• State
  • Locked Locked
  • Replies 4 replies
  • Subscribers 98 subscribers
  • Views 350 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TDA4APE-Q1: Question about TDA4APE video Encode/Decode capacity

Mian DENG
Prodigy 200 points

Part Number: TDA4APE-Q1

Hi experts,

We are developing a custom board based on TDA4APE. The system is equipped with multiple 12M cameras. Therefore we have high requirements for image encoding and decoding capabilities.  
According to the chip manual, the maximum image encoding capability is 960 MP/s. However, in actual test, we used the GStreamer tool to encode four streams simultaneously at 25 fps with a 12M camera, and no issues were encountered. The corresponding theoretical encoding capability is 12 × 25 × 4 = 1200 MP/s, which exceeds the specifications in the chip manual.  

f0889364-86ea-45ba-bf85-8d804c6208cd.png
Is the encoding capability described in the manual accurate? Or is there an issue with our test method?  
This affects our software solution selection, so please help confirm this.

Best Regards.

  • 0
    TI__Mastermind 21060 points

    Hello, 

    The Wave5 does have some capability to go past 960MP/s mostly because 960 MP/s is a simplification of 8MP being 3840x2160 as the resolution. However, this heavy of a throughput is not officially supported by our spec due to other potential limitations we were generally concerned about such as DDR b/w. 

    Mian DENG said:
    Is the encoding capability described in the manual accurate? Or is there an issue with our test method?

    Based on your high-throughput use-case I take it you are splitting the load between the two VPU cores. Could you share your application/dataflow? What use-case are you targeting with this high throughput?

    Since our spec doesn't officially support this pixel throughput then if any problems do arise TI is not going to be able to commit to fixing your issue. 

    Regards,
    Sarabesh S.

  • 0 in reply to Sarabesh Srinivasan
    Prodigy 200 points

    Hi,

    You may refer to the picture below for our current data pipeline.

    Note that currently that we have only one camera so that we encode it 5 times to test the encode capacity, which exceeds the spec quite a lot. Meanwhile I understand that this test has a lighter DDR burden than normal case since there is only one input.

    Later we'll have 2*12M and 1*3M cameras connected and we need both H265 and H264 strems. So we'll need to encode 4 *12M  and 2 * 3M. Plus, there would also be image alogrithms applied.

    I wonder how do we calculate the DDR throughput for encode & decode?

    Regards,

    Mian

  • 0 in reply to Mian DENG
    TI__Mastermind 21060 points

    Hi Mian, 

    I'll get back to you on this. Will clarify some details internally.

    Thanks,
    Sarabesh S.

  • 0 in reply to Sarabesh Srinivasan
    TI__Mastermind 21060 points

    Hi Mian, 

    Repoening this ticket. Apologies for the delay. Will get you an update this week.

    Thanks,
    Sarabesh S.