CPU usage exceeds 100% in vSphere Performance Charts due to Turbo Boost

Products

• VMware vSphere ESXi

Issue/Introduction

This article documents a change in behavior from 7.x to 8.0U3 in the UI where it displays the actual used% as opposed to capping the used% at 100. It explains why CPU utilization metrics for virtual machines and ESXi hosts may display values greater than 100% in vSphere 8.0 Update 3 and later environments.

Symptoms:

• Virtual machine CPU utilization alarms report values greater than 100%.
• vCenter alarms for CPU usage trigger frequently despite guest OS reporting lower usage.
• Performance graphs show usage over 100% instead of capping at the traditional 100% limit.

Environment

• VMware vSphere ESXi 8.0 Update 3 and higher
• VMware vCenter Server 8.0 Update 3 and higher

Cause

• Starting with vSphere 8.0 Update 3, the artificial 100% cap was removed on CPU metrics to reflect actual physical resource consumption more accurately.
• vSphere Client measures CPU usage at base frequency.
• It is possible for CPU usage and VM usage to be over 100%, due to turbo boost. When physical CPUs exceed their base clock speed (e.g., via Intel Turbo Boost), the reported frequency can exceed 100% of the rated baseline frequency.
• For VM usage, it can also go above 100% (assuming based on the total number of vCPUs) because of the CPU from IO threads. Due to this, it removes the unnecessary artificial limits on some of the counters ("cpu.usage" and "cpu.usagemhz" are based on "cpu.used", which effectively is already measured in TCs (timer cycles) on VMkernel level, which means that for higher core frequency we are going to receive more TCs (in microseconds) for the same time-interval).
• VMs CPU load graph can exceed 100% which was previously limited to 100%. When VMs CPU usage exceeds 100%, performance graphs no longer show CPU usage in percentage - vCenter UI > VM performance in vCenter > VM > Monitor > Performance.

Resolution

• As of vSphere 8.0 U3, the percentage value should adjust to match the maximum usage over 100% in the graph. 
• You may also use the virtual machine > monitor > performance > advanced to see the correct CPU usage of the virtual machine.
  This graph includes the 16 cores and not 8 vCPU's.
  
  
  
  
  
• Since performance charts now accurately reflect CPU usage exceeding 100% (due to features like Turbo Boost), administrators should review and adjust any existing CPU utilization alarm thresholds set on VMs or ESXi hosts. If a VM is consistently running above 100% but operating normally,  thresholds should be raised accordingly (Consider thresholds above 100% (for example 120 – 150%) or alerts based on sustained usage over time.) to prevent "false alarms" caused by the new, higher metric reporting.
  1. Navigate to the VM or Host or Cluster or vCenter > Configure > Alarm Definitions.
  2. Edit the CPU Usage alarm.
  3. Increase the threshold values (e.g., Warning at 120%, Critical at 150%) to account for Turbo Boost spikes.
• To verify if Turbo Boost is active on an ESXi host, run the following command via SSH:
  - esxcfg-info -a | grep -i turbo

Additional Information

An example of the Alarm Triggered in the vCenter with the CPU Utilization 

---

Some users have reported that the MHz on the right of the Overview graph might incorrectly show the same number for the Percentage (%) on the left side as displayed on the right side for MHz.