DGX Spark – Request AI Enterprise / NIM Entitlement Activation

Accelerated Computing DGX Spark / GB10 User Forum DGX Spark / GB10
, ,
1

Hi,

I recently purchased an NVIDIA DGX Spark system and I am currently unable to run any LLM workloads locally due to missing NVIDIA AI Enterprise / NIM entitlements on my NGC account.

All public toolchains (vLLM, PyTorch, Triton, TensorRT-LLM OSS) fail due to missing support for sm_121a on the Grace–Blackwell architecture. This is expected, since DGX Spark requires the NIM variant-dgx-spark containers and the appropriate Enterprise entitlements.

I would like to request activation of the following for my NGC account:

  • NVIDIA AI Enterprise entitlement

  • Access to NIM DGX-Spark containers (Llama 3.1, Nemotron, Qwen, etc.)

  • NGC API Key with the correct Enterprise permissions

These entitlements are required for running LLM inference locally on the DGX Spark system.

Please let me know what information you need from my side (NGC email, system serial, etc.) so you can proceed with the activation.

Thank you,

Michele Martarello

2

Hi, just following up on this request.
Could someone from the DGX Spark entitlement team
please take a look when possible? Thank you.

1 Like
3

Hi, following up on this request.
Please let me know if you need my NGC email or the DGX Spark serial number.
Thank you.

5

The tools you mention are available without any entitlement and support sm_121 and can run on DGX Spark. Please reference our playbooks on how to do so: Spark Playbooks

6

Thank you for your reply.

The issue is not about running OSS tools.

The problem is that my NGC account is missing DGX-Spark NIM / AI Enterprise entitlements, so I cannot pull or run any of the variant-dgx-spark NIM containers.

In addition:

  • My DGX Spark serial number () shows “on hold or already claimed” on NGC.

  • My NGC Organization (martarello.group) has no AI Enterprise permissions.

  • All NIM container pulls return 403 / permission denied.

  • OSS toolchains fail because the system does not expose the NIM-optimized runtimes for Grace–Blackwell.

Could you please escalate this to the NGC Entitlement / Licensing team?

I only need the correct NIM / AI Enterprise entitlements activated on my account.

My NGC email: martarellosrl@gmail.com

Thank you.

1 Like
8

Any update on this? My DGX Spark still cannot pull the variant-dgx-spark NIM containers.

9

@martarello.group did you manage to solve this issue?
@aniculescu any new info on this?

I have to same problem where I’m unable to pull NIM containers and I’m also unable to add my dgx-spark serial number device on my NGC account and get the proper entitlements to be able to access NIM assets. Opened another ticket [here](Unable to activate subscription for my new dgx spark device on the NGC platform)

10

Update:

Final Staff Support answer:

DGX Spark does not require NGC hardware activation or serial number registration; If you’re blocked generating an NGC API key, this is a known UI bug, please use the direct link https://org.ngc.nvidia.com/setup/api-keys as a workaround.

We only need to generate an NGC API key with proper permissions. No need to register the dgx-spark device.

More info in the other ticket where I encountered a similar, overlapping, problem to this one when trying to pull official dgx-spark NIM contianers.

Best,

M

11

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.