← Back

How to check your Mac's performance

Four numbers tell you almost everything: CPU activity, memory pressure, free storage, and thermal state. Activity Monitor shows you three of them, and the fourth, the one that explains most cases of "it was fast last month", it does not show at all. The bigger problem is that the headline number everybody reads, %CPU, is not a measure of performance. Here is what each number really is, how to read it on Apple Silicon, and how to test the machine in a way that separates a slow Mac from a hot one.

The four numbers

NumberWhere to read itWhat a bad reading means
CPU activityActivity Monitor, CPU tabSomething is busy, not necessarily something wrong
Memory pressureActivity Monitor, Memory tabYellow means compressing, red means swapping
Free storageSystem Settings, General, StorageNear-full disks slow everything down
Thermal stateTerminal, or a menu-bar readerThe machine is trading speed for heat

If you have not opened Activity Monitor before, it lives in Applications, Utilities, and the Task Manager equivalent on a Mac covers getting to it and what each tab does. This page is about reading it correctly once you are there.

Why %CPU is not a performance number

The %CPU column measures active residency: the proportion of the sampling window during which a core was not idle, added up across cores. That is why one process can read 400% on an eight-core machine, and why the scale tops out at 100 times your core count rather than at 100.

Residency says nothing about how much work happened in that time. Howard Oakley at The Eclectic Light Company measured this directly on Apple Silicon: efficiency cores running a low quality-of-service thread sat at 744 MHz, against 2,748 MHz for the same cores at a high QoS, a 3.7-fold gap. Two tasks each showing 200% finished in 10.2 seconds and 3.0 seconds respectively. His conclusion is the one to carry around: what Activity Monitor shows as 100% can be anywhere between 27% and 100% of the real capacity, depending on which cores took the work and at what frequency.

The practical consequence is that a high percentage is a starting point, not a verdict. A background indexer pinned to the efficiency cores can look alarming and cost you nothing. A modest-looking percentage on the performance cores can be the thing heating your Mac. For the per-process view and how to find the offender, see how to check CPU usage on a Mac.

See which cores are doing the work

Activity Monitor's CPU History window draws a graph per core, and on Apple Silicon the efficiency cores are the group that stays busy while the machine feels idle. To know which is which, ask the system directly in Terminal:

sysctl hw.perflevel0.physicalcpu hw.perflevel1.physicalcpu

Perflevel 0 is the performance cluster, perflevel 1 the efficiency one. The older hw.ncpu just returns the total, which is the number About This Mac shows you and the one that tells you least. Knowing the split turns the CPU History graph from decoration into information: eight bars busy on the efficiency side is a completely different situation from eight bars busy on the performance side.

Memory pressure, not free memory

macOS is designed to use all of your RAM, so "free memory" is close to meaningless as a health metric. The graph that matters is Memory Pressure at the foot of the Memory tab. Green means the system is handling demand normally. Yellow means it has started compressing memory to cope, which you may feel slightly. Red means it is leaning on swap, writing memory out to disk, and that is the state where a Mac genuinely feels slow rather than busy.

Judge it over time rather than at a glance. A spike into yellow while a big export runs is the system working as intended. Sustained red during ordinary work is the honest signal that the machine wants more RAM, which on Apple Silicon means a different machine rather than an upgrade.

The number Activity Monitor will not give you

Nothing in Activity Monitor reports temperature or thermal state, which is a problem, because on a modern Mac that is often the actual answer. Apple Silicon runs hot by design and protects itself by cutting clock speed as it approaches its limits, so a Mac that benchmarks fine cold can lose a chunk of its speed twenty minutes into a video export while every number on screen still looks reasonable.

The built-in way to see it is Terminal:

sudo powermetrics -s thermal

That reports thermal pressure, where Nominal means the system is not constraining performance and anything higher means it is. Note that the smc sampler quoted in a lot of older guides is an Intel-era thing and is not available on Apple Silicon, so copying those commands gets you an error rather than a reading. If you would rather see a live number than run a command, how to check your Mac's temperature covers the menu-bar route, and what thermal throttling is explains what the machine is doing when it happens.

One caution from building a tool that reads these sensors. A Mac exposes many temperature keys, and several of the non-die ones read roughly 15 degrees Celsius above the actual CPU and GPU die sensors, because they are aggregate or modeled hotspot values rather than measurements of the chip. An app that grabs the highest key it can find will happily show you a frightening number that is not your die temperature. If a reading alarms you, check what sensor it came from before you act on it.

Benchmark twice, not once

A benchmark answers one question well: is this machine performing like others of its model. Geekbench or Cinebench, run once on a cool Mac, gives you that peak figure to compare. What it does not tell you is anything about the ten minutes after, which is where most real complaints live.

So run it again, in a loop, for ten minutes or so, and watch the score across passes. A machine holding its number is fine. A machine whose second and third passes drop away is thermally limited, and that is a cooling story rather than a performance story. A fanless MacBook Air will do this by design; a MacBook Pro doing it in an air-conditioned room is worth investigating. The comparison is spelled out in why the fanless Air throttles, and if the machine is both slow and hot, the slow and hot loop is the page you want.

Watch out for the monitor itself

This one I learned the hard way, building ChillBlades. Its window shows live temperatures, and idle die temperatures wobble by a degree or so every polling tick. With SwiftUI animations keyed to those values, the animation was effectively running non-stop, and the app cost between 12 and 32 percent CPU with its window open. With the animation removed, the same window sat between 0.8 and 1.9 percent. The two-second sensor poll was never the expensive part.

The general lesson applies to whatever you use to check performance. A monitoring app with a live graph, a menu-bar readout updating every second, a dashboard left open on a second display: each of these is a process, and one written carelessly can consume more than the thing you are investigating. If you are chasing a mystery load, close the monitoring tools you are not reading and see what the number does.

About this guide

I make a Mac fan control app, so I spend more time than is healthy reading these numbers. The CPU frequency figures above are Howard Oakley's published measurements rather than mine, and I have named them as such because that testing is his. The sensor caution and the animation cost are first-hand from my own app's source and profiling. Where I have held back is on absolute temperature and RPM figures for your machine, because they vary by chip and chassis, and a number that is true on my Mac is not a number you should trust on yours.

FAQ

How do I run a performance test on my Mac?
Run a benchmark twice: once cold, then in a loop for ten minutes. Geekbench or Cinebench will give you a score in a couple of minutes, and that score tells you the peak your Mac can hit. Looping the same test is what tells you what it can hold, because sustained load is where a thin or fanless Mac drops clock speed to control heat. A single run that matches other machines of the same model plus a big fall on the second and third pass is the classic thermal signature, not a broken Mac.
What does %CPU in Activity Monitor actually mean?
It is active residency: the share of the sampling window in which a core was not idle, summed across cores. That is why a single process can show 400% on an eight-core Mac. It is not a measure of work done. On Apple Silicon a thread parked on efficiency cores at a low quality-of-service level can sit at a high percentage while running at a fraction of the clock speed the performance cores would use, so the same percentage can represent very different amounts of computing.
How do I check how many cores my Mac has?
In Terminal, run sysctl hw.perflevel0.physicalcpu hw.perflevel1.physicalcpu. The first is your performance core count, the second your efficiency core count. On Apple Silicon the older hw.ncpu just adds them together, which hides the split that matters. About This Mac gives you a combined figure too, so the perflevel keys are the quickest way to see the real shape of your chip.
How can I tell if my Mac is thermally throttled?
Run sudo powermetrics -s thermal in Terminal while the machine is under load and watch the thermal pressure level. Nominal means the system is not constraining anything; anything above it means macOS is managing heat. The smc sampler people quote from Intel-era guides is not available on Apple Silicon, so -s thermal is the route on a modern Mac. A clock speed that falls while the workload stays constant is the same story from the other direction.

Try it

ChillBlades is a one-time $25 purchase with a seven-day free trial, no account and no card up front. It runs entirely on your Mac.

Get ChillBlades