Hi guys,

I didn't see it mentioned anywhere on the forums yet, but it looks like Raspberry Pi silently introduced a Raspberry Pi 2 Model B revision 1.3, that contains a BCM2837B0, instead of the BCM2837A (revision 1.2) or BCM2836 (revision 1.1). Official information here: https://www.raspberrypi.com/products/ra ... 2-model-b/

They also released an official EOL notice, but that's only for revision 1.1 and 1.2: https://pip.raspberrypi.com/documents/R ... i2-eol.pdf

How do you guys think this compares to the BCM2837A in terms of performance and power consumption? Due to the low power, the revision 1.2 specifically is and was very suitable to run as lightweight 24/7 server for tasks that don't require a lot of performance, such as a TVheadend server in combination with the TV HAT.

Now to see where to order one.. :D

Hmmm... While the picture in the link wouldn't seem so, I wonder if this revision is to use the "flipped" BCM2837 with the heat spreader on top....

Will have to see when they actually show up in the channel, though that's hard to determine as most resellers aren't specific enough.

W. H. Heydt wrote: ↑

Mon Nov 03, 2025 6:42 pm

Hmmm... While the picture in the link wouldn't seem so, I wonder if this revision is to use the "flipped" BCM2837 with the heat spreader on top....

Yes - that's what the BCM2837B0 is. BCM2837A0 is the old one with a black package - I guess they just haven't updated the picture yet.

So it's now a 3B+ without the WiFi/ Bluetooth/ POE ?

andrum99 wrote: ↑

Mon Nov 03, 2025 7:05 pm

W. H. Heydt wrote: ↑

Mon Nov 03, 2025 6:42 pm

Hmmm... While the picture in the link wouldn't seem so, I wonder if this revision is to use the "flipped" BCM2837 with the heat spreader on top....

Yes - that's what the BCM2837B0 is. BCM2837A0 is the old one with a black package - I guess they just haven't updated the picture yet.

Ah... Good. I've been waiting for that change for a long time. I'll just have to keep an eye on the Usual Sources.

gordon77 wrote: ↑

Mon Nov 03, 2025 7:33 pm

So it's now a 3B+ without the WiFi/ Bluetooth/ POE ?

Exactly! That's also what the revision 1.2 was, a Pi 3 but without the extra fluffs, and clocked at 900 Mhz. Even when clocked at 900 Mhz, it gave a considerable performance boost compared to the original BCM2836 of the 1.1 revision, and also consumed less power than either Pi 2 rev 1.1 or Pi 3. You also got 64 bit support as a bonus. ;)

W. H. Heydt wrote: ↑

Mon Nov 03, 2025 9:00 pm

Ah... Good. I've been waiting for that change for a long time. I'll just have to keep an eye on the Usual Sources.

Wondering what the usual sources are, seeing as it is already really difficult to specifically buy a Pi 2 rev 1.2, even when it is listed as such. Most of the time you get a new old stock rev 1.1.

boe wrote: ↑

Mon Nov 03, 2025 9:39 pm

W. H. Heydt wrote: ↑

Mon Nov 03, 2025 9:00 pm

Ah... Good. I've been waiting for that change for a long time. I'll just have to keep an eye on the Usual Sources.

Wondering what the usual sources are, seeing as it is already really difficult to specifically buy a Pi 2 rev 1.2, even when it is listed as such. Most of the time you get a new old stock rev 1.1.

Yes...that's the problem.

The bcm2836 is going eol so in future only the bcm2837 based Pi2 will be available.

Software guy, working in the applications team.

jamesh wrote: ↑

Tue Nov 04, 2025 6:20 am

The bcm2836 is going eol so in future only the bcm2837 based Pi2 will be available.

But doesn't that contradict the EOL statement? Both revision 1.1 and 1.2 are EOL, so also the BCM2837(A) counterpart is going EOL.

I've been in contact with several of the industrial distributors, none of which are able to specifically order the new 1.3 revision, even though the product page specifically recommends revision 1.3.

Do you have any idea how one could order a revision 1.3? The product information page of the Pi 2 (https://pip.raspberrypi.com/categories/ ... -2-model-b) lists either part number SC0021 or SC1029, what are the differences between those?

Once stocks of older devices are exhausted only 1.3 will be available.

Software guy, working in the applications team.

jamesh wrote: ↑

Tue Nov 04, 2025 7:34 pm

Once stocks of older devices are exhausted only 1.3 will be available.

Well... That's the problem. If a company still has stocks of the earlier revision, and they're selling slowly, how does one tell when they run out of the old stock? The reseller is very unlikely to say. The times I've asked which revision a reseller had in stock, they have declined to say because they would have to go through all the ones they have to determine that and there is a LOT of pushback if one wants a particular revision.

Now..if the RPT could add to the vendor lists which revisions any give reseller has in stock, it'd be a lot easier for us out here in the cheap seats.

I doubt we would do that. We sell the 2B as the 2B, irrespective of revision number. It's quite possible there are no 2826 based devices left out there, but it's difficult to tell.

Software guy, working in the applications team.

With the release of v1.3 are the website statements still true ?

jamesh wrote: ↑

Wed Nov 05, 2025 11:35 am

I doubt we would do that. We sell the 2B as the 2B, irrespective of revision number. It's quite possible there are no 2826 based devices left out there, but it's difficult to tell.

I understand the issues with old stock, but still find it strange that the website seems to recommend the 1.3 revision specifically, but there is no way to actually get a 1.3 revision.

For me as a consumer that's fine (but unfortunate!), I just wondered how it'd work for larger (industrial) businesses that really want/need to know what revision they're getting. I work in the medical sector, and just "getting" a random revision of a product is usually unacceptable. The embedded IPCs (industrial PCs) we use get new catalog numbers each time a new revision pops up, and sometimes all that changed (for various reasons, including EOL) was a single chip.

One could also argue that with the additional benefits of the BCM2837 SoC (most importantly the 64 bit support) it maybe should've been a Pi 2 Model B+, but that's a different discussion I suppose.

Apart from a few, almost the entire Pi 2 production runs do go to businesses*, and we do talk with them regularly about future plans and how to transition between models as needed.


* Very limited demand from general users for three generation old equipment.

Software guy, working in the applications team.

I have an old application for which the Model 2 is ideal - just enough capacity to do the task, lower power consumption and less overheating than the 3 or 4. Both of the Model 2s I own are rev 1.1, so therefore I'm limited to 32-bit. I also have some zeros that will never be 64-bit, but all of my other pi: the Zero 2Ws and 3s and 4s and 5 are all now 64-bit. I would love, for consistency alone, to have a Model 2 that could be 64-bit. But it is much harder to find one, as even rpilocator.com doesn't track model 2s.

The consumer outlets in the US - all of those listed on the Model 2 product page - either don't carry it anymore, or are out of stock: (microcenter, pishop.us, adafruit, canakit, vilros.)

Off the industrial tab, depending on how I get to their Model 2 page, Newark either says available or no longer available. Pishop.us, none there, same as off the home tab. Mouser and digikey say they have stock on hand, but I've never ordered from them. Perhaps I'll give it a try.

Edit: But, even after looking at the mouser and digikey sites, I wouldn't be sure which rev I would be getting, to the OP's point.

tinker2much wrote: ↑

Fri Nov 28, 2025 12:20 am

Edit: But, even after looking at the mouser and digikey sites, I wouldn't be sure which rev I would be getting, to the OP's point.

At this point in time, it's rather unlikely that you'd still be able to get a Pi2Bv1.1 as the changeover to the Pi2Bv1.2 was several years ago. Much harder to determine if what you'd get would be a Pi2Bv1.3.

I just ordered this item from Mouser; when I see which one I get, I'll report back.

tinker2much wrote: ↑

Sat Nov 29, 2025 9:42 pm

I just ordered this item from Mouser; when I see which one I get, I'll report back.

Please post which version you get. If it's the v1.3, I'll order a couple from them.

W. H. Heydt wrote: ↑

Fri Nov 28, 2025 5:11 am

tinker2much wrote: ↑

Fri Nov 28, 2025 12:20 am

Edit: But, even after looking at the mouser and digikey sites, I wouldn't be sure which rev I would be getting, to the OP's point.

At this point in time, it's rather unlikely that you'd still be able to get a Pi2Bv1.1 as the changeover to the Pi2Bv1.2 was several years ago. Much harder to determine if what you'd get would be a Pi2Bv1.3.

I thought the same would be true, but I received a v1.1 from Reichelt just several months ago. I returned it (because I also wanted a 64 bit capable one) and got a full refund.

W. H. Heydt wrote: ↑

Sat Nov 29, 2025 11:25 pm

tinker2much wrote: ↑

Sat Nov 29, 2025 9:42 pm

I just ordered this item from Mouser; when I see which one I get, I'll report back.

Please post which version you get. If it's the v1.3, I'll order a couple from them.

Yes, please do! It would be really cool if you get a v1.3.

I've been keeping track on the stock at Digikey, and it looks like November 28th was the LTB date for them. They've marked the product now as EOL and it will not be restocked once the stock depletes. Interestingly the reason listed is that the product is no longer being made, so maybe they're not willing to invest in the 1.3 revision.

If you don't get a v1.3, I'd be willing to order one from Digikey, just to confirm what revision is available there. They have about 1200 left in stock.


I got that you're in the US, but I found one reseller which at least explicitly lists v1.2 as option available for purchase: https://www.welectron.com/Raspberry-Pi-2-Modell-B

This could be a nice alternative if you're willing to pay for shipping from Europe.

boe wrote: ↑

Mon Nov 03, 2025 6:06 pm

How do you guys think this compares to the BCM2837A in terms of performance and power consumption? Due to the low power, the revision 1.2 specifically is and was very suitable to run as lightweight 24/7 server for tasks that don't require a lot of performance, such as a TVheadend server in combination with the TV HAT.

You can start with adding:

dtoverlay=disable-wifi
dtoverlay=disable-bt
arm_freq=900

in config.txt of an RPi3B and measure what this means for powerconsumption, per day or week or month. TV-HAT requires real HW, but many 24/7 functions would fit on more or less random computing platforms. So sort of shared power consumption is possible.
I find it quite amazing that so much effort and searching seems to be spent to get older and less HW (as no BT+WiFi, but you might want later). I assume prices of an RPI2 and RPI3B are the same, maybe that is not the case.

redvli wrote: ↑

Mon Dec 01, 2025 11:18 am

boe wrote: ↑

Mon Nov 03, 2025 6:06 pm

How do you guys think this compares to the BCM2837A in terms of performance and power consumption? Due to the low power, the revision 1.2 specifically is and was very suitable to run as lightweight 24/7 server for tasks that don't require a lot of performance, such as a TVheadend server in combination with the TV HAT.

You can start with adding:

dtoverlay=disable-wifi
dtoverlay=disable-bt
arm_freq=900

in config.txt of an RPi3B and measure what this means for powerconsumption, per day or week or month. TV-HAT requires real HW, but many 24/7 functions would fit on more or less random computing platforms. So sort of shared power consumption is possible.
I find it quite amazing that so much effort and searching seems to be spent to get older and less HW (as no BT+WiFi, but you might want later). I assume prices of an RPI2 and RPI3B are the same, maybe that is not the case.

The reason why the Pi 2B is preferred over an under-clocked 3B is because even when under-clocked, the Pi 2B still runs cooler and draws less power. I think this explanation sums it up nicely: viewtopic.php?t=180286#p1150260

I can't speak for others, but the reason why the Pi 2B is for me preferred over the 3B(+), 4 or 5 is because:

• It shines in applications that require a Pi 0/1 as minimum (such as TVheadend, which is really lightweight), but is much faster and therefore has ample room to breathe.
• It has a really low power consumption (almost the same as a Pi 1), especially when idle, which is great for Pis that run 24/7. A Pi 3B+ uses almost the double amount of power when idle (!), let alone a Pi 4 or 5. If there is no specific use for the additional performance, I think having a higher spec Pi is just wasteful.
• It has an Ethernet controller, which is (IMO) a must when tasking a Pi with any server task.
• 1.2 and 1.3 revisions have the same BCM2837 SoC as the Pi 3, enabling 64 bit support, which gives it a brighter future along with some nice free performance improvements.
• The BCM2837 SoC also gives it some additional features such as USB and network boot, which are handy at times.

Sure, if WiFi, Bluetooth or other improvements introduced in later models may become relevant later, then the Pi 2B probably isn't the best choice.

boe wrote: ↑

Mon Dec 01, 2025 3:02 pm

The reason why the Pi 2B is preferred over an under-clocked 3B is because even when under-clocked, the Pi 2B still runs cooler and draws less power. I think this explanation sums it up nicely: viewtopic.php?t=180286#p1150260

I can't speak for others, but the reason why the Pi 2B is for me preferred over the 3B(+), 4 or 5 is because:

• It shines in applications that require a Pi 0/1 as minimum (such as TVheadend, which is really lightweight), but is much faster and therefore has ample room to breathe.
• It has a really low power consumption (almost the same as a Pi 1), especially when idle, which is great for Pis that run 24/7. A Pi 3B+ uses almost the double amount of power when idle (!), let alone a Pi 4 or 5. If there is no specific use for the additional performance, I think having a higher spec Pi is just wasteful.
• It has an Ethernet controller, which is (IMO) a must when tasking a Pi with any server task.
• 1.2 and 1.3 revisions have the same BCM2837 SoC as the Pi 3, enabling 64 bit support, which gives it a brighter future along with some nice free performance improvements.
• The BCM2837 SoC also gives it some additional features such as USB and network boot, which are handy at times.

Sure, if WiFi, Bluetooth or other improvements introduced in later models may become relevant later, then the Pi 2B probably isn't the best choice.

OK, I did not realize about AVS, I know what can be done and how it works. I actually forgot about the 900Mhz w.r.t. Pi2B. I have a v1.1 and I could operate it from USB-serial console cable only (so with the 4th red wire also connected, on 5V). arm_freq=500 in config.txt, although I saw it is probably 600 running as minimum. 5V pin measured only 4.25V, permanent undervoltage, but it kept running whole night also ethernet operational.
With normal good 5V from a DC-DC module connected to USB-C PD 12V, it measured 1.8W idle. Now that is low, but note that my NanoPi-R6C idles at 1.5W, that is a thing faster than a RPi5. The example TVHeadend triggers me a bit as I put my TV-HAT on a RPi4, so it could do local transcoding of TV-channel content, using the on-chip HW codecs. All that is not there on older devices, so that is why I look at it a bit differently. It is tempting to hook up the TV-HAT to a RK3588 so I can do for sure transcode 3x 1080p50 in real time, but now that browsers play HEVC, it isn't worth the complex kernel driver hacking effort anymore.

I plan to do some network re-wiring and also solder some passive PoE things. I think I can quite easily measure what forced lower clock does (on a Pi3B, not so much the Pi3B+ as that comes with Gbit ethernet troubles (only 4 wires in cable).

redvli wrote: ↑

Mon Dec 01, 2025 5:15 pm

With normal good 5V from a DC-DC module connected to USB-C PD 12V, it measured 1.8W idle.

There must be something wrong here, as I see from here:
https://www.pidramble.com/wiki/benchmar ... onsumption
that it should be 1.1 or so.

I now have a nanopineo connected now, same way as Pi2B and idles at 0.84W. That is including a 50 cents DC-DC that does step-down 12V from USB powermeter via passive PoE (3 meters RJ45 cable) to 5V. Is also a 4x Cortex-A7 (AllwinnerH3).

My Pi2B is actually doing random reset/reboot when loaded a bit and supplied with stable/steady 5V, it seems in the past, something is damaged w.r.t. 1-wire GPIO, more I do not know. So that could explain a too high idle power consumption. I guess I first free up a RPi3B from existing enclosure and see how that behaves.

redvli wrote: ↑

Mon Dec 01, 2025 8:13 pm

There must be something wrong here, as I see from here:
https://www.pidramble.com/wiki/benchmar ... onsumption
that it should be 1.1 or so.

I now have a nanopineo connected now, same way as Pi2B and idles at 0.84W. That is including a 50 cents DC-DC that does step-down 12V from USB powermeter via passive PoE (3 meters RJ45 cable) to 5V. Is also a 4x Cortex-A7 (AllwinnerH3).

My Pi2B is actually doing random reset/reboot when loaded a bit and supplied with stable/steady 5V, it seems in the past, something is damaged w.r.t. 1-wire GPIO, more I do not know. So that could explain a too high idle power consumption. I guess I first free up a RPi3B from existing enclosure and see how that behaves.

I did some more tests and measurements. It is (and was actually) clear that extra voltage convertors and isolaters in active PoE modules are not without loss. I started with just including them, so not measuring as close as possible to the sink ( 5V 40-pin header) but at the source (12V battery), because that is what counts/matters. But now also did measure boards at microUSB, so just an USB cable and maybe some reverse polarity protection on the board, don't know exact components. So then the USB meter displays:
RPi3B RPiOS64 bookworm 6.12.x rpi kernel RF off: 1.26W
RPi2B Raspbian bookworm 6.12.x rpi kernel: 1.57W
NanoPiNEO Debian trixie armhf 6.16.x sunxi kernel: 0.57W

Use-case is idle, or better, basic capability to react to a ping via IP from just a LAN, so networking is up, it is not about off-line computing. I have not looked at optimizations like turn off LEDs. Also no forced low-clock as at least for the RPI3B measured via active PoE it made no difference, clock is already at 600 (is minimum). The NanoPiNEO is 480 minimum (1300 max/load). At least now clear that the own cheap DC-DC module causes about 0.2W loss.

But big question-mark is why is the RPi2B so high? It is a v1.1, so 4x Cortex-A7. Maybe someone can confirm the numbers. The pidramble site states 'nothing plugged into USB', so that means also no ethernet as that is USB connected. I know from long ago that such is killing for power optimizations but up to almost half a Watt seems much too high. That the RPi3B is lower could be in theory because it can do same work in less instructions and bus-width (more the externals) is anyway mostly the same, but also that is old know-how.