Shadow of the Tomb Raider
Shadow of the Tomb Raider is a benchmark you can actually check. Its built-in run walks three fixed scenes and writes the numbers to disk itself - a per-frame log with a time and delta for every frame, plus a summary naming the average, the frame count and the full settings block it ran with. You are not reading a figure off someone's screenshot, you are re-deriving it from the game's own file.
That is why it leads our Linux numbers. It is a Feral Interactive native port, so there is no Proton translation layer between the benchmark and the hardware - what you see is the machine, not the compatibility shim. The run is long enough to move through genuinely different loads: an open jungle stretch, a dense market hub and a night-time set piece, rather than one flattering corridor.
The catch worth knowing: the game's own summary reports a Min FPS of 0.0, because the level load sits inside the captured window as a single enormous frame. The average is sound. Treat that minimum as an artifact, not a stutter.
Recommended settings
- Resolution
- Not specified
- Preset
- Not specified
- Upscaling
- Not specified
- Ray tracing
- Not specified
- Built-in benchmark
- Yes
- Scene
- Built-in benchmark, all three scenes run end to end (Day of the Dead, Lost in the Jungle, Paititi hub) — ~157s total.
CPU sensitivity
0.3 - 0 is GPU-bound, 1 is CPU-bound.
DSOGaming's Shadow of the Tomb Raider PC performance analysis is the CPU-scaling source I retrieved and read for this title. Testing simulated core configurations at 720p on Max settings, it measured a dual-core without Hyper Threading at 22 fps minimum and 39 fps average, rising to 31 fps minimum and 54 fps average once Hyper Threading was enabled - so thread count matters sharply below four threads. In the same run its GTX 980 Ti was maxed out even at 1280x720, and the analysis frames the GPU as the limiter at playable settings. Two limits on that evidence: it simulates core counts rather than comparing real CPUs, and it is 2018-era hardware. So 0.3 here is a deliberately conservative low-sensitivity value - give the game four or more threads and it goes GPU-bound, which is what our own 1440p run on the Steam Machine shows. Revisit this if a modern multi-CPU scaling test for the title turns up.
Benchmark results
| Build / device | Preset | Upscaling | fps |
|---|---|---|---|
| Valve Steam Machine | highest | Native | 80.96 fps Verified built-in benchmark |
How we measured this
Native 1080p. Upscaling, VSync, dynamic resolution and ray tracing off. Game Mode, frame limiter off, power profile at maximum. We report the median of 5 passes. | |||
Suggest a fix. Know a correction? Sign in to submit one - it’s credited to your handle when we publish it.
Spot an error on this page? Report a correction.