Problems · Under load
How much do soft shadows cost in frame rate?
Often more than any other single setting. Realistic soft shadows sample the shadow map dozens of times for every pixel, every frame, to decide how blurred each edge should be. On a mid-range or integrated GPU that can take a large share of the frame rate; on a fast dedicated card the cost is smaller but still real.
The only honest way to know what they cost on your card is to measure the same scene with soft shadows and with hard shadows. That comparison is exactly what the soft shadows test does.
First, notice when it happens
The causes, ranked
1. Soft shadow filtering is sampling heavily per pixel
most commonHow to confirm it
Frame rate rises clearly when shadow quality drops from high or ultra to medium.
The fix
Use medium or high shadows. The visual difference from ultra is usually small; the frame rate difference often is not.
2. Shadow map resolution is set very high
commonHow to confirm it
Memory use and frame rate both improve when shadow resolution is reduced separately from shadow softness.
The fix
Lower shadow resolution one step. On cards with limited video memory this also prevents stutter.
3. Ray-traced shadows are enabled
commonHow to confirm it
The game offers ray-traced shadows and they are on.
The fix
Turn them off or to their lowest quality, or pair them with upscaling. They are among the most expensive ray-tracing effects.
4. The card is heat-limited, and shadows just add load
occasionalHow to confirm it
Performance fades over time even with shadows reduced.
The fix
Address cooling first. Settings cannot fix a card that is throttling.
Measure it instead of guessing
The SystemCheck soft shadows test renders one scene at a fixed 1600 by 900 with percentage-closer soft shadows (16 blocker samples and 32 filter samples per pixel) and with a single hard-shadow sample, alternating every ten seconds. It reports the soft-shadow frame rate and the cost as the share of frame rate soft shadows take away. It runs standalone on its own page and as a stage of the Deep run.
What software cannot tell you here
- What a specific game's shadow setting costs. Games use different techniques, so the test measures the principle on your card, not a particular title.
- Why a game stutters. Shadow cost shows as lower frame rate, not as irregular frames.
Asked, in people’s own words
- Should I turn off shadows for more FPS?
- Rarely all the way. Dropping shadow quality from ultra to medium usually recovers most of the frame rate while keeping shadows that anchor objects to the scene, which matters for how a game reads.
- Why are soft shadows more expensive than hard shadows?
- A hard shadow needs one lookup per pixel: lit or not. A soft shadow has to estimate how far the blocker is from the surface and blur accordingly, which takes many lookups per pixel.
- Do shadows use the CPU or the GPU?
- Rendering and filtering them is GPU work. Some games also spend CPU time deciding which objects cast shadows, so very high shadow distance can load the processor too.