Memory configuration error when VGA card installed - IBM System x3650 M2, x3650 M3 (4255, 7945)

Resolving The Problem

Source

RETAIN tip: H203415

Symptom

With the NVIDIA Quadro FX 580 Graphics Adapter, Option part number 45K1671, installed, after the system powers on, light emitting diodes (LEDs) on the Lightpath diagnostics panel indicate a memory configuration error.

Affected configurations

The system may be any of the following IBM servers:

• System x3650 M2, type 4199, any model
• System x3650 M2, type 7947, any model
• System x3650 M3, type 4255, any model
• System x3650 M3, type 7945, any model

The system is configured with one or more of the following IBM Options:

• NVIDIA Quadro FX 580 Graphics Adapter, Option part number 45K1671, replacement part number (CRU) 46R2786

This tip is not software specific.

Workaround

Follow these steps to try to resolve the error:

1. Ensure all system and adapter firmwares are at latest levels. Firmware updates may reduce Read Only Memory (ROM) size requirements, so this may be a solution.
2. Disable the Preboot eXecution Environment (PXE) ROM of on-board Network Interface Controllers (NICs). The simplest solution to some PCI Option ROM space errors is to reduce the base system ROM requirements to the minimum necessary. Typically, this can be achieved by disabling the PXE (Network boot) ROM capability of the on-board Ethernet.
   
   Note: This does not disable the device in the operating system. It only disables its capability to perform a network boot.
   
   
   1. Select F1 setup
   2. Select System Settings, then Network, and then PXE Configuration
   3. Select the Media Access Control (MAC) address of the on-board Network Interface Controller (NIC) #1
   4. Change the Enable PXE / PXE Mode setting to "Disabled"
   5. Select Save Changes
   6. Select the MAC address of the on-board NIC #2
   7. Change the Enable PXE / PXE Mode setting to "Disabled"
   8. Select Save Changes
      
      
3. Disable all legacy option ROMs for all devices that are not booting.
   
   1. Boot the machine and press F1 to enter the Unified Extensible Firmware Interface (UEFI) menu.
   2. Select System Settings, then Devices and I/O Ports, and then Enable /Disable legacy ROM execution.
      
      Note: If the machine uses PXE or Storage Area Network (SAN) booting, do not disable the Legacy ROMs for adapters that are actually booting a Legacy (non-Extensible Firmware Interface (EFI)) OS on the machine.
      
      
4. Change the ROM execution order.
   
   1. Select F1 setup
   2. Select System Settings, then Devices and I/O, and then Set option ROM execution order.
   3. Ensure that the on-board LSI storage controller is first in the list, before the on-board ethernet devices.
      
      
5. Move Fibre Channel and Fibre Channel over Ethernet cards to earlier slots as they tend to require more ROM space and should initialize first to reduce the chance that the issue might reoccur.
6. Make sure PCIe adapters with PCI firmware spec 2.1 are in earlier slots. PCIe adapter with PCI firmware spec 3.x does not need as much option ROM space due to design changes.
7. Change PCIe ASPM power setting:
   1. Select F1 setup
   2. Select System Settings, then Power, then PCI Express ASPM, and then Enable L1 only

Additional information

There are two (2) root cause failures for this issue:

1. First one was originally discovered on Legacy (non-UEFI) machines and extensive documentation was provided to explain how to resolve the symptoms. For more details, see the PDF file, "IBM Support Info - Resolving 1801, 1802 Errors", which is referenced in RETAIN Tip H194252 (MIGR-5078445), and available at the following location: ftp://ftp.software.ibm.com/systems/support/system_x_pdf/resolving_1801-1802_errors.pdf
   
   These errors are due to a permanent restriction of legacy PCI ROM space architecture.
   
2. The Intel chipset has a limitation with the Active State Power Management (ASPM) L0s function. Changing the ASPM function to L1 will cover the chipset limitation.