Is there any data available about the effect of temperature on the lifetime of the nano production module?
We see in our application that the current heatsink is not sufficient enough to keep the processor under 95°C without throttling.
To overcome this, we are designing a new heatsink but need some more information that cannot be found in the thermal design guide.
- what is the expected lifetime at 95°C?
- how does the lifetime change when the temperature is reduced?
- what is the thermal resistance between the chip and the outside of the package?
The operating lifetime is 5 years which needs components in module working in temperature range as listed in “Jetson nano thermal performance” table in thermal design guide. The thermal resistance is a calculated value as you can see in thermal DG.
What I can’t find in the thermal DG is the thermal resistance between the die and the case. It is a crucial part of the calculation. The resistance of the thermal interface and the heatsink are known from their respective datasheets but the part of the NVIDIA SoC is missing.
Are there also calculations available about the lifetime of the product when the temperatures are lower than what is defined in the table?
What the resistance you said is covered by θjp (Tj to TTP).
No calculation for lifetime of that as it will be better if temperature is lower than that in the table.
The list of all the component temperatures is a nice start but are there measurements or thermal images available?
All those components are unreachable under the heatsink or between the module and the carrier board so we have no clue how they relate to the CPU and GPU temperature
About the thermal resistance I cannot find a number that defines θjp. Can you please give me that value or point me to the right page in the manual?
The temperature spec of components in table is just for component itself, not related to CPU/GPU. Nano module is sold without TTP, these components are all reachable, customer should design TTP based on the spec.
As said prior, θjp is a tested and calculated value ((T2-T1)/P) because the heat dissipation might be different even if the power consumption is same, it needs customer to test with their target use case to get P and T.
I’m sorry if I keep coming back to the same point but I don’t really see how that is possible.
The nano module is designed in a way that only the CPU/GPU can be thermally connected to the heatsink so we have control over how much heat can be dissipated from that one part but not all the other components in the list.
On the other hand, we cannot measure the temperature of all these components because we cannot run the board on full load without a heatsink to make a thermal image.
Can NVIDIA please support us in this by stating which of the components are at risk of running at their maximum temperature with the stock heatsink that is provided with the development board?
About the thermal resistance I think we have a misunderstanding. To calculate/simulate the temperature of the die at a certain load with our new thermal solution I am only missing one value: the thermal resistance between the die and the outside of the package. That is not related to the temperature or our solution but only based on the design of the chip.
To give you an example, this is a list of of thermal resistances for intel devices:
In this example I would need qJC (junction to case)
Yes as said in table 3-1, only X1 chip must be contacted, others except memory are miscellaneous. And also as said in guide: Customer must determine if system thermal solution needs to contact all / select components on the PCB to make sure that they are maintained within the maximum temperature specifications listed in Table 3-1.
Whether the components in table 3-1 need to be contacted depends on use case, in general it is unnecessary, but as I said, it is a real tested result.
Regarding the θjc you said, we don’t have such data, while in simple you can take Tj as Tc since they are so close. Jetson nano has no θjp as it is sold without TTP. And as you can see in other Jetson platform thermal DG, such as Xavier, even the θjp is a calculated value by testing based on customer’s use case.