatar
Posts: 3626
Joined: Sun Apr 22, 2012 4:50 pm
Location: Cambridge, UK
by AndrewS » Wed Jul 04, 2012 2:26 pm
Just noticed https://github.com/Hexxeh/rpi-firmware/ ... 649ce9f48a which says "lower powered poweroff state". Fancy doing some more measurements Alex? 8-)
User avatar
Posts: 3626
Joined: Sun Apr 22, 2012 4:50 pm
Location: Cambridge, UK
by godFather89 » Wed Jul 04, 2012 2:28 pm
Alligator wrote:Slightly different than the previous discussion, but has anyone tried unsoldering all of the LEDs on the board?

With five LEDs, each pulling (at the most) 20 mA, we could easily save 50 - 100 mA (hopefully).

I need my Pi to consume as little power as possible since it will be a solar powered web server.


I unsoldered the onboard hub (and ethernet) and got 120mA on idle (from a 4.2V Li-Ion battery). I now have a USB Wi-Fi adapter but I didn't measure the power drawn (I connect it to a powered hub). If you don't need real networking maybe you could use some other communication protocol (RS232, RS485, CAN, etc.).
Posts: 150
Joined: Fri May 18, 2012 9:40 am
Location: Timisoara, RO
by 0117blocky » Wed Jul 04, 2012 2:31 pm
I think you will find the current draw from the LEDs is lower than that as each LED needs about 1.8v to 2V applied to them. Which would mean the 1K resistor would drop about 1.3V to 1.5V therefore the current would be in the region of 2mA for each LED.
Posts: 43
Joined: Mon Apr 30, 2012 8:03 am
by alexeames » Wed Jul 04, 2012 3:33 pm
AndrewS wrote:Just noticed https://github.com/Hexxeh/rpi-firmware/ ... 649ce9f48a which says "lower powered poweroff state". Fancy doing some more measurements Alex? 8-)


I'll get that done soon. Dom mentioned it in another thread last night. :D
Alex Eames RasPi.TV HDMIPi.com RasP.iO
User avatar
Posts: 2070
Joined: Sat Mar 03, 2012 11:57 am
Location: UK
by Forcie » Wed Jul 04, 2012 5:56 pm
0117blocky wrote:I think you will find the current draw from the LEDs is lower than that as each LED needs about 1.8v to 2V applied to them. Which would mean the 1K resistor would drop about 1.3V to 1.5V therefore the current would be in the region of 2mA for each LED.


I would also guess that each LED draws less than 5mA!
Posts: 14
Joined: Sun Jun 10, 2012 9:31 am
by alexeames » Wed Jul 04, 2012 6:54 pm
OK I made a new debian squeeze image and rpi-updated it.

First thing I noticed before doing any measurements is that when shutdown -h you have only the red LED left on. I'm sure this is different.

OK. Onto the measurements. Connected. Logitech kbd/mouse dongle, network cable, HDMI to TV.
Powered by battery and reg.
Before connecting. 5.03V
After connecting 5.00V
Once booted and not logged in: 5.00V 0.43A
logged in: 5.00V 0.40A
startx 5.00V peaked 0.47A settled 0.40A
logout of x - crashed - poo couldn't ssh so pulled the plug :evil:
Start again
logged in 5.00V 0.42A
sudo halt
screen goes blank after "will now halt". LEDs go off apart from power
5.00-5.02V 0.12A - BINGO. Houston - we have a powerdown. :lol:
Alex Eames RasPi.TV HDMIPi.com RasP.iO
User avatar
Posts: 2070
Joined: Sat Mar 03, 2012 11:57 am
Location: UK
by AndrewS » Wed Jul 04, 2012 7:18 pm
Brilliant! Next question for Dom... ;) - is there any way the Pi can be "woken up" again from this 'deep sleep' mode (watchdog timer? external interrupt?), or is the only option to momentarily remove and then reapply power?
User avatar
Posts: 3626
Joined: Sun Apr 22, 2012 4:50 pm
Location: Cambridge, UK
by dom » Wed Jul 04, 2012 8:02 pm
AndrewS wrote:Brilliant! Next question for Dom... ;) - is there any way the Pi can be "woken up" again from this 'deep sleep' mode (watchdog timer? external interrupt?), or is the only option to momentarily remove and then reapply power?


In theory yes. Trouble is we are almost stateless at this point. I sleep before sdram has been enabled, and this is just after a reset.
It turns out there is a register (the one written to in the kernel power down function) that does maintain a few bits of state over the reset, so I can distiguish a power on/reboot from a halt.

There's not enough information to set a wakeup alarm duration, or GPIO interrupt pin.

Possibly we could say, a rising ede on GPIOn (for a fixed n) wakes us up again. (Maybe even the HDMI hotplug line, so you could possibly provoke it by powering on your TV).
Although whatever you choose, someone will grumble...
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 4036
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge
by alexeames » Wed Jul 04, 2012 8:07 pm
dom wrote:Possibly we could say, a rising ede on GPIOn (for a fixed n) wakes us up again. (Maybe even the HDMI hotplug line, so you could possibly provoke it by powering on your TV).
Although whatever you choose, someone will grumble...


You can fool all of the people some of the time. You can fool some of the people all the time. But you can't please all of the people ANY of the time. :lol:
Alex Eames RasPi.TV HDMIPi.com RasP.iO
User avatar
Posts: 2070
Joined: Sat Mar 03, 2012 11:57 am
Location: UK
by AndrewS » Wed Jul 04, 2012 8:13 pm
dom wrote:Possibly we could say, a rising ede on GPIOn (for a fixed n) wakes us up again.

I'd be happy with that :D
(Maybe even the HDMI hotplug line, so you could possibly provoke it by powering on your TV).

But if you're plugged into an HDMI TV, there's a good chance you're running off mains power, in which case a deep-sleep mode isn't so relevant. I'm thinking battery-powered situations... ;)

Although whatever you choose, someone will grumble...

Yup :|
User avatar
Posts: 3626
Joined: Sun Apr 22, 2012 4:50 pm
Location: Cambridge, UK
by dom » Wed Jul 04, 2012 8:15 pm
alexeames wrote:5.00-5.02V 0.12A - BINGO. Houston - we have a powerdown. :lol:


Thanks for meauring. I'll check with Gert if that's the best we can do. My guess is the LAN chip and the 2835 are using almost no power, but the power supply related bits are still wasting power. Perhaps an IR thermometer would spot who's responsible.

Does unplugging USB/ethernet/HDMI have any effect?
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 4036
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge
by dom » Wed Jul 04, 2012 8:17 pm
AndrewS wrote:But if you're plugged into an HDMI TV, there's a good chance you're running off mains power, in which case a deep-sleep mode isn't so relevant. I'm thinking battery-powered situations... ;)


If no HDMI TV, then that pin is available and reasonably accessible. Or you'd prefer one of the GPIO header pins?
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 4036
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge
by alexeames » Wed Jul 04, 2012 8:20 pm
dom wrote:Does unplugging USB/ethernet/HDMI have any effect?


I'll check.
idle cli 5.00V 0.39-0.42A
sudo halt 5.02V 0.10A
unplug kbd/mouse dongle no change
unplug ethernet no change
unplug HDMI no change still 5.02V 0.10A
unplug SD card (just kidding) :lol:

So, in a word NO. :)

Just for fun, I then powered it down, removed the SD card and plugged the power back in to see what it would do.
5.02V 0.08-0.09A
Alex Eames RasPi.TV HDMIPi.com RasP.iO
User avatar
Posts: 2070
Joined: Sat Mar 03, 2012 11:57 am
Location: UK
by godFather89 » Wed Jul 04, 2012 8:33 pm
I also got 120 mA at idle with onboard USB hub removed so my guess is (also considering the LEDs go off) that, when halted, the onboard USB hub is turned off, thus the power save.
Posts: 150
Joined: Fri May 18, 2012 9:40 am
Location: Timisoara, RO
by alexeames » Wed Jul 04, 2012 8:41 pm
godFather89 wrote:I also got 120 mA at idle with onboard USB hub removed so my guess is (also considering the LEDs go off) that, when halted, the onboard USB hub is turned off, thus the power save.


Thanks - it's good to get independent corroboration :D
Alex Eames RasPi.TV HDMIPi.com RasP.iO
User avatar
Posts: 2070
Joined: Sat Mar 03, 2012 11:57 am
Location: UK
by jbeale » Wed Jul 04, 2012 9:02 pm
Having a lower-power shutdown mode is certainly good, but if 120 mA is the lower limit, my hunch is that the most useful thing for a limited-power setup is some kind of fast-boot option (running memory image saved to flash - hibernate mode?) so you can have it fully shut down (0 mA) when not in use.
User avatar
Posts: 2058
Joined: Tue Nov 22, 2011 11:51 pm
by azerty » Tue Jul 10, 2012 2:45 am
With the new firmware I measured 0.107 A after shutdown -h now.
Then I waited 15 for the temperature to re-equilibriate and took this picture:
Image
Looks like the LAN chip is still drawing quite a bit of current.
Posts: 9
Joined: Thu Jun 07, 2012 4:06 am