IMX28 vs IMX6 CPU tests on LogicMachine
Up to now all LM2, LM3, LM4 series came with i.MX28 ARM9 architecture based CPU. A new CPU generation based on i.MX6 ARM Cortex A7 is optionally available starting from LM5 series devices (LM5 has i.MX6 CPU by default, LM5 Lite and Reactor can have it optionally installed).
| Features | New hardware | Actual hardware |
|---|---|---|
| CPU | ARM Cortex A7 | ARM9 Architecture |
| Maximum clock speed | 528 MHz | 454 MHz |
| Embedded SRAM | 32 KB / 32 KB L1, 128 KB L2 | 16 KB / 32 KB |
| DDR Memory | 128 MB | 128 MB |
The following results are based on tests done in Schneider Electric CZ laboratory.
For testing purposes test bench was built with 21 pcs of LM controllers.
For simulation of real scenario there is running KNX bus traffic generator in each LM. KNX bus traffic simulates real-time work of KNX devices and caused by people in building. Structure of objects is as follows:
- Each LM contains 300 objects (128x 1-bit objects, 32x 4-bit objects, 88x 1-byte objects, 52x 2-byte objects ) which are randomly changed every 5 minutes to simulate real traffic on KNX network.
- From these 300 objects are visualized 209 objects (100x 1-bit objects, 77x 1-byte objects, 24x 2-byte objects 4x 1-byte gauges and 4x 2-byte gauges) in each plan.
- Users viewing the visualization have significant affect to CPU load of LM controller. Overloading of CPU cause unavailability of controller as well as its rebooting. That is the reason why is important to specify how many users can be connected to visualization in the same time.
Test example: CPU load (5 users connected, traffic)
Actual CPU, average load = 0.77
New CPU, average load = 0.15
Conclusion
Thanks to higher clock rate and new architecture ARM Cortex A7 we reached approximately 3x more performance in comparison with actual ARM9 used in LM2, LM3 and LM4.
New CPU also can handle more users simultaneously accessed the visualization. We observe faster response in Configurator too which brings benefits to integrators during controller’s commissioning.
| Test | Actual CPU (load) | New CPU (load) | Performance increase | |
|---|---|---|---|---|
| Multiuser access to visualization (traffic) | 1 user | 0,59 | 0,24 | 2,45x |
| 3 users | 0,74 | 0,25 | 2,96x | |
| 5 users | 0,82 | 0,37 | 2,22x | |
| 10 users | 1,10 | 0,33 | 3,33x | |
| 15 users | N/A | 0,41 | ||
| Multiuser access from smart devices (traffic) | 5 users | 0,77 | 0,15 | 5,13x |