AM62P: Power Consumption for Encoder & Display

Hi Kevin,

I wasn't able to identify significant power savings in the PET that would lower the power to <1.5W at 90C with fast/strong silicon process (worst case leakage). Anshu Madwesh from our software team is also looking into this. 

Thanks,

Brenda 

Hi Brenda,

We have some good news today that customer agrees to give some concessions on the requirements:

• Customer can accept not using video encoder during runtime, and only used it in a short time when needed.
• After knowing our Junction-to-free air is only 13.8 degree/w, customer thinks the maximum Tj is only about 75 degree (decrease from previous 90 degree).
• Customer aggress to decrease the input CSI resolution from 1080P 30 fps to 720*1280 30 fps
• Customer aggress to decrease MIPI-DSI resolution from 1080P 30 fps to 640*480 30 fps
• Customer finds a way to increase their battery capacity, and could accept AM62P power consumption about 1.7W (increase from previous 1.5W)

That being said, we need evaluate if AM62P could meet 1.7W at 75 degree Tj if only the below applications running:

• CSI receiving 720*1280 30 fps
• MIPI-DSI displaying 640*480 30 fps
• Encoder & DPI are enabled but not running (DPI enabled because there will be short time usage to connect external HDMI, but this will not be keep using, so not being considered for power consumption, but need to make sure it is enabled)

I have updated the attached excel sheet, reducing encoder utilization from 60% to 0%, keeping only 1 A53 core running at 600MHz, and reserving only 30% for DSI & CSI. Now it seems we could meet this target about 1.6W.

Please also help double check if the current configuration is reasonable to meet this application requirement, thanks.

 4062.SPRUJD9_AM62P_PET_1_1(1).xlsm

Thanks,

Kevin

Hi Kevin,

The PET does not take into account all the architecture power saving mechanism (i.e LPSCs) that are supported on Linux. Anshu Madwesh  from our software team is looking into this. 

Thanks,

Brenda

Hi Brenda,

There are 3 points here:

1: The most important point is to review our latest PET which already achieves 1.6W at 75 degree Tj to see if it is a real reasonable value that is achievable on a real board with the applied configurations.

2: Please help review the applied configurations (single A53 with 600MHz, 1600M bps DDR 60% utilization, 30% utilization of CSI & DSI) are reasonable to meet the below application usage:

• CSI receiving 720*1280 30 fps
• MIPI-DSI displaying 640*480 30 fps
• Encoder & DPI are enabled but not running

3: If the above two points are aligned, customer will kick off and start building their own board, then we could invite Anshu to provide some further optimizations (e.g. LPSCs).

Could you help on the first two points please?

1805.SPRUJD9_AM62P_PET_1_1(1).xlsm

Thanks,

Kevin

Hi Kevin,

Kevin Peng said:

1: The most important point is to review our latest PET which already achieves 1.6W at 75 degree Tj to see if it is a real reasonable value that is achievable on a real board with the applied configurations.

This is exactly where Anshu is helping us; Understanding if the IP/resources configured or disabled on PET can be supported by Linux and implemented on a real customer application. From the SoC architecture perspective, it can be implemented but we are in the process of understand that is supported on software/Linux. The limiting factor is not the power architecture but the software.  

Thanks,

Brenda