Unable to Flash and Apply Different Power Mode Configurations on Jetson Orin

[sunil.ap]

Hi NVIDIA Team,

We are flashing Jetson Orin devices using different .conf files (e.g., default and super configurations) to evaluate the performance under various power modes (MAXN, 10W, 15W, 25W).

However, after flashing with different config files, we observe that the system always shows the same set of power modes (MAXN, 10W, 15W, 25W) — even when using a config intended for MAXN_SUPER. It seems like the system is only applying the default configuration, regardless of the .conf file used.

Could you please guide us on:

• How to properly flash and apply a specific config file (e.g., super-mode .conf) so that the corresponding performance profiles are actually enabled after boot?
• How to confirm at runtime that the intended config (e.g., super) is applied correctly?

Could you please help us to verify and compare performance across different flash configurations.

Best regards,
Sunil

@DaneLLL @WayneWWW

[WayneWWW]

How did you flash your board?

[sunil.ap]

Hi Wayne,

We’re using the NVIDIA-provided config:

jetson-orin-nano-devkit-super-maxn.conf

And flashing the board with the following command:

sudo ./tools/kernel_flash/l4t_initrd_flash.sh --external-device nvme0n1p1 -c tools/kernel_flash/flash_l4t_t234_nvme.xml -p “-c bootloader/generic/cfg/flash_t234_qspi.xml” --showlogs --network usb0 jetson-orin-nano-devkit-super-maxn internal

The flash completes successfully, but after boot:

• We still see the same power mode list as the default config: MAXN, 10W, 15W, 25W
• There’s no clear indication that MAXN_SUPER or higher frequency profiles are active
• We want to compare performance between default and super configs, but it’s unclear if the custom DTBs and clock settings from the super config are being applied

Could you please confirm:

• Is there any post-flash verification step or setting to activate MAXN_SUPER?
• How can we ensure the super config is active at runtime?
• Are additional changes to nvpmodel or device tree overlays needed?

Regards,
Sunil A P

@WayneWWW

[WayneWWW]

please check your tegrastats or jtop result after you run the jetson_clocks.

[sunil.ap]

Hi Wayne,

We rechecked after flashing with the jetson-orin-nano-devkit-super-maxn.conf config and running jetson_clocks.

Here’s what we’ve observed:

• The power mode list remains the same (MAXN, 10W, 15W, 25W)
• Even in MAXN, we do not see clear evidence that MAXN_SUPER limits (e.g. higher CPU/GPU/EMC clocks) are applied
• Screenshot of the current GUI power menu is attached for reference
• jtop still showed Jetson Clocks = Inactive before manually running jetson_clocks

Our goal is to verify performance differences between default and super configurations, but the visual and runtime outputs appear identical so far.

Could you help clarify:

1. Should we expect MAXN_SUPER to appear explicitly in the UI or through nvpmodel -q?
2. Are there any additional steps or indicators that confirm the super config is actually active?

We’re happy to share logs or outputs (nvpmodel -q, tegrastats, etc.) if needed.

Best regards,
Sunil

@WayneWWW

[WayneWWW]

Hi,

Please share us the result of tegrastats too. That will be more precise than jtop in this case.

[sunil.ap]

Hi Wayne,

We captured tegrastats after flashing with jetson-orin-nano-devkit-super-maxn.conf and running jetson_clocks.

Summary:

• CPU: 4 cores online, all at 1420 MHz
• EMC: 3199 MHz
• GPU: Idle at 611 MHz
• VDD_IN: ~6.3–6.6 W
• NVENC/DLA/etc.: All off

This confirms the EMC clock is hitting expected MAXN_SUPER levels, but CPU cores/freq seem unchanged from default config.

Let us know if this looks correct based on the super config behavior.

Thanks,
Sunil

@WayneWWW
tegrastats_jetson_clk_logs.txt (40.4 KB)

[WayneWWW]

This seems not correct.

jetson-orin-nano-devkit-super-maxn shall give you emc clk to be 3200 but it is still 3199. That is still jetson-orin-nano-devkit-super setting.

Could you share me the host side flash log?

[sunil.ap]

Hi Wayne,

Thanks for the clarification.

We’ve captured the host-side flash log from flashing with jetson-orin-nano-devkit-super-maxn.conf.

Best regards,
Sunil

@WayneWWW

jetson_orin_nano_super_maxn_flash_logs.txt (338.3 KB)

[WayneWWW]

This log is telling a flash process on a custom board. Could you test this on NV devkit too?

Are you sure you didn’t modify anything inside of jetson-orin-nano-devkit-super-maxn.conf? The BPMP dtb file got flashed here is not expected.

[sunil.ap]

Hi Wayne,

You’re right — we’re flashing on a custom carrier board at the moment.

We’ll test the same jetson-orin-nano-devkit-super-maxn.conf on an official NVIDIA devkit and re-check the results.

Also, we’ll cross-verify that the .conf file is unmodified and matches the default one provided in the BSP.

Will share updated results and logs shortly.
jetson-orin-nano-devkit-super-maxn.conf.txt (2.2 KB)

Best regards,
Sunil

@WayneWWW

[WayneWWW]

Hi,

This conf content is not correct. Also, not same as default one at all.

[sunil.ap]

Hi Wayne,

Yes, we did modify the config for our custom board.

The only change we made to the original jetson-orin-nano-devkit-super-maxn.conf was to add our custom DTB in the p3767_super_maxn_overlay() function, like this:

if [ "${board_sku}" = "0000" ] || [ "${board_sku}" = "0002" ]; then
    if [ "${board_FAB}" = "TS1" ] || [ "${board_FAB}" = "EB1" ]; then
        echo "Error: ${board_FAB} can't be used for super maxn";
        exit 1;
    fi
    DTB_FILE=tegra234-mstl-neuron-orin-a01+p3767-0000.dtb;

We didn’t touch the BPFDTB or any other logic — just inserted our custom DTB for SKU 0000.

Let us know if this affects MAXN_SUPER behavior or if there’s a better way to integrate the custom DTB while preserving the super-maxn clock configs.

Thanks again,
Sunil

@WayneWWW

[WayneWWW]

No, I don’t think you didn’t touch anything related to BPFDTB there.

My original file is like this.

29 source "${LDK_DIR}/p3768-0000-p3767-0000-super.conf";
 30 
 31 p3767_super_maxn_overlay()
 32 {
 33         if [ "${board_sku}" = "0000" ] || [ "${board_sku}" = "0002" ]; then
 34                 if [ "${board_FAB}" = "TS1" ] || [ "${board_FAB}" = "EB1" ]; then
 35                         echo "Error: ${board_FAB} can't be used for super maxn";
 36                         exit 1;
 37                 fi
 38 
 39                 BPFDTB_FILE="tegra234-bpmp-3767-0000-3768-super-maxn.dtb";
 40         elif [ "${board_sku}" = "0001" ]; then
 41                 BPFDTB_FILE="tegra234-bpmp-3767-0001-3768-super-maxn.dtb";
 42         else
 43                 echo "Error: module SKU ${board_sku} detected";
 44                 exit 1;

[sunil.ap]

Hi Wayne,

You’re right — our modified config missed setting the BPFDTB_FILE, which your original file includes. That’s likely the reason the super-maxn behavior didn’t take effect.

To clarify, I’m attaching both:

• The original NVIDIA-provided jetson-orin-nano-devkit-super-maxn.conf
• Our modified version for the custom board

jetson-orin-nano-devkit-super-maxn.conf_ORG.txt (2.2 KB)
jetson-orin-nano-devkit-super-maxn.conf.txt (2.2 KB)

Thanks,
Sunil

@WayneWWW

[WayneWWW]

I am not sure why you are asking me such question. Just add your DTB_FILE in it.

[sunil.ap]

Hi Wayne,

We tested with the Jetson Orin NX devkit using the default unmodified jetson-orin-nano-devkit-super-maxn.conf.

Please find the following attached:

1. jetson-orin-nano-devkit-super-maxn.conf
2. Flash log: jetson_orin_nx_super_maxn_flash_logs.txt
3. Screenshot after boot: jetson_orin_nx_dvkit_sprmaxn_pwr_modes.png
4. jtop output after jetson_clocks: jetson_orin_nx_dvkit_sprmaxn_jtop_with_jtson_clk.png
5. tegrastats output after jetson_clocks: jetson_orin_nx_dvkit_sprmaxn_tegrastats_with_jetson_clk.txt

Kindly review and let us know whether everything looks correct.

Thanks,
Sunil

@WayneWWW

jetson_orin_nx_super_maxn_flash_logs.txt (338.4 KB)
jetson_orin_nx_dvkit_sprmaxn_tegrastats_with_jetson_clk.txt (12.6 KB)

jetson-orin-nano-devkit-super-maxn.conf.txt (2.2 KB)

[WayneWWW]

Yes, the result looks good.

[sunil.ap]

Hi Wayne,

Thanks for confirming the results.

Just a quick follow-up:

1. Is there a way to explicitly verify at runtime that MAXN_SUPER mode is active?
2. Does MAXN_SUPER allow a higher power budget (e.g., 40W), and how can we confirm that on the board?

Attaching a screenshot from the NVIDIA documentation for reference.

regards,
Sunil

@WayneWWW

[WayneWWW]

If you scroll down the table in the document you will notice a note there to tell you what is the exact difference between “super” and “maxn_super”. And actually just EMC clock has slight difference. The power budget is still same between super and maxn super. But of course both of them have higher budget than non-super mode.