EdgeXpert MS-C931 — SOCFW 2.152.15 and PD FW 5.22 needed to fix driver-EC communication

K

ka15e902e8

New member
Joined
May 3, 2026
Messages
2
System: MSI EdgeXpert MS-C931
Current BSP: GA2 OTA2 (2026-03-04)
OS: Ubuntu 24.04, Kernel 6.17.0-1014-nvidia
Driver: nvidia-driver-580-open 580.142

Issue

The NVIDIA GPU driver cannot establish full communication with the EC on the EdgeXpert. Specifically, the driver-EC
mailbox (used to read power budget and GPU configuration) is not initializing correctly. Symptoms:

- nvidia-smi reports GPU power limit as N/A and GPU model as Unknown
- GPU defaults to a conservative 2418 MHz clock (max is 3003 MHz) due to unknown power budget
- SW Power Capping events accumulate, periodically locking GPU power to ~13W
- nvidia-spark-ota-check reports EC/SOCFW/USBPD as <missing> — the tool cannot read firmware versions via its
detection protocol on MSI hardware

GPU compute is functional (~70W during inference) but performance is degraded due to the missing power budget
information.

---
Diagnostics (sudo dmidecode -t 45)

Firmware Component Name: FLASH
Firmware Version: SBP:R:2.148.24

Firmware Component Name: UEFI
Firmware Version: MSI_UEFI_40_1.6.0

Firmware Component Name: EC Firmware
Firmware Version: 4.1.60

Firmware Component Name: PD Firmware
Firmware Version: PD0 FW1: 5.0, FW2: 4.10
PD1 FW1: 5.0, FW2: 4.10

---
Root Cause

Comparing against the NVIDIA DGX Spark Founders Edition (OTA2604), which has full driver-EC communication working:

┌───────────┬─────────────────────┬────────────────────────────┐
│ Component │ EdgeXpert (current) │ Founders Edition (working) │
├───────────┼─────────────────────┼────────────────────────────┤
│ SOCFW │ 2.148.24 │ 2.152.15 │
├───────────┼─────────────────────┼────────────────────────────┤
│ PD FW2 │ 4.10 │ 5.22 │
├───────────┼─────────────────────┼────────────────────────────┤
│ EC │ 4.1.60 (MSI) │ 3.3.2 (NVIDIA) │
└───────────┴─────────────────────┴────────────────────────────┘

SOCFW 2.152.15 is the version that fixes the power delivery and mailbox communication on the Founders Edition
(confirmed on NVIDIA developer forums:
https://forums.developer.nvidia.com/t/dgx-spark-performance-degradation-gpu-power-draw-issue/361294). The EdgeXpert is
currently one SOCFW version behind.

---
Request

Please publish an updated BSP for the MS-C931 that includes:
1. SOCFW 2.152.15 (or MSI-equivalent)
2. PD firmware FW2: 5.22 (or MSI-equivalent)

The current GA2 OTA2 BSP (paired with EC9310N1MSI.40170.bin) is the only available release. A follow-up BSP pairing
newer SOCFW + PD firmware with an updated EC binary would resolve this.

Happy to provide additional diagnostic output if helpful.
 
Thank you for your detailed analysis and for sharing the diagnostic data.
MSI has already started the development and validation of the next BSP release aligned with OTA2604, which includes updated firmware components (SOCFW, PD, and EC).
The validation process is currently in progress, and the updated BSP & OS is targeted to be completed and published via LVFS around May.
We appreciate your patience in the meantime.
 
Thank you for your patience.
The DGX OS OTA2604 release, along with the related firmware updates, has completed validation and has been uploaded to LVFS.
The latest DGX OS image is now available for download from the MSI EdgeXpert official website below:

MSI EdgeXpert Download Page

Please proceed with the update and let us know if there are any further questions or issues.
 
@Athrun how can I just apply the PD firmware without re-imaging the whole device (for the 4th time)? I have two MSI Edgexperts and one is fine but the other has PD0 FW0 5.22, PD0 FW1 4.10, PD1 FW0 5.22 and PD1 FW2 4.10. I have re-imaged using the OTA264 twice now and it's not updating the firmware. If I look for updates it says there aren't any (I've seen on another post that perhaps the driver manufacturer ID is set to lenovo for the MSI one which is why it's not being picked up??). Is there a way to just force an update without trying to re-image the whole machine again??
Also, can anyone explain the difference between PD0 and 1 and FW 0 and 1 please?
BTW, the impact to me isn't just the terrible nccl slowdowns but also screen blanking and reboots :/
 
pau153d02d2 said:
@Athrun how can I just apply the PD firmware without re-imaging the whole device (for the 4th time)? I have two MSI Edgexperts and one is fine but the other has PD0 FW0 5.22, PD0 FW1 4.10, PD1 FW0 5.22 and PD1 FW2 4.10. I have re-imaged using the OTA264 twice now and it's not updating the firmware. If I look for updates it says there aren't any (I've seen on another post that perhaps the driver manufacturer ID is set to lenovo for the MSI one which is why it's not being picked up??). Is there a way to just force an update without trying to re-image the whole machine again??
Also, can anyone explain the difference between PD0 and 1 and FW 0 and 1 please?
BTW, the impact to me isn't just the terrible nccl slowdowns but also screen blanking and reboots :/
Click to expand...
Thanks for the detailed report and clarify as below:

* PD0/PD1 & FW0/FW1:
- PD0 and PD1 are the two independent USB-C PD controllers on the MS-C931. FW0 is the active slot, FW1 is the backup. Your active slot (5.22) has updated correctly — the backup slot retaining the prior version is expected behavior by design.

* Re-imaging:
- OS re-imaging has no effect on PD firmware, which is flashed independently. For further update assistance, please contact MSI Support directly rather than attempting manual methods.

* LVFS / DMI issue:
- This is a backend issue on the LVFS side, already reported for correction. No action is required on your end, and it is unrelated to your PD firmware status.

* NCCL slowdowns / screen blanking / reboots:
- These symptoms can have multiple causes and deserve proper investigation rather than a quick guess here. To make sure nothing gets missed, please open a support case at support.msi.com include your full firmware versions (SOC FW, EC FW, PD FW), DGX OS version, and a description of when the issues occur.
 
@Athrun Thanks. I just saw you're other response on the LVFS. Just for my understanding the other node I have show PD0\1 FW0/1 all with 5.22, hence why I thought this was odd? Why is it different for another node, is it based on build time and what firmware is available on build?
 
Back
Top