How do I view GPU monitoring data?

Dashboards

Dashboards:

You can view GPU data from the following dashboards. 

Summary dashboards:

GPUs:

Note, these dashboards are only shown for a server if the server has the GPU Connection Method field configured on the Add Asset form during asset management.

• GPUs - this is a Table dashlet. This dashlet lists and details the GPUs in the selected server. Click a GPU to navigate to its Summary dashboard. 
• Fabric Switches - this is a Table dashlet. This dashlet lists and details the fabric switches of the GPU that you select in the GPUs dashlet. Click a fabric switch to navigate to its Summary dashboard. 
• Utilization (selected GPU) - this is a Chart dashlet. This dashlet displays the current and historical GPU utilization in the selected server for the GPU that you select in the GPUs dashlet, shown as a percentage of total available GPU, memory, and power.
• GPU Temperature (selected GPU) - this is a Chart dashlet. This dashlet displays the current and historical temperature of the GPU and GPU memory in the selected server for the GPU that you select in the GPUs dashlet, shown in celsius. 
• GPU Memory (selected GPU) - this is a Chart dashlet. This dashlet displays the total available memory, current amount of memory used, and the amount of memory free for use on the GPU that you select in the GPUs dashlet, in gigabytes.
• GPU Power (selected GPU) - this is a Chart dashlet. This dashlet displays the total amount of power used by, and the power limit for, the GPU that you select in the GPUs dashlet, in watts.
• GPU Clock Frequencies (selected GPU) - this is a Chart dashlet. This dashlet displays the current and historical operating frequency of the GPU card's processor core, in megahertz, for the GPU you select in the GPUs dashlet.
• GPU PCIe Throughput (selected GPU) - this is a Chart dashlet. This dashlet displays the current and historical rate at which data moves between the GPU and the rest of server over the PCI Express (PCIe) slot of the GPU that you select in the GPUs dashlet.

GPU device:

• Key Info - this is a Key Info dashlet. This dashlet displays information about, and the status of, the selected GPU.
• Incident Summary - this is an Incidents Summary dashlet. This dashlet displays a pie chart of the incidents by severity level on the selected GPU.
• Events Over Time - this is a Chart dashlet. This dashlet displays a history of the events on the selected GPU.
• GPU Links - this is a Table dashlet. This dashlet lists and details the GPU links on the GPU. Click a GPU link to navigate to its Summary dashboard. 
• GPU Utilization - this is a Chart dashlet. This dashlet displays the current and historical GPU utilization on the selected server, shown as a percentage of total available GPU (0-100%).

• Temperature - this is a Chart dashlet. This dashlet displays the current and historical temperature of the selected GPU.

   

GPU fabric switch:

• Key Info - this is a Key Info dashlet. This dashlet displays information about, and the status of, the selected fabric switch of the GPU.
• Incident Summary - this is an Incidents Summary dashlet. This dashlet displays a pie chart of the incidents by severity level on the selected fabric switch of the GPU.
• Events Over Time - this is a Chart dashlet. This dashlet displays a history of the events on the selected fabric switch of the GPU.
• Temperature (C) - this is a Chart dashlet. This dashlet displays a history of the events on the selected fabric switch of the GPU.
• Bandwidth - this is a Chart dashlet. This dashlet displays the current and historical rate at which data can move across the selected fabric switch's internal backplane or switching engine, shown in kilobytes per second (TX/RX).

GPU link:

• Key Info - this is a Key Info dashlet. This dashlet displays information about, and the status of, the selected GPU link.
• Incident Summary - this is an Incidents Summary dashlet. This dashlet displays a pie chart of the incidents by severity level on the selected GPU link.
• Events Over Time - this is a Chart dashlet. This dashlet displays a history of the events on the selected GPU link.
• Link Status - this is a Chart dashlet. This dashlet displays the real-time speed, lane width, and operational state of the connection between the GPU and the motherboard (with PCIe) or between multiple GPUs for the selected GPU link, shown as up or down.
• Bandwidth - this is a Chart dashlet. This dashlet displays the current and historical rate at which data can be transferred directly between the GPU and another device over the selected GPU link, shown in gigabytes per second.

GPU instance:

• Key Info - this is a Key Info dashlet. This dashlet displays information about, and the status of, the selected GPU instance.
• Incident Summary - this is an Incidents Summary dashlet. This dashlet displays a pie chart of the incidents by severity level on the selected GPU instance.
• Events Over Time - this is a Chart dashlet. This dashlet displays a history of the events on the selected GPU instance.
• Compute Instances - this is a Table dashlet. This dashlet lists and details the GPU compute instances on the GPU instance. Click a GPU compute instance to navigate to its Summary dashboard. 
• Framebuffer Memory - this is a Chart dashlet. This dashlet displays the total, free, and used dedicated VRAM (video random-access memory) that's allocated to the selected GPU instance, shown in gigabytes.
• Engine Activity (%) - this is a Chart dashlet. This dashlet displays the percentage of time the graphics and compute engines on the selected GPU instance are actively processing instructions.

GPU compute instance:

• Key Info - this is a Key Info dashlet. This dashlet displays information about, and the status of, the selected GPU compute instance.
• Incident Summary - this is an Incidents Summary dashlet. This dashlet displays a pie chart of the incidents by severity level on the selected GPU compute instance.
• Events Over Time - this is a Chart dashlet. This dashlet displays a history of the events on the selected GPU compute instance.
• Engine Activity (%) - this is a Chart dashlet. This dashlet displays the percentage of time the graphics and compute engines on the selected GPU compute instance are actively processing instructions.