A useful performance number can still hide an uncomfortable moment. Here’s how to read it without asking it to answer the wrong question.
A game runs at 40 frames per second on average. Would you buy it for your Steam Deck?
The honest answer is: That number alone isn’t enough. It could describe a consistently paced experience or a session where quiet corridors run quickly and busy fights struggle. Both can produce a respectable headline figure. What matters is whether performance holds up when you actually need to move, aim, or react.
Start by Asking Where the Number Came From
Valve’s Steam Deck Framerate announcement describes averages based on anonymised, opt-in player data, accompanied by a histogram. These are community samples, not one reviewer’s controlled test. Valve filters for default device frame limits but does not standardise custom in-game graphics settings.
That makes the feature useful for an initial check, not a promise about your save file. Our GameCast guide to finding Steam Deck FPS data and reading the histogram covers where the figures appear on Steam. The next question is what they leave out.
Frame Times Explain What Averages Flatten
Think of frames as deliveries. Knowing how many arrived during a minute doesn’t tell you whether they arrived evenly or in bursts.
At a perfectly even 60 FPS, a new frame arrives about every 16.7 milliseconds. At 40 FPS, that interval is 25 milliseconds; at 30 FPS, roughly 33.3 milliseconds. These are mathematical examples, not measurements of a particular game.
A frame-time trace adds the missing sequence. A sudden long interval can expose a hitch that a session-wide average barely moves. It also shows when the problem happened: during combat, a camera turn, or a transition. The graph alone does not prove the cause. Calling every spike “shader compilation” would be another unsupported shortcut.
A Histogram Is Not a Recording of Your Session
A distribution and a timeline answer different questions. A histogram groups observations into ranges; a timeline preserves their order.
That distinction matters even before comparing settings. A distribution cannot tell you whether slower observations occurred together in one difficult encounter or were scattered across many sessions. Nor does a small bar automatically predict a pause on your own Deck. Use it to decide what to investigate rather than turning it into a verdict.
Read 1% Lows with the Methodology Attached
Technical reviews often add a “1% low” figure to describe the slower end of a capture. NVIDIA’s FrameView documentation distinguishes its slowest-one-percent average from a percentile threshold. Those are not interchangeable statistics.
Check which tool and calculation a reviewer used before comparing two charts. A 1% low is not the single worst frame, and it does not explain every hitch. Capture length and the section being played also matter. This is a point about reading benchmark reports, not a recommendation to install the Windows-based FrameView on SteamOS.
Before buying, look for a useful comparison
Ask five questions:
- Does the footage show your Deck model and a recent game version?
- Are resolution, graphics settings, and upscaling stated?
- Does the test include demanding gameplay, not just the opening area?
- Are frame-time behaviour and slower moments visible?
- Does the chosen frame-rate target suit the way you want to play?
When testing your own copy, repeat the same route and change one setting at a time. Keep notes on the settings and game version so a comparison means something. Include the parts you expect to spend time in, not only the scene that produces the biggest number.
The goal isn’t to reject average FPS, it is to stop treating an average as a guarantee. Use it to shortlist a game, then look for evidence that the experience stays comfortable when the game gets demanding.
Prepared by the GameCast editorial team. This feature uses published technical documentation and explanatory examples; no original performance measurements are claimed.
