--- title: "GPUs View" slug: "gpus-view" updated: 2026-09-14T13:16:46Z published: 2026-09-14T13:16:46Z canonical: "support.controlup.com/gpus-view" --- > ## Documentation Index > Fetch the complete documentation index at: https://support.controlup.com/llms.txt > Use this file to discover all available pages before exploring further. # GPUs View > [!NOTE] > Feature available only in version 9.2.5 > > This article describes monitor features only available in ControlUp Real-Time DX version **9.2.5**. The **GPUs** view provides comprehensive, host-level monitoring for physical NVIDIA GPUs. This view in the Real-Time DX Console and the VDI web UI gives you deep visibility into hardware performance, resource consumption, and environmental metrics, helping you right-size infrastructure capacity and control costs. GPU metrics are displayed across the real-time topology, appearing in the dedicated **GPUs** view widgets and data grid, as well as in new GPU metrics columns on the existing [**Hosts**](/v1/docs/hosts-view-actions) and [**Machines**](/v1/docs/machine-view-actions) views grids. You can focus on a specific host or machine in the topology view (Details) or organization tree, and drill down directly to individual user sessions utilizing GPU resources. ![VDI details GPUs view element selected showing GPUs widgets and GPUs metrics](https://cdn.document360.io/098100b7-b9da-4bea-b4b9-017140ab863e/Images/Documentation/VDIGPUs.png) You can also use GPUs to configure Stress, Scheduled, and Advanced [triggers](/v1/docs/setting-up-triggers), as well as [Script-based Actions](/v1/docs/script-based-actions-sba), to help your organization get ahead of rising GPU costs. For more details about GPU monitoring, see [here](/v1/docs/gpu-monitoring-with-controlup). > [!NOTE] > Note > > Because NVIDIA doesn’t expose GPU metrics through standard VMware APIs, host-level data collection for this feature requires direct SSH access to your ESXi hosts. ## Prerequisites Because NVIDIA doesn't expose GPU metrics through standard VMware APIs, host-level GPU data collection requires configuring SSH access to your ESXi hosts directly from the Hypervisor Connection settings in ControlUp. To enable GPU data collection for a vSphere Hypervisor Connection: 1. In the Real-Time DX Console, open the **Hypervisor Connection** settings for your vCenter connection. 2. Under **GPU Data Collection**, select **Enable**. 3. Click **Configure ESXi Hosts...** 4. In the **GPU Data Collection Settings** window, set the **Default SSH Credentials** and **Default SSH Port** to be used across your ESXi hosts, or configure credentials per host in the table below. 5. For each host listed, enable GPU data collection individually, or click **Enable All** to apply it to every host. 6. Click **OK** to save the settings. This covers host-level GPU data collection via vSphere. If you also want to collect GPU metrics at the machine, session, or process level, see the registry-based prerequisites [here](/v1/docs/gpu-monitoring-with-controlup#system-prerequisites). ## GPU Triggers and Actions GPUs are fully integrated targets within ControlUp for VDI's automation engine, empowering you to move from passive monitoring to proactive management. ### Stress Settings & Alerting You can configure customized stress level thresholds for GPU metrics like utilization, memory depletion, or high temperatures. When a host's GPU exceeds these health parameters, its color-coded stress state shifts in real-time, alerting you to potential bottlenecks before end-user experience degrades. ### Advanced & Scheduled Triggers Set up automated workflows using Advanced Triggers. For example, to manage cost control: Schedule reports or trigger alerts on underutilized GPU assets to optimize allocation across your host clusters. ### Script-Based Actions (SBAs) Contextual SBAs can be executed directly against objects in the **GPUs** view. You can run diagnostics, gather deeper NVIDIA SMI log telemetry, or reset stuck display sub-systems right from the grid without opening a separate host management shell. ## GPUs View Column Reference The following columns are available in the Information Grid while using the **GPUs** view. | **Column** | **Description** | | --- | --- | | Name | Name/model of the GPU | | Hostname | The name of the host which hosts GPUs | | Driver version | Installed NVIDIA driver version | | Timestamp | The timestamp of when the query was made | | GPU Utilization | Percentage of GPU core usage | | Memory Utilization | Percentage of memory bandwidth usage | | Current temp | Current temperature of the GPU | | Power Draw | Real-time power consumption | | P-State | Power state (performance level), e.g., P0 = max performance, P8 = idle | | User sessions | Total number of user sessions on virtual machines using this GPU | | Active vGPUs | Number of active virtual GPUs (vGPUs) currently running on the physical GPU | | Bus Id | PCI bus ID of the GPU | | Clocks - graphics | Current graphics clock speed | | Clocks - SM | Current Streaming Multiprocessor clock speed | | Clocks - Video | Current video encoder/decoder clock speed | | Clocks Memory | Current memory clock speed | | Free memory | Free (unused) GPU memory | | Index | GPU identifier on the host (e.g., 0, 1, 2, 3) | | Installed memory | Total GPU memory available | | Max operating temp | Maximum safe operating temperature defined by the GPU manufacturer | | Power limit | Maximum power draw allowed | | Scheduler policy | vGPU scheduler policy configured on the physical GPU (e.g., Equal Share, Best Effort) | | Shutdown temp | The critical temperature at which the GPU will automatically shut down to prevent hardware damage | | Slowdown temp | The temperature at which the GPU begins to throttle performance to prevent overheating | | Stress Level | Stress Level is a flexible load measure which you can customize using the "Stress Settings" pane | | Target Temperature | Target temperature the GPU attempts to maintain under load by adjusting cooling mechanisms such as fan speed | | Used memory | Currently used GPU memory | | UUID | Universally unique identifier of the GPU |