Performance metrics are numbers that measure different aspects of the working state of a cpu and reveal useful diagnostic information. These include cpu usage, clock speed, power consumption, temperature etc.
There are many more but for simplicity we shall be focusing on these ones in this article. The metrics can be monitored using tools like Hwinfo, Open hardware monitor, core temp etc on windows.
In this article we shall be discussing about 4 key cpu metrics which are:
- CPU Usage (%)
- CPU Clock Speed/Frequency (Mhz)
- Power Consumption (watts)
- Temperature (C)
At any given moment, a cpu core is technically either executing an instruction (100% usage) or not executing any instruction (0%). Then how do we get these usage indicators in the form of a percentage number like 35% or 95%. The answer lies in how the usage is measured.
A cpu has a clock that runs at a certain speed, for example 1000 hz which means there are 1000 clock cycles each second. Now the cpu can execute an instruction at each clock cycle (this is an over-simplification but bear for now).
So if the cpu was executing something in 300 out of the 1000 clock cycles in a particular span of 1 second, and not executing anything in the other 700 clock cycles, we can say that the usage was about 30%. Such a sample can be taken over any arbitrary period of time, like 1 second or 1 minute.
Further more, if there are multiple cores in the cpu, then the average of usage of all cpus can be taken combined and a final number is calculated.
To check your cpu usage, open the Task Manager in Windows 10 or the System Monitor on Ubuntu. They would show the usage per cpu core. In case of hyperthreading, it would show twice the number of cores each with its own usage history.
Its worth mentioning that on linux systems there is another unique metric for measuring cpu load and the command is:
$ cat /proc/loadavg 1.08 1.20 0.85 2/1977 59113
The loadavg metrics shows the number of processes/threads that are running or waiting. This metric is very different from the cpu usage and there is no direct correlation between the 2. For instance a system might have very low loadavg values but still high cpu usage and vice-versa.
When it comes to cpu usage, its always a good idea to monitor the per-process cpu usage using tools like Task manager. This will report the load of each application. Sometimes, some applications might get in a bad state and start killing the cpu by consuming all of it. Then you know its time to kill that process from the task manager.
There might be bloat-ware running in the background that is constantly consuming cpu. You can discover such un-needed applications and get rid of them, to keep your cpu less busier.
2. Clock speed/frequency
Clock speed of frequency is the number of times the clock ticks each 1 second. For instance 2.1 Ghz. This means that the clock ticks around 2100 million times each second and with each tick, the cpu can execute some instruction.
Turbo Boost: In most modern cpus the clock speed is not always constant. There is a base value, called the base clock which is the frequency maintained all the time (even when the cpu is doing nothing). If there is heavy load of instructions from multiple programs, the cpu can increase its clock frequency (called boost clock) to execute more instructions per second.
Turbo boost has become a pretty standard feature in almost all desktop and laptop cpus. Furthermore cpus are configurable and the turbo boost can be disabled to keep the cpu cooler.
Overclocking: Besides boost, another way the cpu clock frequency can be increased is through overclocking. Overclocking allows to increase the maximum boost levels of the cpu, provided the target system can actually handle the outcome of higher clock frequency which is higher power consumption and consequently higher heat output.
Sometimes, if the system is not able to dissipate the cpu's heat then the cpu will throttle and down-clock its frequency to lower the thermals. This is a frequent phenomenon in laptops which have limited cooling capacity.
Higher frequency also needs more power, so if the system is running on low power, it might keep the cpu clock down to save power. Laptops for instance keep cpu frequencies low when running on battery. They do this by disabling boost clocks and lowering the minimum frequencies below the base levels.
Desktop pcs on the other hand have no such mechanism of know the maximum power available, and if the psu is not able to keep up with the power needs of the cpu or any other component like the gpu, the system will simply "crash" and reboot. This is seen when the psu is low capacity or has worn out over time and can't sustain higher power output.
Details about base and boost clock frequencies of any particular cpu model can be found on the official page on manufacturers' website (whether intel or amd). The official page would also tell whether the cpu supports overclocking or not.
Here is a small table comparing the clock speeds of some current and old cpus:
|Base Clock (Ghz)
|Boost Clock (Ghz)
|Ryzen 7 5800H
|Ryzen 5 5500U
|Ryzen 9 5900X
Some cpus like the ryzen "U" series are designed for low power consumption devices like ultrabooks and light-thin notebooks. They have much lower base frequencies and lower boost frequencies compared to regular cpus in the same line.
To monitor the cpu clock frequencies on Ubuntu check the commands discussed on this page
3. Power consumption
Power consumption (measured in simple watts) tells the amount of power that is being consumed by the cpu at any given time-span.
As we discussed above that in each clock cycle, the cpu is able to execute an instruction and each execution consumes different amount of power. Sometimes you would see 100% cpu usage consuming less power compared to other times. This depends on what kind of instruction or workload is being executed on the cpu.
Some instructions consume more power than others.
The power consumption also rises with the number of cores on a cpu. A cpu with more cores will consume more power compared to one with fewer cores built using similar technologies.
Other factors that affect the power consumption include number and size of the transistors. Modern cpus are fabricated with smaller transistors compared to previous generations and hence consume less power and generate less heat, but at the same time the number of transistors has increased which need more power but also deliver higher performance.
So the actual power requirements of cpus has been practically rising over time.
When it comes to power consumption there are 2 important hardware factors that come into play. The first is the power supply which must output enough wattage to the cpu and the second is the cooling capacity which should keep the cpu cool enough to prevent thermal throttling.
You can tell that your cpu is consuming more power simply by listening to the noise generated by the fan which will be loud as the fan spins faster to cool the hotter cpu.
High fan noise is also an easy to pickup alert that some application is stressing your cpu. Incase you did not intend to run anything, its a good idea to quickly check the running processes and see whats going on.
Both Intel and AMD provide information about the power rating of their cpus in a unit called TDP, which gives us an indication of how much power a cpu would consume to deliver maximum peformance. This metric is useful in choosing the right power supply unit (psu) for a desktp build for instance and also choosing the right kind of after market cooler.
A cpu with higher tdp ratings needs a more powerful psu to drive it and a more powerful cooler to cool it!
Here is a small comparison of some cpus and their tdp values:
|Base Clock (Ghz)
|Boost Clock (Ghz)
|Ryzen 7 5800H
|Ryzen 5 5500U
|Ryzen 9 5900X
Laptop cpus usually have much lower tdp values.
Monitor Power usage
On linux you can use command line programs like powertop, powerstat and s-tui to monitor cpu usage/frequency/power consumption.
The temperature of the cpu is a direct consequence of the power being consumed, and the cooling provided. With good cooling a cpu can maintain low temperatures even with high power consumption which is ideal.
On gaming laptops the cooling is often limited and inadequate, leading to thermal throttling. This process downclocks the cpu so that it consumes less power and consequently generates less heat.
On windows, the cpu temps can be monitored using tools like Open Hardware Monitor, HwInfo and Coretemp. The tools would report the temperature of individual cores/threads as well as the total package.
Gaming laptops also have tuning utilities that report system performance metrics related to cpu, ram and gpu. For instance the Asus gaming laptops have Armoury Crate.
On linux there are commands like lm-sensors, s-tui, psensor that can be used to monitor the temperature.
I remember seeing temperatures close to 100 C under heavy workloads on my earlier desktop pc that ran on the older intel Q8400 cpu. The temps revealed that the thermal paste on the cpu had dried completely and it was time to re-apply fresh paste. After applying new thermal paste the temperatures dropped by over 20 C.
High temperatures on laptops is a common phenomenon, but too high temperatures could indicate some problem with the placement of the device resulting in poor cooling.
For example if you put a laptop on a couch or bed, the vents on the bottom side can get badly blocked and the whole laptop would heat up within few minutes. I remember doing this silly mistake with my acer swift ultrabook once.
If you are seeing high temperatures on laptops then try tweaking the system parameters to make it run cooler because little can be changed in the hardware. With desktop pcs, if you are seeing high temperatures on modern cpus, then get a better third party cooler with higher cooling capacity.
So that was a quick and dirty introduction to some of the basic concepts about cpu metrics. You can right away download free tools to monitor your cpu and see how things work.
Some free tools include HwInfo and Open Hardware Monitor that provide extensive information about various system components and not just the cpu.
Monitoring cpu performance can sometimes provide critical insight into system health. For example if the cpu is running with high temperatures even under moderate work loads, it could indicate poor or failed cooling which needs immediate replacement to avoid damage.
If the cpu usage is constantly high without any intended application, it could indicate some poor application that needs to be stopped right away.