Umamusume: Pretty Derby

Steam store page

Verdict

Runs

32 runs · 3 Macs · on dormison-r8

It starts and draws, but most sessions end early or crash.

Excluded from the verdict and frame rate: 13 runs stopped by a script through sevo.

Reached a frame
90%
Played 5 min or more
31%
Crashed
0%
First window after
12.7 s
Median frame rate
45 fpson a 120 Hz variable display · 49% of refreshthe game holds itself to 60 fps
Best renderer
DXMT

Reports

What players and the Sevoflurane team found on their own Macs: a verdict, the configuration the run was on, and a note. Runs speak for themselves above; a report is the sentence a run cannot say.

  1. Fails
    • dormison-r7
    • DXVK
    • lanczos
    • msync
    • macOS 27.0.1
    • M3

By Mac and renderer

Each figure is the median across Macs of each Mac’s own median. Frame rates count gameplay only, from runs with half a minute or more of it on a real display; the display column is the refresh rate most of those runs were played at, and the frame rate as a share of it.

MacRendererMacsRun countAvg fps1% lowp99 frameDisplayTypical windowEngines
M1 Max · 24-core GPUD3DMetal19301360.7 ms60 Hz4566×2632dormison-r8
M1 Max · 24-core GPUDXMT11————4566×2632dormison-r8
M3 · 10-core GPUD3DMetal18————2490×1464dormison-r7
M3 · 10-core GPUDXMT11604618.9 ms60 Hz2490×1464dormison-r7
M3 · 10-core GPUDXVK12—————dormison-r7
M3 Pro · 18-core GPUD3DMetal16————2192×1296dormison-r8
M5 Pro · 20-core GPUDXMT116592919.3 ms120 Hz variable2520×1482dormison-r9, dormison-r8
M5 Pro · 20-core GPUDXVK11—————dormison-r8
M5 Pro · 20-core GPUwined3d (OpenGL)11—————dormison-r8

Renderers

DXMT played best: the highest share of sessions that lasted, then the highest frame rate.

RendererMacsRun countPlayedAvg fps
D3DMetal3230%30
DXMT31816%60
DXVK230%—
wined3d (OpenGL)110%—

Engines measured

EngineVerdictMacsRun countReached a framePlayedCrashed
dormison-r9Not enough runs12100%0%0%
dormison-r8Runs33290%31%0%
dormison-r7Runs11191%0%73%

Last updated · newest run

How the data is collected

Every figure here was measured by Sevoflurane on a player’s Mac, after the player chose to share run statistics. Written reports are shown separately.

A shared run carries the game’s Steam id, the engine and renderer, the Mac’s chip, GPU core count and memory tier, the window size, the frame rate and how the session ended. It carries no name, no path and no serial number.

Each install signs its runs with a key kept in the Mac’s Secure Enclave, and Apple’s App Attest and DeviceCheck vouch that it is a real Mac. Each install counts once per game: the page shows the median of every Mac’s own median, so a thousand runs from one Mac weigh as much as one.

A frame rate counts gameplay only. The app starts counting ten seconds after the game’s first frame, once five seconds pass with no frame slower than 100 ms, and leaves out every stretch in which the game’s window was not the focused app, was hidden, or the display slept, and every frame slower than a quarter second. A run needs thirty seconds of such gameplay to carry a frame rate. Runs on a virtual display, where nothing paces the frames, carry none. Each frame rate is shown beside the refresh rate of the display it was played on.

A verdict needs three runs on the newest engine that the player or the game ended. Runs a script stopped through Sevoflurane’s command-line tool count toward neither the verdict nor the frame rate; a player pressing Stop does. Playable means most sessions lasted five minutes or more and ended when the player chose. Perfect adds a frame rate within a tenth of 60 fps, or of the display’s refresh rate when that is lower, or of the game’s own cap: when every Mac fast enough holds a game to one steady rate, such as 30 fps, that rate is its target.