Not all RPI 4's are equal?

[Aardappeltaart]

The first thing I did after getting my new RPI 4 was connecting it and trying out some simple commands, and yes also measure temperature.

vcgencmd measure_temp
61°

Wow, that's hot for being idle. But it was confirmed by some reviews, while others seem to have lower temperatures.

Then I read this post:
https://www.cnx-software.com/2019/06/29 ... mperature/

It seems not all RPI 4's run the same firmware by default, early production units have older firmware, lacking some thermal optimizations.

Can someone elaborate on which/how much models are `early production units` and how they can be identified?

And are there more differences? (When I compare my model to the Explaining Computers review, my model does not have any text printed on the ethernet connector.)

Any idea when and how the firmware will be updated?

[W. H. Heydt]

There are a couple of issues here. The first is, that by the very nature of things, chips vary some from one to the next. The factory tests them to meet basic requirements (will it run at the specified clock speed? do all functions work correctly? can you read and write all the cache? stuff like that). Some chips can be overclocked further than others, for instance. And pertinent you your questions, some will run hotter than others.

As regards the firmware... The Pi4B has updateable firmware. I've seen references to people getting firmware updates and applying them, so that functionality already exists as a practical process. I suspect that, going forward, along with doing update/upgrade setup will include getting and writing the latest firmware. Not sure if all of that is in place yet without using bleeding edge versions.

[Aardappeltaart]

There are a couple of issues here. The first is, that by the very nature of things, chips vary some from one to the next. The factory tests them to meet basic requirements (will it run at the specified clock speed? do all functions work correctly? can you read and write all the cache? stuff like that). Some chips can be overclocked further than others, for instance. And pertinent you your questions, some will run hotter than others.

Sure, but that doesn't explain shipping units with different firmware. In Eben Upton words: “some thermal optimizations that are not installed by default on early production units.”

It would be nice for buyers to know whether they have a somewhat lower quality batch unit, or just old firmware, before they rush out to buy heatsinks or fans.

[fruitoftheloom]

There are a couple of issues here. The first is, that by the very nature of things, chips vary some from one to the next. The factory tests them to meet basic requirements (will it run at the specified clock speed? do all functions work correctly? can you read and write all the cache? stuff like that). Some chips can be overclocked further than others, for instance. And pertinent you your questions, some will run hotter than others.

Sure, but that doesn't explain shipping units with different firmware. In Eben Upton words: “some thermal optimizations that are not installed by default on early production units.”

It would be nice for buyers to know whether they have a somewhat lower quality batch unit, or just old firmware.

All Raspberry Pis are fully tested, there is no such thing as lower quality, it is insult to state that....


All previous models of Pis had kernel / firmware updates, the 4B will also get kernel / firmware updates during its life cycle.


Just have to patient and wait for updates as and when RPF / RPT feel they are fit for public consumption.

[W. H. Heydt]

It would be nice for buyers to know whether they have a somewhat lower quality batch unit, or just old firmware, before they rush out to buy heatsinks or fans.

As noted, early firmware is NOT "lower quality". It's just earlier firmware. For the first time, the firmware is field upgradeable, so it isn't even a case of being stuck with early firmware. (Want to take a look at my very early RPF 7" screen that will only do the backlight on or off?)

Given the rate at which stock is selling out, if you order now (let alone later), you are extremely unlikely to get that early firmware. They've all gone by now.

And that makes an interesting point... Anybody got a quick, easy way to check the EEPROM program version and date?

[Aardappeltaart]

And that makes an interesting point... Anybody got a quick, easy way to check the EEPROM program version and date?

He, that's was my question in OP: `and how they can be identified?`

[jerrm]

All Raspberry Pis are fully tested, there is no such thing as lower quality, it is insult to state that....

That just isn't true.

Over the years at least a dozen Pis have failed my memory bench tests. Early Pi 3's were the worst. Outside of that period only a two or three.

All tests were done at stock settings - no overclocking or other tricks, using memtester from the repos.

[Aardappeltaart]

All Raspberry Pis are fully tested, there is no such thing as lower quality, it is insult to state that....

I'm not stating that. I was just asking about different firmware shipped. W. H. Heydt made the suggestion that the thermal (throttling) issues could be caused by some chips simply running hotter than others.

Just have to patient and wait for updates as and when RPF / RPT feel they are fit for public consumption.

Isn't the time fit to explain that some models were shipped with old firmware and that owners of those models can experience thermal issues for the time being? To be clear we're talking about 20% performance difference (throttling) in some benchmarks (7-zip). See link OP.

[energyi]

How they can be identified?`

[FTrevorGowen]

How they can be identified?`

Just to clarify perhaps: there's a difference between "identifying" the bootloader version installed in a given P4B's eeprom (which will need to be updated a some point by those folks wanting to USB boot when that's available in any case) or, what is usually called "firmware" (for all Pi's) which is on the uSDHC card, or identifying a P4B with a specific bootloader and/or firmware. Most, if not all, "firmware" updates (once stable) usually happen along with O.S. updates for all Pi's AFAIK.
Trev.

[HawaiianPi]

The command below should update the OS, installed software, kernel and firmware to the latest version.

Code: Select all

sudo apt update && sudo apt full-upgrade -y

I'm not sure if that includes the bootloader EPROM.
To reinstall (and presumably update) the bootloader EPROM see here: https://www.raspberrypi.org/documentati ... teeprom.md

[Timescale]

My first (headless/no desktop) test unit is currently running 65'ish degrees science when idle. A bit toasty, but still fine really. It certainly is good enough for test purposes before deployment and I expect that new firmware or other measures will be available then.

I mean, surely nobody is going to deploy new hardware in the field less then a week after release?

Eventually I suspect these units will fit the temperature profile of the 3B more, perhaps with some heat sinks added, but at this point in the game, I really do not see any problem. In fact, in testing having it throttle makes for a nice test-case in how an application handles in that scenario. I could never get my 3B+'s to do that!

Regardless of the thermal impact, we are still getting at least 2.5 times the performance compared to the 3B+. Lets just work with what we got and see what comes up in the following weeks.

[chwe]

There are a couple of issues here. The first is, that by the very nature of things, chips vary some from one to the next. The factory tests them to meet basic requirements (will it run at the specified clock speed? do all functions work correctly? can you read and write all the cache? stuff like that). Some chips can be overclocked further than others, for instance. And pertinent you your questions, some will run hotter than others.

As regards the firmware... The Pi4B has updateable firmware. I've seen references to people getting firmware updates and applying them, so that functionality already exists as a practical process. I suspect that, going forward, along with doing update/upgrade setup will include getting and writing the latest firmware. Not sure if all of that is in place yet without using bleeding edge versions.

if you would have 5°C differences of temperature only by different quality of the components on your SBC, you should hire a new qc employee... On default settings under the same conditions there shouldn't be much a difference.. Otherwise something is fishy. Obviously if we talk about overclocking chips may differ.. out of specification, not the hardwaremakers responsibility to ensure all can be similar overclocked..

as for the firmware upgrade, I'm quite sure an apt update/upgrade wont change anything here it's generally not a good idea, if something goes wrong you can easy brick thousands of devices within a few days (yeah they can be unbricked by the special recovery binary but you want be responsible for that).. I'm quite sure the famous 'rpi-update' will trigger a 'firmware' update.

Sure, but that doesn't explain shipping units with different firmware. In Eben Upton words: “some thermal optimizations that are not installed by default on early production units.”

It would be nice for buyers to know whether they have a somewhat lower quality batch unit, or just old firmware, before they rush out to buy heatsinks or fans.

Probably they started with the production a bit earlier cause they knew they will sell a bunch of boards once the RPi is announced? Likely that the software wasn't really ready at this time (it's not even fully ready yet, otherwise USB-boot and PXE would work, but okay for the average use-cases)? You want unpack thousands of boards only to send them with the newest 'firmware' and risk a even bigger delay.

As noted, early firmware is NOT "lower quality". It's just earlier firmware. For the first time, the firmware is field upgradeable, so it isn't even a case of being stuck with early firmware. (Want to take a look at my very early RPF 7" screen that will only do the backlight on or off?)

What else? If the temperature lowers 3-5°C between before and after firmware upgrade without any other significant drawbacks, the older firmware is by definition of lower quality.. Doesn't mean that the hardware is of lower quality, but the product at all is definitively. If it is upgradable, it's not much an issue but you probably agree that you would prefer the board with the new firmware over the one with the old, don't you?
And if you read the CNX post which the OP mentioned.. it's likely that all/most current boards are more or less affected by it.
https://www.cnx-software.com/2019/06/29 ... mperature/

The guys at the Raspberry Pi Foundation somehow noticed my post, and I received an email from Eben Upton explaining a new Raspberry Pi 4 VLI firmware had “some thermal optimizations that are not installed by default on early production units.”

...

The Raspberry Pi Foundation provided me with a test version of the firmware, which they’ll release in the next few days, or weeks after testing is completed.

otherwise, this new firmware would already be out and somewhere stated how to upgrade it...

[W. H. Heydt]

How they can be identified?`

Just to clarify perhaps: there's a difference between "identifying" the bootloader version installed in a given P4B's eeprom (which will need to be updated a some point by those folks wanting to USB boot when that's available in any case) or, what is usually called "firmware" (for all Pi's) which is on the uSDHC card, or identifying a P4B with a specific bootloader and/or firmware. Most, if not all, "firmware" updates (once stable) usually happen along with O.S. updates for all Pi's AFAIK.
Trev.

I think we're going to run into a terminology issue with the Pi4B (and, most likely, subsequent Pis). We have the ROM bootloader that is part of the SoC and will only change if there is a version step. It is, for all intents and purposes, fixed. There is the EEPROM--the contents of which is what I would refer to as "firmware"--which can be reloaded. Then there is the binary blob, which heretofore has been referred to as "firmware", but is actually just a file on the boot media. So... Do we now have "hard firmware" (the EEPROM contents) and "soft firmware" (the binary blob)?

[milhouse]

And that makes an interesting point... Anybody got a quick, easy way to check the EEPROM program version and date?

He, that's was my question in OP: `and how they can be identified?`

SPI Bootloader version (this is the current "production" version):

Code: Select all

rpi4:~ # vcgencmd bootloader_version
May 10 2019 19:40:36
version d2402c53cdeb0f072ff05d52987b1b6b6d474691 (release)

[W. H. Heydt]

And that makes an interesting point... Anybody got a quick, easy way to check the EEPROM program version and date?

He, that's was my question in OP: `and how they can be identified?`

SPI Bootloader version (this is the current "production" version):

Code: Select all

rpi4:~ # vcgencmd bootloader_version
May 10 2019 19:40:36
version d2402c53cdeb0f072ff05d52987b1b6b6d474691 (release)

Thank you.

[Aardappeltaart]

People are experiencing thermal issues
The official case could be dubbed prefab sauna, raising the temperature to 70° idle.

SPI Bootloader version (this is the current "production" version):

Code: Select all

rpi4:~ # vcgencmd bootloader_version
May 10 2019 19:40:36
version d2402c53cdeb0f072ff05d52987b1b6b6d474691 (release)

Thank you.

AFAIK new firmware (VIA VL805 PCIe USB 3.0) enables ASPM (Active-State Power Management) for the USB controller. All from the post linked in OP.
Old firmware:

Code: Select all

sudo lspci -vvv | grep ASPM
--
LnkCtl: ASPM Disabled; RCB 64 bytes Disabled- CommClk-

New firmware:

Code: Select all

sudo lspci -vvv | grep ASPM
--
 LnkCtl: ASPM L0s L1 Enabled; RCB 64 bytes Disabled- CommClk+

[HiassofT]

To clarify things: RPi4 not only has an upgradeable bootloader in an SPI EEPROM but also the VIA USB3 controller has upgradeable firmware in an EEPROM - the latter is what the CNX article is talking about.

I also did some quick tests with the newer USB controller firmware (which is currently in internal testing) and could verify that it reduces power consumption by about 0.3W.

I tested with LibreELEC and Kodi idling on the home screen consumed about 3.3W with the current release USB3 firmware, with the updated one it dropped to about 3W.

10% less power consumption means about 10% less heat has to be dissipated which is nice and helps to run the RPi a bit cooler.

But the important thing now in testing is to find out if the updated firmware works fine and doesn't introduce any regressions - when that's finished RPi folks will probably officially release the firmware and updater tool.

so long,

Hias

[jamesh]

That article pretty much explains it. But note, the new USB device firmware has not been released yet, so at the moment all Pi4 are indeed the same.

It should help greatly when using a case, the power savings are up to 300mA.

[lb]

300 mA? That's crazy. Now I really want to try this. So if you need some users to test this...

[HiassofT]

300mW, not 300mA so about 60mA less.

These are the screenshots from my bench top power supply, setup to output 5.1V:

current release firmware:

00013701.png (3.56 KiB) Viewed 4440 times

updated testing firmware:

000137a8.png (3.58 KiB) Viewed 4440 times

so long,

Hias

[jamesh]

300mW, not 300mA so about 60mA less.

These are the screenshots from my bench top power supply, setup to output 5.1V:

current release firmware:
00013701.png

updated testing firmware:
000137a8.png

so long,

Hias

Sorry, always getting my Watts and amps mixed up.

[Aardappeltaart]

That article pretty much explains it. But note, the new USB device firmware has not been released yet, so at the moment all Pi4 are indeed the same.

Thx, but that contradicts Eben Upton words “some thermal optimizations that are not installed by default on early production units.”

And it doesn't explain some users experiencing higher idle temperatures (~5°) than others

Are you're sure all `official reviewers` got units with old firmware shipped?

Why then would Eben Upton drop his fork on his first family diner on his first free evening in months (well deserved) to send an email and experimental firmware to a reviewer that published disappointing benchmarks. Because he likes cold soup?

(Never release a Pi in the summer)

[jamesh]

Not sure what Eben was talking about specifically, but I know that the vli firmware has not been released.

Reviewers have had devices for some time. In that time many things have changed, bootloader, kernel etc. But I don't think vli changes were sent out.

It may be just this one guy was sent it. I don't know.

As for idle temperature variations, depends what the device is doing while idling. Desktop? Console? 4kp60?

[Aardappeltaart]

Not sure what Eben was talking about specifically, but I know that the vli firmware has not been released.

Reviewers have had devices for some time. In that time many things have changed, bootloader, kernel etc. But I don't think vli changes were sent out.

It may be just this one guy was sent it. I don't know.

As for idle temperature variations, tdepend what the device is doing while idling. Desktop? Console? 4kp60?

Well, the important thing is: improvements are on the way.