Skip to main content

Install the GPU board

Use this information to install the GPU board.

About this task

S001
shock hazard
DANGER
danger
Electrical current from power, telephone, and communication cables is hazardous.

To avoid a shock hazard:

  • Connect all power cords to a properly wired and grounded electrical outlet/source.
  • Connect any equipment that will be attached to this product to properly wired outlets/sources.
  • When possible, use one hand only to connect or disconnect signal cables.
  • Never turn on any equipment when there is evidence of fire, water, or structural damage.
  • The device might have more than one power cord, to remove all electrical current from the device, ensure that all power cords are disconnected from the power source.
Attention
  • Read the following sections to ensure that you work safely.
  • Turn off the corresponding DWC tray that you are going to perform the task on.
    Note
    If Shared I/O adapters are installed, power off the auxiliary node (node 1/3/5/7/9/11) first, and then power off the primary node (node 2/4/6/8/10/12).
  • A torque screwdriver is available for request if you do not have one at hand.

  • To avoid damaging the water loop, always use the water loop carrier when removing, installing or folding the water loop.

  • Hold the long sides of the GPU board with two hands while removing the new GPU board from the package box.
    Figure 1. Removing GPU board from the package box
    Removing GPU board from the package box
  • Once the GPU board is removed from the plastic protective bag, hold two handles with both hands to move the GPU board.

    Figure 2. Removing GPU board from the package box
    Removing GPU board from the package box
Note
  • Ensure you have “SD650 V2 or SD650-N V2 Neptune DWC Waterloop Service Kit “ in hand to install components.

  • To prevent potential thermal issues, change the Misc setting in the BIOS from Option3 (default value) to Option1 if the following two conditions are met:
    • The server is equipped with a GPU adapter.

    • The UEFI firmware version is ESE122T or later.

    For the method of changing the Misc setting, see https://support.lenovo.com/us/en/solutions/TT1832.
Prepare the following screwdrivers to ensure you can install and remove corresponding screws properly.
Table 1. Torque screwdriver type list
Torque screwdriver type listScrew type
Torx T10 head screwdriverTorx T10 screw
Phillips #1 head screwdriver or 3/16" hex head screwdriverPhillips #1 screw
Phillips #2 head screwdriverPhillips #2 screw
The following illustration shows the GPU numbering.
Figure 3. GPU numbering
GPU numbering

Procedure

  1. Gently place the GPU board down and install the five Phillips #1 screws with a torque screwdriver sets to the proper torque.
    Note
    1. For reference, the torque required for the screws to be fully tightened/removed is 0.5-0.6 newton-meters, 4.5-5.5 inch-pounds.

    2. When you install the GPU board from the node, avoid touching the connectors on the GPU board. Be careful not to damage any surrounding components inside the node.

    Figure 4. GPU board installation
    GPU board installation
  2. Install the two Torx T10 screws with a torque screwdriver sets to the proper torque.
    Note
    For reference, the torque required for the screws to be fully tightened/removed is 0.5-0.6 newton-meters, 4.5-5.5 inch-pounds.
    Figure 5. Screws installation
    Screws installation
  3. Reinstall the retimer board (see Install the retimer board).
  4. Apply the new thermal grease on GPUs.
    1. If there is any old thermal grease on four GPUs and the cold plates, gently clean the top of the four GPUs and the cold plates using an alcohol cleaning pad.
    2. If you have cleaned the top of the GPUs with an alcohol cleaning pad, make sure to apply the new thermal grease after the alcohol has fully evaporated.
    3. Apply gray thermal grease to the top of the four GPUs with a syringe by forming four dots spaced as shown below, with each dot consisting of about 0.5 gram (about 0.225 ml) of gray thermal grease. Each syringe contains 1 gram of thermal grease, sufficient for two dots of thermal grease.
    Figure 6. Thermal grease application
    Thermal grease application
  5. Replace the existing putty pad with the new one.
    Figure 7. Putty pad
    Putty pad
  6. Check the gap pads on the water loop, if any of them are damaged or missing, replace them with the new ones.
    Figure 8. Gap pads on the water loop
    Gap pads on the water loop
  7. Reinstall the water loop.
    1. ❶ Carefully rotate the top side of the water loop.
    2. ❷ Carefully insert the quick connect into the tray opening as shown.
    3. ❸ Carefully align the water loop with eight guide pins on four GPU cold plates; then, gently put the water loop down and ensure it is firmly seated on the GPU board.
      Figure 9. Water loop installation
      Water loop installation
  8. Loosen all water loop carrier screws (15x Phillips #2 screws).
    Figure 10. Loosening water loop carrier screws
    Loosening water loop carrier screws
  9. Carefully lift the water loop carrier up and away from the water loop.
    Figure 11. Water loop carrier removal
    Water loop carrier removal
  10. Reinstall water loop screws (7x Torx T10 screws) with a torque screwdriver sets to the proper torque.
    Note
    For reference, the torque required for the screws to be fully tightened/removed is 0.5-0.6 newton-meters, 4.5-5.5 inch-pounds.
    Figure 12. Water loop screws installation
    Water loop screws installation
  11. Reinstall GPU cold plate screws (4x Phillips #1 screws per GPU cold plate, total of 16x Phillips #1 screws) with a torque screwdriver set to the proper torque. For reference, the torque required for the screws to be fully tightened/removed is 0.34-046 newton-meters, 3-4 inch-pounds.
    Figure 13. GPU cold plate screws installation
    GPU cold plate screws installation
    1. Push down the GPU cold plate with your palm to reduce the gap between the GPU cold plate and the GPU.
    2. Press the torque screwdriver against the screw so that the screw is engaged with the GPU.
    3. Follow the screw sequence specified on the GPU cold plate label, and fasten each screw for 720 degrees with a torque screwdriver set to the proper torque.
      Note
      For reference, the torque required for the screws to be fully tightened/removed is 0.46–0.34 newton-meters, 4–3 inch-pounds.

      GPU cold plate installation
    4. Make sure that the GPU cold plate is lowered into the node and its surface is flat without tilting. If the GPU cold plate is tilted, unfasten the screws, and repeat Step 1 to Step 3.
    5. Repeat Step 3 until the screws are fully tightened.
    6. Make sure the height of each screw is 12.5±0.3 millimeter (0.49±0.01 inch) and is fully compressed. If not, repeat the GPU cold plate installation steps.

      Height of each tightened screw
      Note
      Inspect the screws to make sure they are fully compressed.

      screw height inspection
  12. Reinstall the four Torx T10 screws (per node) to secure the quick connect.
    Figure 14. Screws installation
    Screws installation
  13. Reinstall the clamp plate or the drive depending on your configuration.
    • Clamp plate installation: Install the two screws to secure the clamp plate.

      Figure 15. Clamp plate installation
      Clamp plate installation
    • Drive installation: see Install the drive in the GPU node.

  14. Remove the connector cover if necessary.
    Figure 16. Connector cover removal
    Connector cover removal
  15. Connect GPU power cable.
    Figure 17. GPU power cable installation
    GPU power cable installation
  16. See Cable installation order to connect and route required cables accordingly.
After you finish
  1. Reinstall the front and the rear cross braces (14x Phillips #1 screws).

    Figure 18. Cross brace installation
    Cross brace installation
  2. Reinstall the tray cover (see Install the tray cover).

  3. Reinstall the tray (see Install a DWC tray in the enclosure).

    Note
    For safety, use the lift tool to install the tray into the rack.
  4. Connect all required external cables to the enclosure.
    Note
    Use extra forces to connect QSFP cables to the enclosure if Mellanox ConnectX-6 adapters are installed.
  5. Check the power LED on each node to make sure it changes from fast blink to slow blink to indicate all nodes are ready to be powered on.

  6. Update the Universally Unique Identifier (UUID) and DMI/SMBIOS data with new vital product data (VPD). Use the Lenovo XClarity Provisioning Manager to update the UUID and DMI/SMBIOS data. See Update the Universal Unique Identifier (UUID) and Update the asset tag.

  7. Enable TPM/TCM. See Enable TPM/TCM

  8. Optionally, enable Secure Boot.

Demo video

Watch the procedure on YouTube