Skip to content
nodeza.learn
DOCA course/8. Dell OEM/Lesson 8.2

iDRAC, BIOS, CPLD and the DPU

S8·E203:40, and the DPU is gone · Bridge call, the imaging bureau's data center, night shift

S8·E2Analyze~35 minsources checked todayverified against Dell KB 000227031 (v3, 21 May 2026), KB 000379421 (17 Jun 2026), KB 000300192 (22 May 2026), Dell DUP 4RR0M, Broadcom KB 379391, DOCA 3.5.0 modes page, 2026-09-06

Builds on: PowerEdge support matrix, Switching modes safely, Dell-specific bring-up: iDRAC, BIOS, CPLD, aux power, DSE

Before you read: what do you already know?

3 quick questions. Wrong answers are fine and expected; trying first makes the lesson stick.

After this lesson you can

  • Explain why a BlueField-3 in DPU mode adds a fourth state machine to a PowerEdge boot and what goes wrong when it is not synchronized.
  • Match the three Dell KB symptoms (No Memory Found, Device not detected, fatal PCIe error) to KB 000227031, 000379421 and 000300192 and their fixes.
  • Decide whether a given firmware comes from a Dell DUP through iDRAC or from NVIDIA's DOCA downloads, and what power action must follow.
  • State precisely what Dell does and does not document about iDRAC and the DPU, including OS-to-iDRAC pass-through.
  • Triage "the card is gone after a BIOS update" from the POST message and Lifecycle Log before proposing a replacement.

Episode 2 — 03:40, and the DPU is gone

The situation · Bridge call, the imaging bureau's data center, night shift

Twenty to four, and the bridge is crowded: the night-shift operator, the Dell SE with a coffee going cold in a different time zone, and procurement, asking the lead time on a replacement card. An R760 took a BIOS update in last night’s change window; the operator says the DPU has vanished and an RMA screen is already open. You ask for two artifacts before anyone touches anything: the POST screen text and an export of the iDRAC Lifecycle Log. The network lead writes both in a notebook that no longer opens flat.

Those two decide the call, because a BlueField-3 in DPU mode is not a card. It is a small computer whose Arm complex runs its own firmware and OS, shuts down gracefully, and comes back on its own timeline — joining a boot the BIOS, the iDRAC and the CPLD were already sequencing between themselves. That independence is the whole point of the offload, and it costs a handshake. Dell says what happens when the handshake is mistimed: after BIOS updates, PCIe Switch Board updates, BIOS configuration changes or a cold reboot, “the timing of the various state machines are not synchronized, leading to the event” — the POST message “No Memory Found”, after which the server waits about a minute, reboots itself, and runs normally with all its memory.[1]

In DPU mode the boot is a negotiation between four state machines, not three. The operator reads the screen aloud; the words decide whether this is a timing event, a training failure, or a power cable.

1Four state machines on one boot

A PowerEdge boot already coordinates three firmware actors: the BIOS, the iDRAC baseboard controller and the CPLD that sequences power and resets. A BlueField-3 in DPU mode adds a fourth, because its Arm complex runs its own firmware and OS, shuts down gracefully, and expects to come back on its own timeline. Dell KB 000227031 describes what happens when those timelines disagree: a firmware combination introduced a graceful DPU shutdown, and after BIOS updates, PCIe Switch Board (PSB) updates, BIOS configuration changes or a server cold reboot “the timing of the various state machines are not synchronized, leading to the event”.[1] The event is the POST message “No Memory Found”; the server “waits for approximately 1 minute, then it reboots automatically” and is then “fully functional and memory is seen as expected”.[1]

The same coupling appears in the other two Dell articles. KB 000379421 documents a PCIe training failure where the Lifecycle Log repeats “Device not detected: Nvidia Network Adapter”, caused by the BlueField-3 device firmware, not by the server.[2] KB 000300192 documents a fatal PCIe error after a DPU firmware update that was followed by a warm reboot instead of a power cycle.[3] NVIDIA’s own modes page states the general rule: after a mode change “Arm and NIC components must undergo a reset. Power cycle is recommended.”[7]

The triage tree below starts at the Dell branch. Each leaf is one of these documents or a field failure from the forums; the point of this lesson is that you can walk it from the message on the screen, before anyone proposes a replacement.[2][11]

SymptomBlueField-3 on a Dell Pow…SymptomDell platform: POST, iDRA…Symptom"No Memory Found" POST e…SymptomvSphere Distributed Serv…SymptomCard disappeared after a…SymptomFirmware: Dell DUP vs NV…
Symptom

Dell platform: POST, iDRAC, vSphere, firmware

Start at the Dell branch: POST message or Lifecycle Log entry → KB → first action. Click a leaf for the exact KB wording.

2KB 000227031: 'No Memory Found'

The title is “PowerEdge: ‘No Memory Found’ event is displayed on NVIDIA Bluefield 3 enabled servers during POST”; the article is version 3, last modified 21 May 2026.[1] The impacted platforms are PowerEdge R660, R760, R760XA and XE9680.[1] The impacted versions are specific: CPLD 1.1.5 and 1.1.7, iDRAC 7.10.50.00 and BlueField-3 firmware 32.40.1000.[1] The affected DPUs are the B3210e and B3220 “when operating in DPU mode”; the B3140H “comes set in Super NIC mode. If it is changed to DPU Mode, it can also encounter this event.”[1]

Read that last sentence as a rule: exposure follows the mode, not the SKU. A SuperNIC customer who flips to DPU mode to run a DOCA service inherits the DPU-mode boot behaviour, including this one.[1][7]

The resolution is not on the host and not on the card. “Graceful shutdown is disabled in an upcoming CPLD release”; until that release is applied, the interim action is to wait for the automatic reboot.[1] Nothing in the article asks for DIMM reseating, a BFB reflash or a card replacement, and an FAE who lets any of those happen has misdiagnosed a timing event as a hardware fault.[1]

When you verify a customer’s exposure, collect the four versions in one table: CPLD, iDRAC, BIOS (for the change history) and BlueField-3 firmware from mlxfwmanager --query.[1][2] If they match the KB and the card is in DPU mode, the customer is exposed; the plan is the CPLD release, scheduled with the firmware bundle discussed in the next lesson.[1]

3KB 000379421 and KB 000300192: 'Device not detected' and fatal PCIe errors

KB 000379421, “BlueField-3 DPU PCIe Initialization Failure”, last modified 17 June 2026, describes a Lifecycle Log that repeats “Device not detected: Nvidia Network Adapter” (in the example, Slot 33, PR8).[2] Dell calls it a PCIe training failure caused by the BlueField-3 device firmware, “a known NVIDIA issue”, fixed from firmware 32.46.3048, released 14 August 2025.[2] The resolution path has three rungs. First, a full power cycle. Second, if the card returns, update to 32.46.3048 or later via the “NVIDIA DOCA Software Framework” download and keep “the BFB image aligned with the updated firmware”. Third, if the card is still undetected, “open a service request for a replacement card”.[2]

Note what that path says about ownership: the firmware that fixes a Dell-logged failure comes from NVIDIA’s DOCA downloads, and the BFB (the Arm OS image) must move with it.[2] The Spectrum-X validated stack v2.3.1 row pairs BlueField-3 firmware 32.50.1002 with DOCA-Host 3.5.0-082, which is the current reference point when you choose the target version.[13]

KB 000300192, for the XE9680L (Dell part HFRWM), last modified 22 May 2026, is the corollary. After updating DPU firmware without a power cycle, the Lifecycle Log reports “A fatal error was detected on a component at bus [number], device 0, function 0”; per NVIDIA, “a system power cycle is required after updating the DPU firmware”, and a warm reboot is insufficient.[3] NVIDIA’s vSphere HowTo says the same for a mode change: “Power cycle the system after switching to NIC mode.”[6]

One field failure belongs on the same branch even though no Dell KB covers it. A DPU-mode card that enumerates but stays in “FW pre-initializing” with no host netdevs, whose console shows “ATX power not detected! Halting system!!”, is missing its 8-pin PCIe auxiliary cable, and a BIOS update that coincided with a service visit is a red herring.[11]

4DUP or NVIDIA flow: where each firmware comes from

Dell’s Update Package model is documented for BlueField-2. Package 4RR0M, “DPU_Firmware_4RR0M_WN64_24.36.75.06_01.EXE” (30.66 MB, importance Optional, 5 Dec 2023), carries NIC firmware for BlueField-2 DPUs, targets VMware ESXi 8.0 and Windows, and can “ONLY be applied via the iDRAC/Lifecycle Controller”; “Executing this firmware DUP from the host OS is not supported.”[4]

For BlueField-3, the fetched Dell pages point elsewhere. KB 000379421 sends you to the NVIDIA DOCA downloads for the BF-FW-Bundle and BFB and tells you to keep them aligned — that is the bfb-install and mlxfwmanager flow, not a DUP.[2] KB 000300192 does not say how the DPU firmware was updated on the XE9680L — the Lifecycle Controller appears in it only as the log that records the resulting “fatal error … at bus 184, device 0, function 0”, and the KB’s cause cites the NVIDIA BlueField-3 DPU Controller User Manual: “a system power cycle is required after updating the DPU firmware”.[3] Do not infer an iDRAC-driven BF-3 firmware path from it. Whether Dell publishes BlueField-3 firmware, or any BFB, as a DUP is not confirmed by any fetched page; say “not documented” rather than “no”.[2][4]

Whichever path applies, the power rule is the same: a full system power cycle after any DPU firmware update.[3] And the version plan should be one bundle. DOCA’s dependency policy makes the DOCA-OFED host driver Level-1 compatible with older and newer BlueField firmware, while DOCA services and the full DOCA-Host profile are Level 2, guaranteed only within the same October-to-July cycle.[9] Combined with the Dell KBs, whose issues are tied to exact CPLD, iDRAC and firmware versions, that means BIOS, iDRAC, CPLD, BlueField firmware and BFB move together, with a power cycle at the end.[1][9]

5What iDRAC does and does not document about the DPU

Be exact here, because customers repeat what they hear. The iDRAC9 7.x User’s Guide has a top-level topic titled “Data Processing Unit (DPU)”; the fetched page rendered only the table of contents, so the topic exists but its body is unverified from a primary source.[8] Search snippets attribute to it a DPU definition, an iDRAC9 Enterprise or Datacenter licence requirement, Redfish one-time boot and ARM-UEFI/BMC firmware updates, and attributes such as CSP Mode and DPUOSDeploymentTaskState; none of that was read in the guide itself.[8]

OS-to-iDRAC pass-through is a documented iDRAC feature for the host operating system over the LOM or a USB NIC. No fetched Dell page ties that feature to BlueField management or describes a path from iDRAC into the DPU’s Arm OS.[8] NVIDIA’s modes page documents NC-SI OEM commands a platform BMC can use to read and set the operating mode — Get Command 0x13 / Parameter 0x33 returns the offload-engine state (0x0 enabled = DPU mode, 0x1 disabled = NIC mode) and Command 0x12 / Parameter 0x33 sets it; the page says nothing about a BMC shutting down or resetting the DPU. Whether iDRAC implements these commands on PowerEdge is unconfirmed.[7] The R760 “Management Interface Card (MIC)” appears only in a search snippet of a community page that returned 403.[1]

What Dell does document about iDRAC and the DPU is the Lifecycle Log: every KB in this lesson identifies the failure from a Lifecycle Log entry — “Device not detected: Nvidia Network Adapter”, “A fatal error was detected on a component at bus [number], device 0, function 0” — and from the POST screen.[2][3][1] So the honest iDRAC story for an FAE is: it is your evidence source and, for Dell-sourced firmware, your update channel; do not promise it as a management console for the Arm OS until you have read the DPU topic in the guide.[8][4]

6DSE mode facts from Broadcom KB 379391

Broadcom KB 379391 is the operational reference for BlueField-3 modes on vSphere. “BlueField-3 DPU out of the factory, by default is in the DPU mode and VMware Distributed Services Engine (DPU) is disabled.”[5] DPU mode for DSE means the “Embedded Arm system controls the NIC resources and data path” with DSE enabled in device settings; in NIC mode “BlueField behaves exactly like an adapter card from the perspective of the external host”; the parameter is INTERNAL_CPU_OFFLOAD_ENGINE, 0 for DPU mode (enabled) and 1 for NIC mode (disabled).[5] The consequence for planning: “When you switch BlueField-3 from NIC mode to VMware Distributed Services Engine enabled DPU mode, you must re-install ESXi”, and this applies in both directions on vSphere 8.0 U3b and higher.[5]

NVIDIA’s HowTo gives the ESXi command. With MFT installed on ESXi 8.0.x: /opt/mellanox/bin/mlxconfig -d mt41692_pciconf0 set INTERNAL_CPU_OFFLOAD_ENGINE=1, then power cycle the system, then verify with mlxconfig -d mt41692_pciconf0 q | grep -i offload showing DISABLED.[6]

Layer Dell’s boundary on top: DSE is not supported on NVIDIA Channel cards, and Dell will not support it if enabled.[10] So on PowerEdge the DSE question has two gates — is the card a Dell SKU, and has the customer budgeted an ESXi reinstall — and a mode switch on a VMware host is a rebuild, not a reboot.[10][5]

Triage: 'the card is gone after the BIOS update'

Situation: PowerEdge R760, B3220 in DPU mode, BIOS updated through iDRAC overnight. The ticket says “the DPU disappeared”.

  1. Collect evidence before theory: the POST message from the virtual console and an export of the iDRAC Lifecycle Log. Three different messages map to three different Dell articles.[1][2][3]
  2. Case A — POST shows “No Memory Found”, the server sits for about a minute, reboots itself and runs normally. This is KB 000227031. Record CPLD, iDRAC and BlueField firmware; if they are 1.1.5/1.1.7, 7.10.50.00 and 32.40.1000 the match is exact. Action: let it reboot, schedule the CPLD release that disables graceful shutdown. No replacement.[1]
  3. Case B — the Lifecycle Log repeats “Device not detected: Nvidia Network Adapter” and lspci shows no BlueField functions. This is KB 000379421. Action: full power cycle from iDRAC (not a warm reboot). If the card returns, run sudo mlxfwmanager --query; if the firmware is older than 32.46.3048, update from the NVIDIA DOCA downloads, install the matching BFB, and power cycle again. If the card is still undetected after the power cycle, open a service request for a replacement.[2]
  4. Case C — the Lifecycle Log shows “A fatal error was detected on a component at bus [n], device 0, function 0” and the change window also included a DPU firmware update. This is KB 000300192: the update was followed by a warm reboot. Action: full power cycle.[3]
  5. Case D — the card enumerates but sits in “FW pre-initializing” with no host netdevs, and the console shows “ATX power not detected”. The BIOS update is a coincidence; the 8-pin auxiliary cable was disturbed during service. Action: reseat the cable, power on.[11]
  6. Close the ticket with a table: message, KB, versions before and after, action, outcome, and the power-cycle rule for the next change.[2][3]

04:15, RMA cancelled

How it ended

The screen said “No Memory Found”, the box rebooted itself a minute later into a healthy server, and the replacement card never shipped.[1] The versions matched exactly — CPLD 1.1.7, iDRAC 7.10.50.00, BlueField-3 firmware 32.40.1000 — which makes it the timing event, whose fix is a CPLD release that disables the DPU graceful shutdown, not a new card.[1] What you leave them with: “Nothing failed tonight; schedule the CPLD release with the firmware bundle, and after any DPU firmware update power cycle the system instead of rebooting it.”[3] The operator prints one more label for the rack door: “POWER CYCLE IS NOT A REBOOT”. By Friday the change board has read the ticket and has one question for the design review: is this supported?

Lab

  1. Pre-flight inventory, all read-only. From the iDRAC firmware inventory record BIOS, iDRAC and CPLD versions. On the host: sudo mst start && sudo mlxfwmanager --query for the BlueField-3 firmware version, and sudo mlxconfig -d /dev/mst/<dev> q INTERNAL_CPU_OFFLOAD_ENGINE for the mode.[1][7]
  2. Compare against the KBs: iDRAC 7.10.50.00 with CPLD 1.1.5 or 1.1.7 and firmware 32.40.1000 in DPU mode = exposed to KB 000227031; firmware older than 32.46.3048 = exposed to KB 000379421. Write the exposure verdict in your notes.[1][2]
  3. Export the Lifecycle Log from iDRAC and search it for “No Memory Found”, “Device not detected” and “fatal error was detected”. Expected on a healthy lab box: none. If present, record the timestamps next to the change history.[2][3]
  4. sudo dmesg | grep -i mlx5 — record the PCIe link and firmware lines for the evidence pack. If the driver reports the device stuck initializing, check the console for the ATX message and the aux cable before anything else.[11]
  5. Do not update firmware or change mode in this lab. If step 2 says the box is exposed, write the change as a plan for lesson 8.3: target firmware, BFB aligned, full power cycle at the end.[2][3]

Retrieval check

10 questions from memory. Answer before looking anything up; misses become flashcards.

Explain it to a Dell SE

Explain to a Dell SE, in four sentences, why 'the DPU disappeared after the BIOS update' is usually one of three different problems, and how the iDRAC Lifecycle Log tells them apart.

12 flashcards for this lesson — 0 in deck. Spaced review lives at /review.

Sources

Facts in this lesson were checked against Dell KB 000227031 (v3, 21 May 2026), KB 000379421 (17 Jun 2026), KB 000300192 (22 May 2026), Dell DUP 4RR0M, Broadcom KB 379391, DOCA 3.5.0 modes page, 2026-09-06. Dates are when each page was fetched.

  1. Dell KB 000227031 — 'No Memory Found' on NVIDIA BlueField-3 enabled PowerEdge during POST · fetched 2026-09-06
  2. Dell KB 000379421 — PowerEdge: BlueField-3 DPU PCIe Initialization Failure · fetched 2026-09-06
  3. Dell KB 000300192 — XE9680L: a power cycle of the system is required after updating DPU firmware · fetched 2026-09-06
  4. Dell driver 4RR0M — NIC firmware for NVIDIA BlueField-2 DPUs (DUP) · fetched 2026-09-06
  5. Broadcom KB 379391 — Configuring BlueField-3 into DPU mode for vSphere DSE or NIC mode · fetched 2026-09-06
  6. HowTo Configure NVIDIA BlueField-3 to NIC Mode on VMware vSphere 8.0 · fetched 2026-09-06
  7. BlueField Modes of Operation · fetched 2026-09-06 · DOCA 3.5.0
  8. iDRAC9 7.x User's Guide — Data Processing Unit (DPU) topic (TOC only rendered) · fetched 2026-09-06
  9. DOCA Dependency Compatibility Policy · fetched 2026-09-06 · DOCA 3.5.0
  10. Dell KB 000225111 — NVIDIA Channel DPU VMware DSE Support · fetched 2026-09-06
  11. NVIDIA forum — BlueField-3 (DPU mode) stuck in FW pre-initializing, no host netdevs (root cause: ATX power not detected) · fetched 2026-09-06
  12. NVIDIA forum — BF3 RShim not working (bfb-install done, no ssh to 192.168.100.2) · fetched 2026-09-06
  13. NVIDIA Spectrum-X Validated Solution Stack · fetched 2026-09-06 · DOCA 3.5.0

The same idea elsewhere

Other lessons that cover this ground, sometimes from another course's angle.