What is maximum tflite model size that can be deployed on MCXN947?

[nnxxpp]

I am newbie with NXP board. I have a board MCXN947 and I read the specification of this board. It mentioned 3 Flash: 2MB with ECC, 64 Mbit fro QSPI flash and 512 Mbit Octal flash. So when deploy tflite model on this board, which flash memory I will flash into? I am concern about this problem because may be I can not deploy a "large" model. I am using MCUXpresso and examples from SDK + Code Hub.

I surprised about small flash memory of this board compared to MIMXRT1xxx boards.

[Harry_Zhang]

Hi @nnxxpp 

When deploy tflite model on this board, you will flash into 2MB Flash.

BR

Harry

[nnxxpp]

Hi @Harry_Zhang 

2 MB is very small. If my model size is larger than 2 MB. I don't want to reduce  model size. What I need to do ? Can I use other flashes or external storage to run model? Thanks.

[Harry_Zhang]

Hi @nnxxpp 

You can use external storage to run model. MCXN947 support XIP.

BR

Harry

[nnxxpp]

@Harry_Zhang 

Sorry, I have one more question.

Whether I can use 64 Mbit QSPI flash for running example from SDK. If I can, I will use larger model on 64 Mbit QSPI flash.

Could you sharing with me guide or tutorial to runing SDK examples with 64 Mbit QSPI flash in the case we can?

[Harry_Zhang]

Hi @nnxxpp 

YES, you can use 64 Mbit QSPI flash for running example.

You can refer to the hello_world_qspi_xip.

BR

Harry

[nnxxpp]

@Harry_Zhang 

Thank you.

It is not easy to custom eIQ running with QSPI. I will investigate it

[nnxxpp]

Thank you so much.