What next?
Render scale trades sharpness for frames. These price that trade.
What your result means
Rendering at is the resolution the game is genuinely drawing before it is stretched or shrunk to fit your screen. Pixels per frame is the work that follows from it, and share of native pixels is the figure the slider does not tell you: because the percentage applies to both width and height, the pixel cost moves with its square. 71% scale is close to half the pixels; 50% scale is a quarter of them. Compared with 1080p and roughly equivalent to translate an abstract percentage into resolutions you can already picture.
Why this one is different
Scaling is expressed in pixels rather than percentages: the resolution actually being rendered, the pixels per frame that implies, and what share of native that is. A comparison against 1080p and the resolution your setting is roughly equivalent to make it obvious whether the performance saving is worth the softness.
Dropping render scale to 71% halves the pixels, not 29% of them
A render scale percentage is applied to each axis, so the pixel count follows the square of it. Set 71% and each axis loses 29%, but the area — the thing the GPU actually shades — falls to 0.71² = 50%. Set 50% and you are drawing a quarter of the pixels, which is why a half-scale image looks so much softer than the number suggests.
It works upwards too. Supersampling at 141% doubles the pixel count and at 200% quadruples it, which is why a modest-sounding "125%" costs over 50% more work. Read the share-of-native figure above rather than the slider, and any settings decision involving resolution becomes a great deal more predictable.
How it works
A render scale multiplies both the width and the height of the render target. The rendered resolution is the display width and height each multiplied by the scale and rounded to whole pixels; the pixel count is the product of those two. Because both axes move together, the number of pixels changes with the square of the scale — the relationship that makes resolution the single most expensive graphics setting in almost every game. The rendered image is then resampled to your panel’s native resolution for display, which is where any softness or shimmer comes from.
How to use this calculator
- Pick your display resolution — the panel’s native resolution, not a lower one you have selected in-game.
- Move the render scale to the value the game is using. Names vary: render scale, resolution scale, 3D resolution, supersampling, or a DLSS/FSR quality preset.
- Read the share of native pixels, which is the honest measure of the change in GPU work.
- Use roughly equivalent to as a sanity check: if 1440p at 67% is drawing fewer pixels than native 1080p, running 1080p natively may look better than scaling 1440p down to the same cost.
Formula
W and H = your display’s native width and height in pixels, s = render scale as a decimal (67% → 0.67). Because both axes scale, pixels ≈ W × H × s² — the square is the part the slider hides.
Example calculation
A 1440p display (2560 × 1440) at a 67% render scale:
Rendered height = round(1440 × 0.67) = 965
Pixels = 1715 × 965 = 1,654,975
Native = 2560 × 1440 = 3,686,400 px
Share = 1,654,975 ÷ 3,686,400 = 44.9% (0.67² = 0.4489)
Native 1080p = 2,073,600 px, so this is 79.8% of 1080p
Two thirds on the slider is under half the pixels, and less work than rendering 1080p natively. That comparison is the practical one: at this setting a native 1080p image would be sharper for slightly more GPU time.
Frequently asked questions
Why is 50% render scale a quarter of the pixels?＋
Because the scale applies to width and height independently. Half the width and half the height gives a quarter of the area, and area is what a GPU shades. The same arithmetic explains why 71% is half the pixels and why 141% is double them — every render-scale figure should be squared before you reason about its cost.
Is 1440p at 67% better or worse than native 1080p?＋
Slightly worse for image quality at slightly lower cost, in most cases. 1440p at 67% draws about 1.65 million pixels against 1080p’s 2.07 million, and it then has to be resampled to a 1440p panel, which softens it further. If your monitor is 1440p, a modern upscaler at a quality preset will usually beat both; if you are choosing between plain scaling options, native at a lower resolution tends to look cleaner than scaled-down at a higher one.
Does this apply to DLSS, FSR and XeSS?＋
The pixel arithmetic does — each quality preset corresponds to a render scale, and the internal resolution follows exactly the maths above. What the tool cannot model is the reconstruction: these techniques use motion vectors and previous frames to recover detail, so their output looks considerably better than a plain resample of the same pixel count. Use this page for the cost, not for the quality.
What are the render scales behind the common quality presets?＋
They are broadly standardised: quality is around 67% per axis, balanced around 58%, performance 50% and ultra performance 33%. Squaring those gives roughly 45%, 34%, 25% and 11% of native pixels respectively, which is why the aggressive presets recover so much frame rate and lose so much detail.
Why does the rendered resolution have odd numbers in it?＋
Because a percentage of a pixel dimension rarely lands on a whole number, and the render target must be whole pixels. 2560 × 0.67 is 1715.2, which rounds to 1715. Some engines round to a multiple of two or eight for alignment reasons, so your game may report a figure a pixel or two from this one.
Is resolution the setting to lower first for more frame rate?＋
It is usually the most effective single change, because it scales the cost of nearly every per-pixel operation at once. Whether it should be first depends on what you are willing to lose: resolution costs sharpness everywhere, while shadows, reflections and volumetrics are often expensive and localised. A common approach is to cut one or two of the heaviest effects, then use render scale for whatever is still missing.
Related calculators
Assumptions & limitations
Every figure here comes from a simplified model. Keep these limits in mind when reading your result:
- Reports the rendered pixel count only. It is a measure of how much the GPU has to shade, not a frame-rate prediction — the relationship between pixels and frames depends on whether your bottleneck is the GPU at all.
- Rounds each axis to whole pixels and multiplies, which is what a render target does. Engines that align to multiples of two or eight will report a marginally different resolution.
- Assumes the render scale applies uniformly to both axes. A few games and VR runtimes scale asymmetrically or vary the scale dynamically per frame.
- Says nothing about image quality. Temporal upscalers reconstruct detail from previous frames and look far better than a plain resample at the same pixel count; a dynamic resolution system changes the scale continuously.
- The equivalence to standard resolutions compares pixel counts, not aspect ratios or shapes. An ultrawide render matching 4K’s pixel count is not showing a 4K image.