Raspberry Pi 3 Model B+


 

Bryan Mayland

TVWBB Hall of Fame
Happy Pi Day everyone! Today, the Pi foundation released a new revision of the Raspberry Pi 3.
No versions of HeaterMeter support the 2018 1.4GHz Pi 3 Model B+

Physically, the design is compatible with the HeaterMeter 4.3, but there is no software support yet. I'll have to get my hands on one to start seeing what is new on the software side. We're definitely going to need new boot firmware and a new wifi driver to start. I'm concerned about them saying it requires more power than the current Pi 3, and they're recommending 2.5A 5V power. We've only got a 1.5A 5V supply that is designed to feed the Pi and the Servo which may be pushing it already. The yellow/green LCD also can pull 120mA by itself (which is part of the reason the new kitsd ship with the more expensive black/white 20mA display), and of course the HeaterMeter needs up to 50mA of power as well.

I'll update as I have more information but the important thing to note is to NOT buy the 2018 Pi 3 Model B+ for use with HeaterMeter at this time.
EDIT: Now available in the latest snapshot "March 22 2018 11:35:45 EDT"
 
Last edited:
Is it just me, but I think it's time Raspberry makes changes in the way they name these thing. It's starting to get very confusing.
 
I’ll grab one for playing with simply for the fact that it now has 5GHz wifi. My RasPis are the only reason I still keep 2.4GHz wifi around.
 
Is it just me, but I think it's time Raspberry makes changes in the way they name these thing. It's starting to get very confusing.
Too right. I added the 2018 to it because if someone doesn't look closely they could just see "Raspberry Pi Model B+" ok that's what I got. I know they're only incremental revisions but I dunno you have a whole alphabet, why does everything have the be a B? Oh because all Bs have the same featureset? Oh no the B/B+/2B don't have wifi but the 3B does so that's right out the window. The 2B and 3B share the same core CPUs so those should be the same number I'd think.

I've also got one on the way to me already for development. Looks like the radio is 2.4Ghz or 5Ghz, not simultaneous so that should save a bit of power (transmit on 2.4GHz is up to 400mA, 350mA for 5GHz).
 
Well the good news is that the new header they've added for PoE just barely misses our schottky diode and power MOSFET in a way that I don't think will be a problem in normal assembly unless sticking up off the board by a lot (2-3mm?).

There was a lot more to getting the software working than I expected after only reading the specs, considering the new wifi chip is also a brcmfmac-class device and there's already a driver for the Ethernet chip. I've got a hacked version of the LEDE firmware up and running on the Pi 3B+ and even connecting to 5GHz wifi! Should just be a little while for me to have a snapshot with full support. Also still haven't done testing on power usage to see if there's a problem with brownouts from the increased power usage.
 
New snapshots are up, with Pi 3B+ support in the Pi 2/3/3B+ firmware image.

In testing I am seeing the 3B+ pull up to 750mA on bootup or when the CPU is busy, idling between 400mA to 500mA depending on the wifi configuration and activity (5GHz uses a little less power, but has shorter range). That's about 150mA more than a regular Pi3 runs at, even at idle, which is a pretty big jump up. I've also tested with the latest Raspbian to make sure I'm not missing something in the core power management and it is just as bad-- I wonder if the Pi people are missing something on the software side as well, or do the new Eternet/Wifi/Power chips just pull that much more power?
 
With everybody wanting these PIs to do everything it makes me wonder if the default kernel tunings are increasingly for performance the cost of power consumption. I.e. even "shifting the power management" into a higher performance range.

The Linux scheduler - particularly in a performance mode - may be spreading the processes and threads amonst the cpus when they're busy - and the scheduler then generally attempts to maintain residency for caches (though it will eventually pack longer term sleepers onto a single cpu).

A quick experiment may be to use either or both of "isolcpus" at boot (probably easiest) and cpusets at run time to constrain the scheduler to one (or desired for test) number of cpus.

To see what's assigned to which processor try something like the following. "ps -eo psr,comm,state -T" Should be close -that's off work RHEL 7.x as I don't have a PI handy right now.
 
Seems like this will be an endless evolution with pi version after version, in rapid fire fashion, what a nightmare to keep up with! Personally IDK why you would look any further than the zero-w for the HM. Cheap, fast, low power, built in wifi... I've been extremely happy with mine!
 
New snapshots are up, with Pi 3B+ support in the Pi 2/3/3B+ firmware image.

In testing I am seeing the 3B+ pull up to 750mA on bootup or when the CPU is busy, idling between 400mA to 500mA depending on the wifi configuration and activity (5GHz uses a little less power, but has shorter range). That's about 150mA more than a regular Pi3 runs at, even at idle, which is a pretty big jump up. I've also tested with the latest Raspbian to make sure I'm not missing something in the core power management and it is just as bad-- I wonder if the Pi people are missing something on the software side as well, or do the new Eternet/Wifi/Power chips just pull that much more power?

Your power consumption readings look to be similar to what others have observed.

http://raspi.tv/2018/how-much-power-does-raspberry-pi-3b-use-power-measurements

Still waiting for my 3B+ to arrive!

I'll be pairing it up with this MegaIO board to redo some of my home automation pieces.
 
Seems like this will be an endless evolution with pi version after version, in rapid fire fashion, what a nightmare to keep up with! Personally IDK why you would look any further than the zero-w for the HM. Cheap, fast, low power, built in wifi... I've been extremely happy with mine!

Yup! The Pi Zero W is the perfect board for the HM. Is there anyone that actually uses the wired ethernet port on their heatermeter?
 
I've only used the wired ethernet to config a fresh build, but since Bryan built the new download page with preconfig'd wifi I haven't had to use it. I used to like the idea that I could use a wire if I had a problem connecting but I haven't had any problems with my zero-w through many updates now, so my need for wired lan has expired.
 
I didn't test anything with the Ethernet plugged in but I am guessing it will pull a whole lot more. The spec sheet for the 78XX chip says that when in SUSPEND2 state (which can easily be activated for testing by ifconfig eth0 down) the power usage should be 5mA. When I do that I get down to about 390mA. It also says when connected to 1000T it will pull 200-210mA or 100-110mA for 100T. That tells me that at idle power could go up another 150mA or more if connected to gigabit ethernet!

There's also mention of other suspend states (0, 1, and 3) but those rely on usb autosuspend unless you want to actually write to the registers to enable it. I have USB autosuspend on, but I don't see control files in /sys/bus/usb/devices/XXX/power/ to make sure it is enabled on this device. Documentation says that they should be there if the driver implements the requirements, which Microchip seems to indicate they do so I'm not sure where to go beyond that. Finally, there's the Energy Efficient Ethernet standard, EEE, which I've checked and is enabled by default but I think that only changes power usage when there is a link.
 
2B+ to 3B+ to Zero W

New snapshots are up, with Pi 3B+ support in the Pi 2/3/3B+ firmware image.

In testing I am seeing the 3B+ pull up to 750mA on bootup or when the CPU is busy, idling between 400mA to 500mA depending on the wifi configuration and activity (5GHz uses a little less power, but has shorter range). That's about 150mA more than a regular Pi3 runs at, even at idle, which is a pretty big jump up.

Bryan, I moved from 2B+ to 3B+ because my Edimax Wifi dongle was overheating. Newer = better, right? That is noob logic, at least, so....

I then discovered that 3B+ with the standard Titan power adapter + Adapt-a-Damper was (probably) power-sapping, taking RasPi offline randomly when blower / damper called for amps. Was correlated enough to be "probably." Then, $42 later, I discovered this thread.

The Zero W with header pins pre-soldered is $14 from Adafruit = problem solved, as draw is tiny. I would encourage you to advocate Zero W more strongly now, especially to noobs, who are going to wander down this same path (hence this post).
 
The Zero W with header pins pre-soldered is $14 from Adafruit = problem solved, as draw is tiny. I would encourage you to advocate Zero W more strongly now, especially to noobs, who are going to wander down this same path (hence this post).

Agreed, I've been very happy with the rPi zero-w, I built a HMv4.3 with one a while ago, and more recently my first rPi Model B started acting flaky so I replaced it with another zero-w. Both are working flawlessly since day 1, and I put them through 24/7 torture tests.
 
The RasPi isn’t needed to use the HM. Try disconnecting the RasPi and use the HM in manual mode to see if the servo is still acting up. That will help rule out the RasPi as the problem.
 
Thanks Steve,
just tried, and I have same issue.
Good news, RasPi is ok...bad news I have really no idea what else to check :p
 

 

Back
Top