Power consumptions of AM62D and AM62 at idle state

Hello Toradex Team,

We have an AM62D ET 1GB and an AM62 512MB to evaluate and wanted to measure power consumption of both at idle state (tdx-reference-minimal-image) with Mallow V1.1A carrier board.

Our measurements show that

  • AM62D ET 1GB: idle power in Mallow V1.1A = 1.1W Ethernet LEDs are on, LED4 green and red are on
  • AM62 512MB: idle power in Mallow V1.1A = 1.4W Ethernet LEDs are off, LED4 green and red are off

It seems single core consumes 0.3W higher than dual core version while we expected it to consume less about 0.1W.

AM62D ET 1GB
Software summary
------------------------------------------------------------
Bootloader:               U-Boot
Kernel version:           6.1.46-6.5.0-devel+git.36991ce6ffec #1 SMP PREEMPT Fri Oct 13 12:50:57 UTC 2023
Kernel command line:      root=PARTUUID=1360bcbd-02 ro rootwait console=tty1 console=ttyS2,115200 consoleblank=0 earlycon=ns16550a,mmio32,0x02800000
Distro name:              NAME="TDX Wayland with XWayland"
Distro version:           VERSION_ID=6.5.0-devel-20231205124640-build.0
Hostname:                 verdin-am62-15190493
------------------------------------------------------------
 
Hardware info
------------------------------------------------------------
HW model:                 Toradex Verdin AM62 on Verdin Development Board
Toradex version:          0073 V1.1A
Serial number:            15190493
Processor arch:           aarch64
------------------------------------------------------------
AM62 512MB
Software summary
------------------------------------------------------------
Bootloader:               U-Boot
Kernel version:           6.1.46-6.6.0-devel+git.278a5418c2ba #1 SMP PREEMPT Fri Feb 23 08:19:02 UTC 2024
Kernel command line:      root=PARTUUID=0e3e05bc-02 ro rootwait console=tty1 console=ttyS2,115200 consoleblank=0 earlycon=ns16550a,mmio32,0x02800000
Distro name:              NAME="TDX Wayland with XWayland"
Distro version:           VERSION_ID=6.6.0-devel-20240305091311-build.0
Distro variant:           -
Hostname:                 verdin-am62-15136132
------------------------------------------------------------
 
Hardware info
------------------------------------------------------------
HW model:                 Toradex Verdin AM62 on Verdin Development Board
Toradex version:          0071 V1.1A
Serial number:            15136132
Processor arch:           aarch64
------------------------------------------------------------

Any idea why it could be?

Thank you,
Fide.

Hello @Fide,

From our tests, the 0073 Verdin AM62 Dual 1GB ET V1.1A consumes about 0.1W more than the 0071 Verdin AM62 512GB V1.1A at idle.

There are a few things that could explain the difference you are seeing:

  • Different BSP versions: You are using different nightly builds in each of your modules. It would be recommended to use the same BSP version on both devices to have a better comparison.
  • Different idle conditions: as in one case the LEDs are on and in the other they are off, there may be mismatched configurations between the modules.
  • Silicon variance: Not all SoCs are made equally, so different SoCs of the exact same model may have different power consumption.

How are you measuring the power consumption of the modules?
Does your measurement also include the power consumption from the Mallow Carrier Board?

Best Regards,
Bruno

Hi @bruno.tx,

Thank you for the suggestions.

For the first point:
We have tested both SOMs using same BSP 6.5.0-devel at the beginning and observed the same problem. (AM62D consumes less but AM62 512 MB consumes more and led status were as reported). Then, because AM62 512 MB is a new module, we thought that there may be some updates for the single core and we decided to upgrade to BSP 6.6.0-devel and repeated the test but results didn’t change.

For the second point:
LED status are different based on modules, we don’t know why. At the end we use same tdx-reference-minimal-image for both SoMs.

For the third point:
The variation observed is about 30%. 0.3W at 1.8V is about 167mA difference. I’m not sure if it can vary from part to part that much.

According to our hardware team, the measurement done from the power input of Mallow carrier board. So the carrier board consumption is included in our report. One note that, we use same Mallow carrier board for our tests, just replaced the SoM on it.

Regards,
Fide.

Hello @Fide,

Thanks for the clarifications and sorry for the delayed response.

We did some additional tests and we saw a higher power consumption on the Verdin AM62 Solo 512MB paired with a Mallow Carrier board has a higher power consumption at idle. The difference we saw was smaller, but the overall idle power consumption was higher:

  • 00711100 - Verdin AM62 Solo 512MB V1.1A + 01611100 - Mallow V1.1A: ~1.7W
  • 00731100 - Verdin AM62 Dual 1GB ET V1.1A + 01611100 - Mallow V1.1A: ~1.6W

It is important to highlight that these results are valid only for the specific hardware used on the version 6.5.0 of the Reference Minimal Image. Further updates may bring changes.

If you require a lower idle power consumption, there are many steps that could be taken to achieve that:

Best Regards,
Bruno

Hi @bruno.tx,

Thank you for the update.

For our new product that will be designed this year, we need to pick a SOM suitable for it.
Because the product is battery powered, the SOM consumption is one of the critical parameters for us after the performance.

We evaluated the performance of AM62 Solo and AM62 Dual. And they are both ok for the project. Second important parameter is idle power consumption. We thought that Solo should consume less comparing to Dual but we couldn’t confirm this on Mallow board.

Since performance wise both Solo and Dual are ok for the project, if we look for smaller idle power consumption values, shall we go with Dual or Solo?

Thank you.

Hello @Fide,

Unfortunately, I cannot make a recommendation for your project.
More information would be needed to really understand what may or may not impact the general power consumption of your application.

Our measurements of SoM power consumption are made under very specific conditions which may not directly translate to different scenarios.

When we are talking about small differences, the best way to know if the power consumption of a given module is adequate for your application is to actually test it.

Best Regards,
Bruno