Orin Nano 8GB with missing DRAM chip

Robotics & Edge Computing Jetson Systems Jetson Orin Nano
1

Hi,

I have a Jetson Orin Nano dev kit (SKU 0005) which I damaged one of the 4GB RAM chip. After remove it I am trying to use it as the 4GB model instead of discard it. Would it be that possible?

I have tried flashing with SDK Manager overriding SKU to 0004 but it fails to boot.

I am debugging via UART and I’m encountering a consistent SDRAM initialization failure very early in boot.

Context

  • Module: Jetson Orin Nano 8GB (production unit)

  • Fuse info from UART:

FUSE_SKU_INFO_0 = 0xd5
SKU: Prod

  • Boot device: QSPI (stock devkit)

  • Debug via UART (115200 8N1)

Issue

The module has a hardware defect (one DRAM chip have been removed and pads damage so it can not be restored ).

Because of this, I am trying to see if it is possible to use it as a reduced-memory (4GB-like) configuration.

nano.log (30.7 KB)

However, boot fails very early in MB1:

I> Task: SDRAM init
I> MemoryType: 4 MemBctRevision: 1
I> MSS CAR: Init PLLM
I> MSS CAR: Init PLLHUB
I> Encryption:   MTS: en, TX: en, VPR: en, GSC: en
E> MSS: Failed to initialize SDRAM.
C> Task 0x2f failed (err: 0x48480112)
E> Top caller module: MSS, error module: MSS, reason: 0x12

The board then resets and repeats the same failure.

What I understand so far

  • SDRAM init is performed by MB1 using MEM-BCT

  • The configuration appears tightly coupled to:

    • SKU (fuses)

    • DRAM topology

  • The current MEM-BCT expects full 8GB layout and fails training

My question

Is there any supported or experimental way to make this module usable as a 4GB configuration?

Specifically:

  1. SKU / Fuse override

    • Is it possible to override or ignore FUSE_SKU_INFO_0 during boot?

    • Or force a different SKU selection in the boot chain?

    • Or writing a new value will solve it?

  2. MEM-BCT / SDRAM config

    • Can I generate or modify a MEM-BCT for a reduced DRAM topology (e.g., half population)?

  3. L4T / flashing side

    • Can flash.sh or board config be forced to use a 4GB Orin Nano profile even if the module is fused as 8GB?

    • Would this realistically work, or will MB1 still fail before that matters?

  4. Any workaround

    • Is there any way to bypass or partially initialize memory to allow boot (even with reduced capacity)?

    • Or is this fundamentally impossible due to MB1/firmware constraints?

Goal

If possible, I’d like to reuse the module as a reduced-capacity device instead of discarding it.

Any guidance from NVIDIA or others who have worked with MB1/MEM-BCT would be very appreciated.

Thanks!

3

*** Please note that this reply is generated by LLM automatically ***

The issue you’re encountering is due to the tight coupling between the SKU fuses, DRAM topology, and the MEMBCT (Memory Boot Configuration Table) in the Jetson Orin Nano module. The MEMBCT is generated based on the SKU fuses, which in your case, are set for an 8GB configuration.

To make the module usable as a 4GB configuration, you would need to modify the MEMBCT to reflect the new DRAM topology. However, this is not a straightforward process, and there are several challenges you’ll need to overcome:

  1. MEMBCT generation: The MEMBCT is generated by the NVIDIA SDK Manager based on the SKU fuses. To generate a new MEMBCT for a 4GB configuration, you would need to modify the SDK Manager’s configuration files or use a custom tool to generate the MEMBCT.
  2. FUSESKUINFO0 override: The FUSESKUINFO0 fuse is used to determine the SKU of the module. To override this fuse, you would need to modify the boot chain to ignore or force a different SKU selection. However, this is not a recommended approach, as it may lead to unpredictable behavior.
  3. DRAM topology modification: The DRAM topology is tightly coupled to the MEMBCT. To modify the DRAM topology, you would need to update the MEMBCT to reflect the new configuration. This would require a deep understanding of the MEMBCT format and the DRAM topology.
  4. MB1 firmware constraints: The MB1 firmware is responsible for initializing the SDRAM. If the MB1 firmware is not designed to handle a reduced capacity configuration, it may fail to initialize the SDRAM, leading to a boot failure.

Given these challenges, it’s unlikely that you’ll be able to reuse the module as a 4GB configuration without significant modifications to the MEMBCT, DRAM topology, and MB1 firmware.

However, if you’re willing to invest time and effort into modifying the MEMBCT and DRAM topology, you can try the following:

  1. Generate a new MEMBCT: Use a custom tool or modify the SDK Manager’s configuration files to generate a new MEMBCT for a 4GB configuration.
  2. Update the DRAM topology: Update the DRAM topology to reflect the new configuration.
  3. Modify the MB1 firmware: Modify the MB1 firmware to handle the reduced capacity configuration.

Keep in mind that these modifications are not trivial and may require significant expertise in the area of memory boot configuration and firmware development.

In summary, while it’s theoretically possible to reuse the module as a 4GB configuration, the challenges involved make it a complex and time-consuming process. It’s recommended to discard the damaged module and use a new one instead.

Comparison with other Jetson modules:

The Jetson Orin Nano module is not the only module that uses a MEMBCT to configure the DRAM topology. Other Jetson modules, such as the Jetson AGX Orin and Jetson Xavier NX, also use a MEMBCT. However, the specific challenges and limitations you’re facing with the Jetson Orin Nano module may not be applicable to other modules.

Recommendations:

  1. Discard the damaged module: Given the complexity and challenges involved in modifying the MEMBCT and DRAM topology, it’s recommended to discard the damaged module and use a new one instead.
  2. Use a new module: If you need to use a 4GB configuration, consider using a new Jetson Orin Nano module that is specifically designed for a 4GB configuration.
  3. Contact NVIDIA support: If you’re experiencing issues with your Jetson Orin Nano module, contact NVIDIA support for assistance. They may be able to provide guidance or recommendations for your specific use case.

*** Please give the thumbs up if you get answers you like. Or provide feedback to help us improve the answer. ***

4

Is the Jetson Orin Nano 4GB version hardware wise different or is the same as 8GB just missing the module? If so, could not be possible to override SKU on EEPROM or similar to make it now work as 4GB with the current firmware available for the 4GB version?

5

We’re not able to support this task, sorry for that.

1 Like