thermal zones

Autonomous Machines Jetson & Embedded Systems Jetson TK1
#1

Hello,

In university i have to hold a presentation which led me to the different thermal zones of the jetson k1 board.

The thermal zones with names “CPU” and “GPU” are self-explaining.
Also i already know what “Tboard” and “Tdiode” stand for.
Now my questions:
Does the one with “mem” refer to the RAM of to the storage? (I would guess the RAM…)
And finally what does “PLL” stand for ? Which part of the Board is meant by that?
I googled, but couldn’t find an answer this question…

I would be very grateful for answers! :)

Daniel

#2

Don’t know about the thermal zones, or physical placement, but PLL is “phase locked loop”. These are associated with clocks and clock speed division/multiplication.

#3

The TK1 chip have eight on-die thermal sensors and one on-die thermal diode
Seven of the sensors are grouped into three thermal zones,one strongly influenced by the CPU, one strongly influenced by the GPU, and one strongly influenced by the DRAM memory controller and interface, the one remaining sensor plays a support role to the other seven, a hardware controller, called SOC_THERM,can directly read the sensors. The temperature are referred to as CPU_therm,GPU_therm,MEM_therm,PLL_therm.
The TK1 also contain a single, one-die thermal diode that is accessed via an external i2c temperature monitor(Ti TMP451) on Jetson design. The thermal diode temperature is referred to as T_diode.
The same temperature monitor IC(TMP451) used for accessing the thermal diode also contain an internal temperature sensor, whose sensed temperature is referred to as T_board, this temperature that is correlated to the internal PCB temperature and not entirely dominated by TK1 temperature.
MEM refer to the memory controller and interface inside TK1.
PLL refer to the clock domain inside TK1

#4

@edli1983:

which temperature is measured by the thermal diode? The question came up, because we measure a temperature difference of approx. 10°C between the internal sensors (Tegra) and thermal diode at stress tests. The sensors and the thermal diode are all on-die, so that they should measure all nearly the same temperature or not?

sensors output:

CPU-therm-virtual-0
Adapter: Virtual device
temp1: +64.0 C (crit = +105.0 C)

GPU-therm-virtual-0
Adapter: Virtual device
temp1: +63.5 C (crit = +105.0 C)

MEM-therm-virtual-0
Adapter: Virtual device
temp1: +63.5 C (crit = +105.0 C)

PLL-therm-virtual-0
Adapter: Virtual device
temp1: +63.5 C

tmp451-i2c-4-4c
Adapter: Tegra I2C adapter
temp1: +41.6 C (low = +0.0 C, high = +120.0 C)
(crit = +122.0 C, hyst = +110.0 C)
temp2: +56.0 C (low = +0.0 C, high = +95.0 C)
(crit = +97.0 C, hyst = +85.0 C)

#5

cat /sys/class/thermal/thermal_zone*/type

CPU-therm: Tsensor inside TK1, close to CPU part.
GPU-therm: Tsensor inside TK1, close to GPU part.
MEM-therm: Tsensor inside TK1, close to MEM part.
PLL-therm: Tsensor inside TK1, close to PLL part.
Tboard_tegra: Tsensor inside external temperatur sensor, take the local temperature as PCB temperature.
Tdiode_tegra: External temperature sensor read the temperature through Tdoide inside TK1, Tdiode is close to CPU and GPU part. it’s more accurate than TK1 internal Tsensor readout.