GPU Memory Allocation

The GPU Memory Allocation chart displays GPU memory allocation for the top 10 graphics devices on the host with the most GPU memory usage.

This chart is located in the GPUs view of the host Performance tab.

Note: The GPUs view is visible only if the host supports GPUs.

Table 1. Data Counters

Chart Labels	Description
Usage for device_id	The amount of GPU memory used in percentages of the total available. Counter: Memory usage Stats Type: Absolute Unit: Percent (%) Rollup Type: Average Collection Level: 4

Chart Analysis

GPU memory usage tracks how much of the GPU onboard memory (VRAM) is in use.

GPU details measurements can show usage trends related to virtual machine power on failures. Device and driver specific placement constraints of vGPU on physical GPU devices, including fragmentation of Device Memory, exhaustion of Device Memory, or exhaustion of Assignable Processor Engines could contribute to power on failures.

If GPU memory usage is almost at capacity and a virtual machine power on failure occurs with a Virtual GPU that would exceed capacity, increase the number of compatible GPU hosts, and migrate one or more virtual machines to the new host to free up GPU Memory resources for the new VM. For further guidance see Knowledge Base article KB 92222.

If a virtual machine is experiencing a Quality of Service (QoS) issue, please refer to the GPU vendor QoS documentation. If you use NVIDIA virtual GPU (vGPU) technology, see the NVIDIA Virtual GPU Software for VMware vSphere documentation for more information.