System info
• Module: Jetson Orin Nano 8GB Developer Kit (P3767-0005, confirmed via /etc/nv_boot_control.conf)
• JetPack: 7.2 / L4T 39.2.0 (installed via Jetson ISO USB path)
• Storage: microSD card
• No Ubuntu x86 host PC available
Problem
GPU frequency is hard-capped at 624,750,000 Hz even in MAXN_SUPER mode. 1007 MHz does not appear in available_frequencies.
Symptoms
cat /sys/devices/platform/bus@0/17000000.gpu/devfreq/17000000.gpu/available_frequencies
→ 306000000 408000000 510000000 612000000 624750000
sudo nvpmodel -q
→ NV Power Mode: MAXN_SUPER, 2
sudo cat /sys/kernel/debug/bpmp/debug/clk/nafll_gpc0/max_rate
→ 624750000
BPMP is reporting 624.75 MHz as the hardware max — this appears to be a firmware-level limit.
What we’ve tried
- nvpmodel symlink — set /etc/nvpmodel.conf → nvpmodel_p3767_0003_super.conf (no 0005 conf exists in /etc/nvpmodel/)
- Disabled is_nvpm_relinking_needed() in nvpower.sh to prevent symlink resets on boot
- Ran nv_bootloader_payload_updater — failed with “Get SPI boot device size failed / Publish block device failed”
- Ran nv_bootloader_capsule_updater.sh -q /opt/ota_package/t23x/TEGRA_BL_3767_super.Cap — succeeded, staged capsule, rebooted. Capsule update status changed from 1→0 (consumed) but GPU frequency unchanged.
- Switched devfreq governor to userspace, wrote 1007000000 to set_freq — kernel accepted the write but cur_freq stayed at 624750000. BPMP hard limit confirmed.
- Verified nvpmodel_p3767_0003_super.conf has correct JetPack 7.2 sysfs paths (old ga10b path is commented out, platform/bus@0 path is active), GPU MAX_FREQ = -1 (uncapped) for MAXN_SUPER mode.
- Checked journalctl — nvpmodel service runs cleanly on boot with no errors. The “golden image” warning only appears on manual runtime attempts.
Key observations
• /etc/nvpmodel/ has no nvpmodel_p3767_0005 conf — 0005 module falls through to 0003 conf
• Device tree compatible string: nvidia,p3767-0005 (correct)
• nvbootctrl dump-slots-info: Current version 39.2.0, Capsule update status: 0 (consumed)
• No nvidia-l4t-jetson-orin-nano-qspi-updater package available under L4T 39.2
Question
How do we get the correct Super Mode BPMP firmware flashed on a P3767-0005 module running L4T 39.2 without an Ubuntu x86 host PC? Is there a hostless path, or is the Jetson ISO reflash (catching the QSPI capsule prompt) the only option?
Is the absence of nvpmodel_p3767_0005.conf in L4T 39.2 expected, and should we be using a different conf for the 0005 module?
Are you talking about using a ISO image to install the board but it could not give you the super mode?
Yes. I bought a new dev kit. Checked that the firmware was 36.0 then used USB image to flash to 39.2. I thought I ran the qspi update. Then installed jetpack/l4t on microsd. And noticed that there was no option for super. Did some digging. And I think the firmware isn’t flashed correctly? Played with nvpmodel, nvpower, looked at available freq, etc. Trying to avoid total reinstall.
Max GPU freq listed at 624mhz. So that’s weird and I only get the 15w/7w power options. Like I said I linked to the _super.conf power profile but still max freq is set to 624mhz
Hi,
This would be fixed in jp7.2.1.
It is a known issue that super mode is missing in ISO image.
when is jp 7.2.1 going to be released? And will I need to reflash? or simply upgrade?
I have this same question will this need reflash or upgrade? I wasted a lot of hours in trying to workaround until I found known issue. Only GPU does not allow in super mode other settings I was able to change to get 25W mode and MAXN just GPU capped at half speed.
For me, the same problem. I tried all the things listed above, but no solution found yet. Any progress from Nvidia. I dit buy the nano for max speed…
Currently, you could only get super mode by flashing your device with sdkmanager.
Which means you would need another Ubuntu host to flash your board.
Thanks for the earlier reply — understood that Super Mode currently requires flashing the device with sdkmanager from a separate x86 Ubuntu host. I’m arranging that host now, but before I wipe a working setup I’d like to confirm the target version.
Current system
- Module: Jetson Orin Nano 8GB Developer Kit (P3767-0005, confirmed via
/etc/nv_boot_control.conf)
- JetPack 7.2 / L4T 39.2.0, installed via the Jetson ISO USB path
- Firmware:
nvbootctrl dump-slots-info reports Current version 39.2.0, capsule update status 0
nvpmodel -q reports MAXN_SUPER (mode 2) as active- GPU clock is hard-capped at 624,750,000 Hz;
jetson_clocks does not change it, and 1007/1020 MHz does not appear in available_frequencies
/sys/kernel/debug/bpmp/debug/clk/nafll_gpc0/max_rate also reports 624750000
My questions
- Which JetPack version should I flash to actually get the Super Mode GPU clock on this module — JetPack 6.2.x, or a 7.x release? My current 7.2/L4T 39.2 install reports MAXN_SUPER but stays at 624 MHz, so I want to avoid flashing back into the same state.
- Is the 624 MHz cap on 7.2/L4T 39.2 a known issue that a future JetPack 7.x release will fix? If a fix is expected soon, I would rather wait than reflash.
- Is there anything specific to the P3767-0005 module I should select in sdkmanager? I noticed there is no
nvpmodel_p3767_0005*.conf in /etc/nvpmodel/ on my current install — it appears to fall through to the 0003 configuration.
Context: this is a production inference workload (TensorRT, multiple camera streams). The GPU clock directly determines whether one device can handle two streams, so the difference between 624 MHz and 1020 MHz matters a lot for my deployment. Reflashing means rebuilding the whole environment, so I want to get the version choice right the first time.
Thanks in advance.
Using manual flash or sdkmanager shall give you the super mode on both jp6 and jp7.
This issue is just ISO image bug that leads super mode not coming.