Power supply for Thor module

Hi, I am currently looking into making a carrier board for the module, and I am looking at the power requirements. Can the module deliver the full 330 W when running from 12 V, or is that only possible at 7 V, as described here?


Because when you look at the Imax specification, it states 22 A at 20 V.

So, can it not pull more than 22 A at any voltage? In that case, it would not be able to reach its maximum 330 W at 12 V, or will it simply draw more current?

I understand that the 330 W figure is for pulse loads, but I want to design the carrier board so there are no power limitations. Therefore, I would like to know: for the Thor module to operate without being power-limited, should SYS_VIN_HV be supplied with 20 V?

hope my question makes sense.

This might
https://docs.nvidia.com/jetson/archives/r38.4/DeveloperGuide/SD/PlatformPowerAndPerformance/JetsonThor.html#system-power-measurement

And further down same webpage to system-power-capping

jetson-power-gui might be helpful. If you don’t have nvidia icon in top right then you can run it with

sudo python /usr/share/jetsonpowergui/__main__.py

Thank you, but I don’t really think this answers my question about which voltage the module should run at to achieve the maximum power draw of 330 W for 1 ms, or whether supplying only 7–9 V would limit the maximum power the T5000 module can run at.

I tried estimating the power consumption using the Power Estimator, and I got this:

image

The exclamation mark is just indicating that the module can reach up to 130 W or be capped at 130 W under normal operation. So the question is whether the power supply should be designed for 130 W or for the 330 W transient peak.

Would it make sense to design the regulator itself for 130 W continuous output, and then size the capacitors to handle the 330 W peak for 1 ms? and then have the output at 7–20 V.

I also read that NVIDIA recommends the power supply should be capable of handling the 330 W peak, but it seemed like this was mainly to avoid triggering the PSU current limit or protection circuitry during short transients.

What I was thinking was to design around 20 V at 22 A, like in the picture before, so there would be no power limitations. But if that’s not necessary, it would be nice to know.

The 22A @ 20V entry is an absolute maximum input current rating based on max stress workloads. The corresponding 440W should be treated as the SYS_VIN_HV rail design limit, not as allowable draw whatever current is needed at lower voltage.
EDP is specified in terms of power on SYS_VIN_HV, not a fixed current independent of voltage. For a lower nominal voltage such as 12V, the peak currents to support the same worst‑case EDP envelope scale approximately with I = P/V, but whether your system ever actually reaches that envelope is highly workload dependent.
Carrier design recommendations:

  • Size the SYS_VIN_HV regulator for ~330W capability on a 100us–1ms moving‑average basis.

  • Set the regulator OCP high enough to tolerate short transients up to the ~440W level (6us) without tripping, then validate this on your actual HW and workload.

  • Add adequate bulk decoupling if workload requires those EDP/peak events.

If the input supply cannot support the transient current/power, the expected behavior is throttling, voltage droop, OCP trip, or reset depending on the overall power path.

Choose a nominal SYS_VIN_HV and design the regulator and decoupling network to keep load step undershoot/overshoot within the recommended operating range with margin under worst-case EDP events. The key requirement is that the voltage at the module pins remains within the datasheet’s min/max limits across all load transients.

This would be the preferable configuration over a lower voltage.

For more information about this topic, please refer to section 1.3.2 Electrical Design Point in Jetson Thor Series Thermal Design Guide.