Jetson Xavier NX clock not stable

Hi, I have noticed that in my Xavier NX, the core clock is not stable (it should be at 1.44Ghz but coming down to 700Mhz). I have configured the Xavuer NX to mode2 (6 core, 15W) and jetson clock is also running. But, the core clock is still varying a lot especially when CPU load increases. Whats the cause for this …?

hello JSP_1,

it’s DFVS to change the loading dynamically, however, could you please enable tegrastats utility and gather the report statistics for reference,
thanks

Hi Jerry Chang, I have attached the Jtop result, In tegra stats I am not seeing any change in clock

RAM 5361/7764MB (lfb 91x4MB) SWAP 332/3882MB (cached 17MB) CPU [93%@1420,82%@1420,84%@1420,83%@1420,76%@1420,77%@1420] EMC_FREQ 0% GR3D_FREQ 5% AO@54.5C GPU@57.5C PMIC@100C AUX@57C CPU@61.5C thermal@58.45C VDD_IN 10338/10542 VDD_CPU_GPU_CV 5079/5234 VDD_SOC 1720/1752
RAM 5363/7764MB (lfb 91x4MB) SWAP 332/3882MB (cached 17MB) CPU [89%@1420,85%@1420,82%@1420,84%@1420,79%@1420,83%@1420] EMC_FREQ 0% GR3D_FREQ 2% AO@54.5C GPU@57.5C PMIC@100C AUX@57C CPU@61.5C thermal@58.15C VDD_IN 10420/10542 VDD_CPU_GPU_CV 5120/5234 VDD_SOC 1720/1752
RAM 5363/7764MB (lfb 91x4MB) SWAP 332/3882MB (cached 17MB) CPU [80%@1420,80%@1420,82%@1420,84%@1420,83%@1420,84%@1420] EMC_FREQ 0% GR3D_FREQ 1% AO@54.5C GPU@57C PMIC@100C AUX@57C CPU@61C thermal@58.5C VDD_IN 10338/10541 VDD_CPU_GPU_CV 5038/5233 VDD_SOC 1720/1752
RAM 5364/7764MB (lfb 91x4MB) SWAP 332/3882MB (cached 17MB) CPU [91%@1420,83%@1420,88%@1420,86%@1420,75%@1420,76%@1420] EMC_FREQ 0% GR3D_FREQ 2% AO@54.5C GPU@57.5C PMIC@100C AUX@57C CPU@61.5C thermal@58.15C VDD_IN 10379/10540 VDD_CPU_GPU_CV 5120/5232 VDD_SOC 1761/1752
RAM 5364/7764MB (lfb 91x4MB) SWAP 332/3882MB (cached 17MB) CPU [88%@1420,85%@1420,88%@1420,86%@1420,79%@1420,76%@1420] EMC_FREQ 0% GR3D_FREQ 2% AO@54.5C GPU@57C PMIC@100C AUX@57C CPU@61.5C thermal@58.35C VDD_IN 10543/10540 VDD_CPU_GPU_CV 5160/5232 VDD_SOC 1761/1752
RAM 5366/7764MB (lfb 91x4MB) SWAP 332/3882MB (cached 17MB) CPU [89%@1420,80%@1420,80%@1420,80%@1420,80%@1420,78%@1420] EMC_FREQ 0% GR3D_FREQ 1% AO@54.5C GPU@57C PMIC@100C AUX@57C CPU@61.5C thermal@58.3C VDD_IN 10297/10539 VDD_CPU_GPU_CV 5079/5231 VDD_SOC 1720/1751
qRAM 5368/7764MB (lfb 91x4MB) SWAP 332/3882MB (cached 17MB) CPU [84%@1420,88%@1420,85%@1420,91%@1420,75%@1420,78%@1420] EMC_FREQ 0% GR3D_FREQ 4% AO@54.5C GPU@57C PMIC@100C AUX@56.5C CPU@61.5C thermal@58.3C VDD_IN 10297/10537 VDD_CPU_GPU_CV 5038/5230 VDD_SOC 1720/1751

hello JSP_1,

that should be an issue with JTop,
actually, JTop and tegrastats they’re parsing the same node for the system info.

Hi Jerry Chang,
So you mean tegra stat is the real indicator and no need to worry about jtop right …?

hello JSP_1,

you should not worry about that,
furthermore, since the default clock value parsing interval is in 1000 milliseconds, could you please try to reduce the values to dump the data to confirm the status.
i.e. $ tegrastats --interval <int>

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.