Is out there a new version of Raspberry Pi 3B?

[manuti]

Hey, first I want to clarify I'm not talking about the 3B+.
After talking with a colleague about an offert in a popular Chinese web store (I don't know if a can`t insert the link) with this images:

Anotación 2020-04-29 123753.jpg (53.12 KiB) Viewed 3013 times

Anotación 2020-04-29 123uu753.jpg (37.3 KiB) Viewed 3013 times

He point me about this "new" 3B. As you can see this 3B have the same thermal plate than 3B+ but not the wireless shield over there.

Thanks in advance for resolving this mystery.

[fruitoftheloom]

Hey, first I want to clarify I'm not talking about the 3B+.
After talking with a colleague about an offert in a popular Chinese web store (I don't know if a can`t insert the link) with this images:

He point me about this "new" 3B. As you can see this 3B have the same thermal plate than 3B+ but not the wireless shield over there.

Thanks in advance for resolving this mystery.

There are two revision of the 3B, though not sure from your pics what is the difference:

https://www.raspberrypi.org/documentati ... /README.md

[manuti]

Thanks!!!
I see the different code revision on 3B models, so I can suppose a new "secret" revision like the RPi2 or RPi1A+.

[HawaiianPi]

Yes, some 3B models were made with the B0 stepping SoC used on the 3B+ model. The chips are basically the same, except for physical packaging. The 3B models are all clocked at 1.2GHz, regardless of SoC stepping (so, no performance difference).

There is a slight difference in physical height with the B0 SoC, so it's something to be aware of when considering one of those heatsink cases for the 3B, as it may not be compatible with the taller SoC (the Flirc case is compatible with either, and includes thermal pads for each).

Note that later Pi 2B models were made with the 3B SoC previously, and that was a bigger difference (32 to 64 bit), even though they were clocked at the slower 2B speed. I haven't seen a 2B with the B0 stepping SoC, but it wouldn't surprise me if there were some. I imagine the old 3B SoC is no longer being produced and the newer one is being phased into the production lines.

[fruitoftheloom]

Thanks!!!
I see the different code revision on 3B models, so I can suppose a new "secret" revision like the RPi2 or RPi1A+.

Like the 2B the SoC may have changed between revisions (as above alluded to), but it is not a recent change, but to all intents and purposes it is not NEW, just revised.....

[HawaiianPi]

There are two revision of the 3B, though not sure from your pics what is the difference

The original 3B did not have the flip-chip SoC with heat spreader. It was just a black square slab.


OP's pics show one of the newer ones with the same B0 SoC used on the 3B+ models.

[fruitoftheloom]

There are two revision of the 3B, though not sure from your pics what is the difference

The original 3B did not have the flip-chip SoC with a heat spreader. It was just a black square slab (like the 2B SoC).


OP's pics show one of the newer ones with the B0 SoC used on the 3B+ models.

https://www.raspberrypi.org/products/ra ... 3-model-b/

There is a revision document regards the SoC on product page..

...should of checked that first

[manuti]

https://www.raspberrypi.org/products/ra ... 3-model-b/

There is a revision document regards the SoC on product page..

...should of checked that first

OOOOOH Many thanks!!!
I never gave any importance to those Notices links, but the devil is in the details.

► BCM2837 package change

So mystery solved: no new 3B model, only a revision and is quite old; November 2018.

SharedScreenshot.jpg (45.18 KiB) Viewed 2837 times

[fruitoftheloom]

https://www.raspberrypi.org/products/ra ... 3-model-b/

There is a revision document regards the SoC on product page..

...should of checked that first

OOOOOH Many thanks!!!
I never gave any importance to those Notices links, but the devil is in the details.

► BCM2837 package change

So mystery solved: no new 3B model, only a revision and is quite old; November 2018.
SharedScreenshot.jpg

There are 2 revision PDFs recent one is Dec 2019 !!!

BCM2837 package change

BCM2837 package change (2)

[davidcoton]

There are 2 revision PDFs recent one is Dec 2019 !!!

BCM2837 package change

BCM2837 package change (2)

And to quote from the second one:

This PCN just adds Made in Japan to the previously issued PCN (RPI-PCN-Pi3B-001) issued in November 2018 which covered this change.

IMHO, that is what gives PCNs a bad name. (I'm not disputing the necessity for them, even this one, but maybe it's why people like the OP ask here first and look for PCNs afterwards.)

[ejolson]

There are 2 revision PDFs recent one is Dec 2019 !!!

BCM2837 package change

BCM2837 package change (2)

And to quote from the second one:

This PCN just adds Made in Japan to the previously issued PCN (RPI-PCN-Pi3B-001) issued in November 2018 which covered this change.

IMHO, that is what gives PCNs a bad name. (I'm not disputing the necessity for them, even this one, but maybe it's why people like the OP ask here first and look for PCNs afterwards.)

Unfortunately PCNs in general seem to be unreliable sources of information.

In the computer and technology industry it is extremely important, in my opinion, for a manufacturer to be transparent about every change made to a product. This is because even slight changes can have unexpected compatibility issues that render a product previously working unfit for the same purpose.

An example of this is when Kingston switched the control chip of their DEC-compatible KNE110 Ethernet card to a cloned version and it started generating lockups under load on many dual socket Pentium III motherboards. At one point I had more than 30 such useless cards; I think Google had hundreds. Maybe they insisted on returning them. Soon after Kingston left the networking market, I think because the brand had been spoiled to the point that it would be impossible to make a profit.

Another example is when Western Digital started shipping hard disks which use shingled magnetic recording last year as Red NAS drivers. Due to the unannounced change, the new Red drives started failing in RAID arrays. Recently Western Digital has announced that their NAS drives were never meant to be used for RAID.

https://blocksandfiles.com/2020/04/14/w ... recording/

However, many people are wondering why would a drive that only works in single-bay systems even be called a NAS drive. Especially, when the identically labelled drive that it replaced worked fine.

With the Raspberry Pi people have been complaining that the 4B USB-C port was also changed without any notice. While that change seems not to have caused trouble, even changes for the better can cause unexpected consequences. In my opinion it is best to be fully transparent with each change, rather than to rely on luck. Of course, luck also helps.

Back on topic, that's interesting that the Pi 3B is still being manufactured using the newer SOC package. Thanks for the long-time support of all models.

[laurent]

I haven't seen a 2B with the B0 stepping SoC, but it wouldn't surprise me if there were some. I imagine the old 3B SoC is no longer being produced and the newer one is being phased into the production lines.

It's exactly what I thought, when I read this topic
Is the Raspberry Pi model 2B will compete with the old venerable Socket 7

[jamesh]

There are 2 revision PDFs recent one is Dec 2019 !!!

BCM2837 package change

BCM2837 package change (2)

And to quote from the second one:

This PCN just adds Made in Japan to the previously issued PCN (RPI-PCN-Pi3B-001) issued in November 2018 which covered this change.

IMHO, that is what gives PCNs a bad name. (I'm not disputing the necessity for them, even this one, but maybe it's why people like the OP ask here first and look for PCNs afterwards.)

Unfortunately PCNs in general seem to be unreliable sources of information.

In the computer and technology industry it is extremely important, in my opinion, for a manufacturer to be transparent about every change made to a product. This is because even slight changes can have unexpected compatibility issues that render a product previously working unfit for the same purpose.

An example of this is when Kingston switched the control chip of their DEC-compatible KNE110 Ethernet card to a cloned version and it started generating lockups under load on many dual socket Pentium III motherboards. At one point I had more than 30 such useless cards; I think Google had hundreds. Maybe they insisted on returning them. Soon after Kingston left the networking market, I think because the brand had been spoiled to the point that it would be impossible to make a profit.

Another example is when Western Digital started shipping hard disks which use shingled magnetic recording last year as Red NAS drivers. Due to the unannounced change, the new Red drives started failing in RAID arrays. Recently Western Digital has announced that their NAS drives were never meant to be used for RAID.

https://blocksandfiles.com/2020/04/14/w ... recording/

However, many people are wondering why would a drive that only works in single-bay systems even be called a NAS drive. Especially, when the identically labelled drive that it replaced worked fine.

With the Raspberry Pi people have been complaining that the 4B USB-C port was also changed without any notice. While that change seems not to have caused trouble, even changes for the better can cause unexpected consequences. In my opinion it is best to be fully transparent with each change, rather than to rely on luck. Of course, luck also helps.

Back on topic, that's interesting that the Pi 3B is still being manufactured using the newer SOC package. Thanks for the long-time support of all models.

You won't be surprised to hear that we disagree. We've been releasing new version of Pi boards without announcement for years, and no-one has noticed.

The real changes are things like the PI3->PI4, which was a major change which we obviously do announce! Point releases of hardware are bug fixes or cost reduction exercises. Of course, eventually we will probably have more complicated changes at point releases that do require documenting, but I don't think it's happened yet. Just make sure your firmwares are up to date, and you'll never notice the HW changes! It's a bit like changing memory chip suppliers - we did that a lot in the early days, each memory chip change required new firmware. No-one noticed.

[manuti]

I started this post because even if the change is old or they made more than one review. The board is mechanically different and can compromise some case design.
I'm not complaining about the review. I don't recommend buy from Chinese web shops.
Thanks for all the information. I'll try to search more deeply before posting new questions here.

[davidcoton]

We've been releasing new version of Pi boards without announcement for years, and no-one has noticed.
...
It's a bit like changing memory chip suppliers - we did that a lot in the early days, each memory chip change required new firmware. No-one noticed.

Well, the "point" revision of the Pi2B when it went from 32-bit to 64-bit definitely was announced. I would argue that it was significant enough to need an announcement (and clearly you (RPF/T collectively) agreed). Though there is no PCN on the product page...

As for the memory chip changes, they were noticed and occasionally discussed here. Mainly because people didn't know there was a need to update firmware? (I may be wrong. Bit-rot in my brain.)

I just hope that the decision not to label the different board revisions does not come back to bite as a severe support nightmare. I do not understand the reasons behind the decision, but I do not challenge your right to make it.

EDIT: Correction as per this post

[ejolson]

You won't be surprised to hear that we disagree. We've been releasing new version of Pi boards without announcement for years, and no-one has noticed.

The real changes are things like the PI3->PI4, which was a major change which we obviously do announce! Point releases of hardware are bug fixes or cost reduction exercises. Of course, eventually we will probably have more complicated changes at point releases that do require documenting, but I don't think it's happened yet. Just make sure your firmwares are up to date, and you'll never notice the HW changes! It's a bit like changing memory chip suppliers - we did that a lot in the early days, each memory chip change required new firmware. No-one noticed.

The policy of many big companies is to not inform consumers unless there is an advertising advantage but to carefully keep their business partners informed. For example, the major cloud providers appear to have been informed which drives were no longer suitable for use in network attached storage appliances. At the same time, consumer branding continued to say NAS on the label.

In that case, consumers still finding out wasn't just a stroke of bad luck, but due to very different quality being provided without warning or notice. The resulting lawsuits, bad publicity and loss of trust would have been good to avoid.

[DougieLawson]

The only time you'll need to worry is if your software is sensitive to /proc/cpuinfo "Revision" code.
https://www.raspberrypi.org/documentati ... /README.md

If you get a new revision code you'll need to update your lookup tables. That includes the kernel and bootcode/eeprom stuff.

The main thing affected by that, that the general population will notice is the, now deprecated, wiringPI. Any new revsion code will break V2.52 as that's the last ever version Gordon is going to publish (if he sticks to his deprecation). There's likely other published code out there that relies on /proc/cpuinfo that will break.

Anything else is likely to be component sourcing related and insignificant board changes that have no effect on how the board runs and doesn't need new firmware.

[W. H. Heydt]

The real changes are things like the PI3->PI4, which was a major change which we obviously do announce! Point releases of hardware are bug fixes or cost reduction exercises. Of course, eventually we will probably have more complicated changes at point releases that do require documenting, but I don't think it's happened yet. Just make sure your firmwares are up to date, and you'll never notice the HW changes! It's a bit like changing memory chip suppliers - we did that a lot in the early days, each memory chip change required new firmware. No-one noticed.

As I recall, people *did* notice because they would tend to move SD cards from an older board to a newer one of the "same model" and when RAM chips suppliers changed, things broke.

While I am fine with introducing minor changes without announcing them--such as the v1.2 Pi4B, switching the SoC on the Pi3B, and (one might presume) on the Pi2Bv1.2--I would really prefer a version number on the board so the less obvious changes (e.g. NOT the stepping change of the BCM2837, which you can't miss) are distinguishable. Even with something as beneficial and benign as the Pi4Bv1.2, anyone of the people that would--for some unfathomable reason--prefer to use a PD-rated PSU or e-marked USB power cable should be able to just look at the board, read off the version number and know if their fancy PSU will work.

What it comes down to is a request that if either the PCB gets changed, or something of significance mounted on the board gets changed, there should be a change to the silk-screen label as well. This would make support here on the forums much easier in any number of cases because we could just as, "What model and version number is your Pi?" and the answer would go far to aid in solving problems.

[HawaiianPi]

I started this post because even if the change is old or they made more than one [revision]. The board is mechanically different and can compromise some case design.

I doubt that affects many users (but I could be wrong). Not sure how popular heatsink cases are, and how many of them are incompatible with the taller SOC. As I said above, the Flirc comes with a thinner thermal pad for the 3B+ and 3B r1.3. Two piece designs might work similarly, and tighter fitting ones could probably be fitted with some spacers (washers) on the mounting screws. The main thing is to be aware of it to avoid potential damage with a too tight heatsink case.

As far as RPF is concerned, they don't make cases like that, and have no control over third party accessories, so the fit of anyone else's product is the responsibility of the maker (RPF can't monitor and test fit every product out there, so the manufacturer of the product has the responsibility of testing or warning buyers about compatibility issues). I know there will be some who disagree, but whatever, you can't please everyone.

Thanks for all the information. I'll try to search more deeply before posting new questions here.

No problem. This is a relatively obscure thing to look up, and your question probably helped get the info out there to others who didn't know about this.

[HawaiianPi]

What it comes down to is a request that if either the PCB gets changed, or something of significance mounted on the board gets changed, there should be a change to the silk-screen label as well.

That ends up being more expensive, so would you still want that if it required a price increase?

How is it more expensive? Let's use the Pi 4B as an example. With all the click-bait press coverage of the USB-C PD issue, you know there are those who absolutely refuse to buy an older model, and dealers won't want to have to open loads of boxes to find a particular revision, so every time a revision change is made, RPF/RPT would need to print new packaging, and discard all the older packaging already printed. Sure, labels could reduce that cost somewhat, but it's still a hassle and additional expense that isn't realistic in the quantities sold. And that assumes most dealers would even bother to hand pick specific revisions for customers.

There are lots of things that would be nice, but are they realistic or beneficial to the goals of RPF?

[W. H. Heydt]

What it comes down to is a request that if either the PCB gets changed, or something of significance mounted on the board gets changed, there should be a change to the silk-screen label as well.

That ends up being more expensive, so would you still want that if it required a price increase?

How is it more expensive? Let's use the Pi 4B as an example. With all the click-bait press coverage of the USB-C PD issue, you know there are those who absolutely refuse to buy an older model, and dealers won't want to have to open loads of boxes to find a particular revision, so every time a revision change is made, RPF/RPT would need to print new packaging, and discard all the older packaging already printed. Sure, labels could reduce that cost somewhat, but it's still a hassle and additional expense that isn't realistic in the quantities sold. And that assumes most dealers would even bother to hand pick specific revisions for customers.

There are lots of things that would be nice, but are they realistic or beneficial to the goals of RPF?

And here I thought you were going to cite the cost of making new silk-screen masters...

People have already--and for some time when changes were made--returned older versions because they wanted the newer one. That is a cost distributed through the supply chain, plus it is negative public relations for the RPF, a cost in itself.

In one case, after looking at an on-line ordering web page, I filed a question with the vendor to determine which variant of a board they were actually selling because the image presented and the text were inconsistent. Just answering the question cost them money...and lost them money because the boards they actually had weren't the ones I wanted.

For many changes--e.g. Pi4Bv1.1 vs. Pi4Bv1.2--it makes no practical difference to me. For others--e.g. Pi2Bv1.1 vs. Pi2Bv1.2--it does make a difference. Even once I have any given board, I would still like to be able to just glance and the board to see which version it is, as that may make a difference for a given intended use.

I'm not asking that the RPF/RPT change their policies with regard to what is worth making an announcement about and what is not, just that PCB revisions be included on the board. If you can boot it up and find a different version number, I'd like to have that indicated with booting it first.

[ejolson]

I doubt that affects many users (but I could be wrong). Not sure how popular heatsink cases are, and how many of them are incompatible with the taller SOC. As I said above, the Flirc comes with a thinner thermal pad for the 3B+ and 3B r1.3. Two piece designs might work similarly, and tighter fitting ones could probably be fitted with some spacers (washers) on the mounting screws. The main thing is to be aware of it to avoid potential damage with a too tight heatsink case.

As far as RPF is concerned, they don't make cases like that, and have no control over third party accessories, so the fit of anyone else's product is the responsibility of the maker (RPF can't monitor and test fit every product out there, so the manufacturer of the product has the responsibility of testing or warning buyers about compatibility issues). I know there will be some who disagree, but whatever, you can't please everyone.

A company that informs their business partners about changes likely to impact the partner's business ahead of time may create a cooperative environment beneficial to both. On the other hand, disruptions that could have easily been avoided by communication eventually lead to loss of trust.

In the same way the manufacture of Raspberry Pi computers depends on part suppliers that don't unexpectedly ship a different item as the same, so the manufacture of other devices may depend on Pi computers, for example, that don't unexpectedly stop booting from a tested image that used to work or fit in a carefully designed case.

When the 2GB Odriod N2 switched to using half as many but higher density RAM chips, a notice was provided ahead of time. In this case, the notification was particularly admirable in that it was not self serving but acknowledged there was also a decrease in performance to keep costs down.

If Western Digital had done the same with their NAS drives, there wouldn't be small companies and individuals who spent a noticeable portion of their IT budget on an item unfit for purpose.

The argument I've seen made is, though small companies and individuals spend less on disk drives in total, the amount they do spend accounts for a greater percentage of their available budget. Thus, the actual damage done to a small businesses is proportionately more than if a similar change were to affect the hyperscalers and cloud providers, which of course didn't happen.

From a Pi perspective, the moral of the story is that people buying cheap stuff are often spending a larger percentage of their budget to do so and consequently are more vulnerable when unexpectedly things don't work as they used to. Hoping they won't notice goes against the ideal of creating a computer literate workforce.

[HawaiianPi]

Ah well, like I said, you can't please everybody.

To be clear, I would like more transparency as well, but I also understand why it isn't always practical.

[fruitoftheloom]

Ah well, like I said, you can't please everybody.

To be clear, I would like more transparency as well, but I also understand why it isn't always practical.

A revision on the board would I believe be beneficial, earlier Pis had that.....

[W. H. Heydt]

Well, the "point" revision of the Pi2B when it went from 16-bit to 32-bit definitely was announced.

Say what? The Pi2Bv1.1 uses 32-bit cores. The Pi2Bv1.2 uses the same 64-bit capable cores that the Pi3B used. The point was to drop the production of the BCM2836. And--at the time--there wasn't any announcement. That came after people noticed the Pi2Bv1.2 in the wild.