Remove a rear B200 GPU

Follow instructions in this section to remove a rear B200 GPU. The procedure must be executed by a trained technician.

About this task

Attention

• Read Installation Guidelines and Safety inspection checklist to ensure that you work safely.
• Power off the server and peripheral devices and disconnect the power cords and all external cables. See Power off the server.
• If the server is installed in a rack, slide the server out on its rack slide rails to gain access to the top cover, or remove the chassis from the rack. See Remove the server from rack.
• Two people and one lifting device on site that can support up to 400 lb (181 kg) are required to perform this procedure. If you do not already have a lifting device available, Lenovo offers the Genie Lift GL-8 material lift that can be purchased at Data Center Solution Configurator. Make sure to include the Foot-release brake and the Load Platform when ordering the Genie Lift GL-8 material lift.
• A torque screwdriver is available for request if you do not have one at hand.

Note

Make sure you have the required tools listed below available to properly replace the component:

• Torx T15 head screwdriver
• 2 x Torx T15 200mm extension bit
• Phillips #1 head screwdriver
• Phillips #2 head screwdriver
• Alcohol cleaning pad
• B200 PCM
• B200 SXM6 PAD-1
• B200 SXM6 PAD-2
• B200 GPU Service Kit

B200 (GPU & Retimer NVSwitch) (service & shipping bkt) Kit are reusable and mandatory when servicing GPUs and GPU cold plate modules. It is recommended to keep them at the facility where the server operates for future replacement needs.

Important

Putty pad/phase change material (PCM) replacement guidelines

• Before replacing the putty pad/PCM, gently clean the hardware surface with an alcohol cleaning pad.
• Hold the putty pad/PCM carefully to avoid deformation. Make sure no screw hole or opening is blocked by the putty pad/PCM.
• Do not use expired putty pad/PCM. Check the expiry date on putty pad/PCM package. If the putty pads/PCM are expired, acquire new ones to properly replace them.

The following illustration shows the B200 GPU numbering and corresponding slot numbering in XCC.

Figure 1. B200 GPU numbering

Physical GPU socket	Slot numbering in XCC	Logical number in nvidia-smi
GPU 1	Slot 21	4
GPU 2	Slot 24	7
GPU 3	Slot 22	5
GPU 4	Slot 23	6
GPU 5	Slot 17	0
GPU 6	Slot 20	3
GPU 7	Slot 18	1
GPU 8	Slot 19	2

Procedure

1. Make preparation for this task.
   1. Remove the front top cover. See Remove the front top cover.
   2. Remove the rear top cover. See Remove the rear top cover.
   3. Remove the fan cage. See Remove the fan cage (trained technician only).
   4. Remove the CPU complex. See Remove the CPU complex.
   5. Remove the power complex. See Remove the power complex.
   6. Disconnect the cables and remove them from the GPU complex if necessary. Before disconnecting the cables, make a list of each cable and record the connectors the cable is connected to. Refer to Internal cable routing.
2. Remove the rear fan cage support bracket.
   1. Unfasten the eight M3 screws that secure the rear fan cage support bracket to the the chassis.
   2. Unfasten the four M3 screws that secure the rear fan cage support bracket to the fan cage.
   3. Grasp the rear fan cage support bracket to lift it from the fan cage.
   Figure 2. Removing the rear fan cage support bracket
   
   
3. Locate the rear GPU.
4. Remove the leakage sensor module cable from the cable clips, route it away from the cold plate and onto the adjacent cold plate.
   Figure 3. Removing the leakage sensor module cable
   
   
5. Loosen the screws by 360 degrees following the screw sequence: specified on the cold plate label, and fully loosen the four Torx T15 screws with a torque screwdriver set to the proper torque.
   Note

   • Loosen the screws with a torque screwdriver set to the proper torque. For reference, the torque required for the screws to be fully loosen is 5.3±0.212 inch-pounds, 0.6±0.024 newton-meters.
   • Ensure the captive screws are completely loosen before removing the cold plate module.
   • Make sure to follow screw sequence to prevent cold plate tilting.
   Figure 4. Removing the GPU cold plate
   
   

   Note

   • If necessary, use the Tim breaker screw to separate the cold plate from the GPU. Ensure to fully loosen all the cold plate screws before fastening the TIM breaker screw.

     • Open the lid of the TIM breaker screw.
       
       
     • Fasten the TIM breaker screw to separate the cold plate from the GPU.
       
       

   • After usage, return the TIM breaker screw to its original position.

     • Loosen the TIM breaker screw to return it to its initial position.
       
       
     • Close the lid. If the lid cannot be closed, the TIM breaker screw needs to be further loosened.
       
       
6. Install the service bracket onto the rear GPU cold plate.
   1. Align the two guide pins at the bottom of the service bracket with the guide holes on the GPU cold plate; then, lower it onto the cold plate.
   2. Fasten the two captive screws (PH1, 2 x M3, 0.5 newton-meters, 4.3 inch-pound) to secure the service bracket to the cold plate.
      Figure 5. Installing the service bracket onto the GPU cold plate
      
      

      Note

      Avoid pinching the leakage sensor module cable when installing the bracket.
      
      
7. Install the service bracket and the GPU cold plate assembly onto the rear B200 GPU cold plate module manifold.
   1. Flip over the service bracket and the GPU cold plate assembly; then, align the captive screw and two guide pins with the screw hole and guide holes on the manifold.
      Note

      Ensure that the hoses are not bent inward to avoid stress damage to the joints.
      
      
   2. Fasten the captive screw (PH1, 1 x M3, 0.5 newton-meters, 4.3 inch-pound) to secure the service bracket and GPU cold plate assembly onto the manifold.
      Figure 6. Installing the service bracket and the GPU cold plate assembly
      
      
      Note

      Ensure to install the service bracket and GPU cold plate assembly in the screw holes and guide holes corresponding to the specific GPU slot number.

      Figure 7. Service bracket and GPU cold plate assembly installation location
      
      

      Table 1. GPU cold plate and service bracket assembly installation location

      Installation location	GPU slot number
      1	GPU 2 and GPU 3
      2	GPU 6 and GPU 7

8. Immediately clean the PCM and putty pads off from the GPU with alcohol cleaning pads. Gently clean the PCM and putty pads to avoid GPU damages.
   Attention

   • It is recommended to clean the PCM while it is in liquid state.

   • The electrical components around the die on the GPUs are extremely delicate. When removing the PCM and cleaning the GPU die, avoid touching the electrical components to prevent damage.

   Figure 8. Cleaning PCM and putty pads off from the GPU
   
   
9. With alcohol cleaning pads, wipe off any remaining putty pad and PCMs from the GPU cold plate.
   Figure 9. Wiping PCM and putty pads off from the cold plate
   
   
10. Remove the GPU. Attach the two Torx T15 200mm extension bit to two torque screwdrivers. Simultaneously loosen the two diagonal Torx T15 screws with the screwdriver set to the proper torque.
    1. Set the torque screwdriver to 0.6 newton-meters, 5.3 inch-pounds to simultaneously loosen the two diagonal screws; then, simultaneously loosen the two diagonal screws.
       Figure 10. Removing the GPU
       
       
    2. Install the two GPU screw handles diagonally. Align the screw handles to the cold plate screw slots; then fasten the screw handles by hand.
       Figure 11. Installing the GPU screw handles
       
       
    3. Hold the GPU screw handles to remove the GPU from the GPU baseboard.
       Figure 12. Removing the GPU
       
       
    4. Carefully lay the GPU on a flat, static protective surface. Remove the two screw handles by loosening them by hand.
       Figure 13. Removing the GPU screw handles