Counter-Strike 2 on a Mac
Measured 13 September 2026 · MacBook Pro, M3 Pro
This is one machine, one map and one person's measurements. Everything below is what we saw ourselves; where we have not measured something, it says so. If you came here to find out whether CS2 is playable on an Apple Silicon Mac, the short answer is yes — with one real limitation you should read before spending anything.
The numbers
Dust II, a MacBook Pro with an M3 Pro, on mains power. Graphics settings and resolution left as the game set them — nothing lowered to make the figure look better.
| Graphics translation | Dust II |
|---|---|
| D3DMetal | about 100–120 fps |
| DXMT | never measured in a matched run |
These figures were taken with fps_max 120 in the autoexec — the upper bound here is that setting, not the machine. Uncapped we have not measured.
Both are usable. DXMT sometimes feels smoother in the hand — that is why it is in the settings — but what it costs in frames we do not know: we never ran a matched comparison on one scene, and an impression from ordinary play is not a measurement. Uncork lets you switch between them per bottle without reinstalling anything.
What the game's own log recorded
The figures above are what you see while playing. Below is what CS2 recorded about itself:
four VProf reports from a single evening, 11 September, pulled out of console.log.
We have not averaged them and have not picked the best one.
| Time | Average FPS | P1 FPS | Frame P99 | Frames excluded |
|---|---|---|---|---|
| 21:01 | 107.5 | 72.9 | 13.71 ms | 275 |
| 21:06 | 110.2 | 70.9 | 14.10 ms | 892 |
| 21:11 | 89.5 | 51.9 | 19.27 ms | 5335 |
| 21:32 | 117.6 | 85.0 | 11.77 ms | 39242 |
Three things are visible in that table and invisible in “roughly 100–120 frames”.
Smoothness lives in P1, not in the average. In the third row the average dropped by about a fifth, and P1 by nearly a third. Those are the frames on which the crosshair jerks, and the average barely shows it. That is why we spend our effort on mouse input rather than on a pretty headline number.
The upper bound is a setting, not the machine's ceiling. That evening the autoexec held
fps_max 120. Figures above 120 simply could not appear in these logs. What this
Mac would produce uncapped we do not know: we have not measured it.
These four rows cannot be compared with one another. The last report excluded 39,242 frames, the third 5,335. Why VProf excluded them, and on which scene, the log does not say. So “it got better by half past nine” is a conclusion the data does not support. The app's “CS2 smoothness” panel refuses, for the same reason, to show a difference between runs whose conditions were not recorded: better an empty space than a figure nobody can stand behind.
P1 as VProf writes it is not the same thing as “average of the worst one per cent”. They are different calculations, and we do not rename one into the other for the sake of a more familiar phrase.
The mouse, which is the part that matters
Frame rate is the number people quote. Aim is the thing they feel.
macOS does not hand an application every movement the mouse makes. It collects them and delivers a batch once per display frame. For a desktop that is invisible. For a shooter it means your aim advances in steps tied to the frame rate, which is exactly the "floaty" feeling people describe when they try to play Windows shooters on a Mac.
Uncork ships its own display driver that reads the mouse straight from the device instead — at the rate the mouse actually reports, up to 1000 Hz. This needs the Input Monitoring permission, which the app asks for before your first game. Only mouse movement is read, and only for games: keystrokes, passwords and other applications are not visible to it. Refuse, and everything still runs — the mouse simply stays as it was.
What does not work
Read this part before anything else.
- FACEIT, Vanguard and Easy Anti-Cheat do not run. They install kernel-level drivers, and there is no such thing here. If you play on FACEIT, this will not help you, and no amount of configuration changes that.
- VAC does run, so official matchmaking works. But Valve does not officially support playing this way, and we cannot speak for what they decide in the future. That residual risk is yours to take, on your own account.
We would rather lose a sale here than have you find this out after paying.
What you need
- A Mac with Apple Silicon — M1 or newer. Intel Macs are not supported.
- macOS 14 or later.
- Your own Steam account with CS2 on it.
- Disk space: here CS2 takes 72 GB — that is what Steam itself reports. The app asks you to keep 85 GB free: the game needs room for updates too.
How it goes
- Sign in on this site. Your profile shows a download link and an activation code.
- Install Uncork, enter the code, and it builds a bottle with Windows Steam inside.
- Sign in to Steam yourself and install CS2 as usual. Your Steam password and Steam Guard code are typed only into Steam's own window — Uncork never asks for them.
- Launch. The first run takes a while as shaders build; after that it starts normally.
What we have not measured
The honest limits of the figures above:
- A locked 120 fps on Mirage at unchanged settings is not established. We tried, and the evidence did not support the claim, so we are not making it.
- Other maps, other Macs and other resolutions are unmeasured. An M1 Air will not behave like an M3 Pro, and we have not tested one.
- Frame rate under a full server with smokes and utility is harder than an empty map. Our numbers come from ordinary play, not a benchmark loop, but they are still one person's play.
If any of this turns out to be wrong on your machine, write and say so — measurements that only hold in one place are worth correcting.