Re: 13.0-RELEASE Poor performance on Raspberry pi 3 B+

From: Mark Millard via freebsd-arm <freebsd-arm_at_freebsd.org>
Date: Thu, 25 Nov 2021 19:29:38 UTC
On 2021-Nov-25, at 09:55, Juan David Hurtado G <jdhurtado@orbiware.com> wrote:
> 
>> There are: v1.1, V1.2, and v1.4 revisions of that.
>> Which?
> 
> Raspberry Pi 3 Model B Plus Rev 1.3
> Raspberry Pi 4 Model B Rev 1.1

In a few days I likely will have access to a 4 GiByte RPi4B
Rev 1.1 .

>> This is not what I've been using (2021.04) but I no
>> longer remember whatever I knew about the differences
>> in behavior.
> 
> I've made a firmware update using `rpi-update` on the rpi 3 and
> `rpi-eeprom-update` on the rpi 4 (while on raspios).
> Same 0.4MHz so far on both of them when using FreeBSD.
> 
> As for the label on the SoC... I have heat sinks in both of them :)
> I wonder if there is a command that I can use to get that information
> without removing the heat sinks.

For the RPi4B: being an older variant (v1.1) would mean
2711ZPKFSB06BOT . 2711ZPKFSB06COT only showed up more
recently. (I'm not aware of any software query equivalent
to the labeling on the SOC. But not being Rev 1.4 (or later),
finding the label explicitly should not matter.)

For the RPi3B+: I'm not aware of ever needing to know any
labeling on the SOC.

I focus on the RPi4B's as I at least have access to a variant
and should soon have access to a match. No RPi3's currently
and I've never had access to a RPi3B+. There haev been past
reports of RPi3B+ oddities on the lists as I remember.

I normally do my own buildworld buildkernel and install
such. I build with tailoring, such as -mcpu-cortex-a72
for targeting RPi4B's and other Cortex-A72 systems. Thus
my normal context would not match in an audit vs. your
context. And I've picked other things like RPi* firmware
versions based on experiments. (Various releases have
historically been problematical.)

Other than running one USB related bug in FreeBSD via
my -mcpu=cortex-a72 builds, I've not had mismatches
in behavior in general.

When I've access to a 4 GiByte RPi4B Rev 1.1 I'll produce
a microsd card from an official 13.0-RELEASE image (so
no -p? at the time) and see what happens when using it to
boot the 4 GiByte RPi4B Rev 1.1 .

I'll look up the eeprom version in use in the RPi4B Rev 1.1
when I have access and report what it is. I definately
updated it in order to have the usb (no microsd card) style
of booting be possible. It will be another thing to compare.
(It is also something that an installed-files audit can
not cover.)

So far the one notable difference in our context's is
your context's U-Boot 2020.10 vs. the 2021.04 in my
context.

I can capture a binary byte sequence of the output of
a boot, such as verbose one. Looking for differences
in such might prove useful, given well matching contexts
if my results do not match yours for the frequency
involved.

===
Mark Millard
marklmi at yahoo.com
( dsl-only.net went
away in early 2018-Mar)