NXP iMX8 series application processors are ARM processors based on Cortex-A72/A53/A35 and Coretex-M4/M7 architectures released by NXP, including the following sub-series. In this paper, representative iMX8, iMX8X and iMX8mini are selected for simple power consumption comparison test, which provides reference for you to choose specific IMX8 product series for the project from the perspective of power consumption.
Please note that since the software and hardware of iMX8 series products have not reached the final stable state, the test data in this paper is limited to the current state for reference, and there may be differences after the subsequent software and hardware optimization.
The iMX8 series ARM platform demonstrated in this paper comes from Toradex ARM embedded platform. In order to minimize the interference of peripherals on power consumption, all simplified carrier boards are adopted (except iMX8MM platform, which has not been simplified for the time being). However, this paper contains the power consumption data of ARM computer module and carrier board, so it will be slightly larger than the actual module.
. /iMX8QM platform? -? Apalis iMX8QM？ 4GB WB IT ARM Core Edition with Ioxra? The carrier board based on NXP iMX8QuadMax processor supports two Cortex-A72+four Cortex-A53 CPU and two GC 7000xVX GPU (each core supports up to 128GFLOPS).
. /iMX8QP platform? -? Apalis iMX8QP 2GB WB? Arm? Core board cooperation? Ixora？ The carrier board based on NXP iMX8QuadPlus processor supports1x cortex-A72+4x cortex-A53 CPU, 2x GC7000Lite XSVX GPU (up to 80GFLOPS per core).
. /iMX8X platform? -? Colibri iMX8QXP 2GB WB? Arm? Core board with iris? The carrier board based on NXP iMX8QuandXPlus processor supports 4x Cortex-A35, 1x GC7000Lite GPU (up to 64GFLOPS per core).
. /iMX8MM platform? -? Verdin iMX8MM 4-way 2gbwit? Arm? The core board is matched with the onboard board developed by Verdin, based on NXP iMX8M mini quad-core processor, and supports 4x Cortex-A53,1x GCnanoUltraGPU (up to 6.4GFLOPS per core).
2).？ prepare
a)。 ? Hardware connection
Different test platforms are powered by digital DC regulated power supply (REK–PS-303 DM+), and peripherals are connected to debugging serial ports. IMX8QM, iMX8QP and iMX8MM platforms are connected with HDMI display, and iMX8X platform is connected with VGA display.
b)。 ? Test software:
. /? Install toradex v3.0b4yco to the console Linux BSP image on all platform modules. Based on Linux kernel version 4. 14. 170.
. / CPU？ The load is realized by BSP's own stress tools.
The load. /GPU is implemented by the glmark2 tool compiled in the Ycoto environment. The compiled glmark2 ipk software package is copied to the platform Linux and installed by the "opkg install" command.
c)。 ? Test flow
. /? The platform is powered on, and there is no load after entering the Linux system. The power consumption data is obtained by testing the current value and voltage value of the DC power supply in the idle state of the system.
. /? Execute the following command, add the CPU load to full load, observe and record the current value and voltage value of DC power supply, and get the power consumption data.
-
$ pressure -c
-
. /? On the basis of full CPU load, eight processes are started by the following Linux shell script to realize GPU pressure loading, and then the power consumption data is obtained by observing and recording the current value and voltage value of DC power supply.
-
#! /bin/sh
for((I = 0； I<8; i++))；
do
GL mark 2-es2-way land-run-forever-full screen & gt； /dev/null 2 & lt； & amp 1。
finished
-
3).？ test result
a)。 ? Because power consumption and temperature are positively related, there will be some differences in power consumption at different temperatures, and the characteristics of GPU architecture also make it unstable at runtime. Combined with the maturity of software and hardware mentioned at the beginning of this paper, the power consumption test data in different scenarios and different States will inevitably be different, so it is suggested that the data in this paper should only be used as a horizontal comparison reference, not as a direct basis for actual heat dissipation design.
b)。 ? Test data for each platform
. /iMX8QM platform
Current (a) voltage (v) power (w)
The idle state is 0.45 12. 15.445.
CPU？ 0.74–0.9512.18.954–11.495 when fully loaded.
CPU+GPU？ Full load status1.2–1.412.114.52–16.94.
. /iMX8QP station
Current (a) voltage (v) power (w)
Idle state 0.44 12. 15.324
CPU？ 0.65–0.812.17.865–9.68 when fully loaded.
CPU+GPU？ The fully loaded state is 0.92–1.112.1.132–13.5438+0.
. /iMX8X platform
Current (a) voltage (v) power (w)
Idle state 0.24 12. 12.904
CPU？ Fully loaded state 0.32 12. 13.872
CPU+GPU？ The fully loaded state is 0.43–0.5712.15.203–6.897.
. /iMX8MM platform
Current (a) voltage (v) power (w)
Idle state 0.33 123.96
CPU？ Full load state 0.42 125.04
CPU+GPU？ Full load state 0.47 125.64
c)。 ? Horizontal comparison of test platform power consumption, here only compare the maximum power consumption data tested.
IMX8QM platform iMX8QP platform iMX8X platform iMX8MM platform
Idle state power (w) 5.4455.2442.4043.96
CPU？ Power under full load (w)11.4959.2825.04
CPU+GPU？ Power at full load (w)16.3338+03.338+06.30080.000000000606
4).？ abstract
This paper tests the power consumption reference data of NXP iMX8 product line. It can be seen that iMX8QM/QP has strong performance, but the corresponding power consumption is relatively large. The power consumption and performance of iMX8X series have reached a good balance, while the CPU performance of iMX8MM series is relatively good and the GPU is relatively weak. Finally, once again, because the power consumption data is influenced by many factors such as software and hardware maturity, temperature, peripherals, test equipment and conditions, the test data in this paper is only for horizontal comparison.