Resolution Scale Calculator

Background.

A resolution-scale slider is labelled in one dimension and paid for in two. Drop it from 100 to 75 per cent and each side of the render target shrinks by a quarter — but the number of pixels the GPU has to shade drops to 56.25 per cent, because 0.75 squared is 0.5625. That square law is the reason a modest-sounding slider move produces such a large change in frame rate, and it is the number this calculator exists to put in front of you.

Everything else follows from it. Rendering at half the linear resolution is a quarter of the pixels. A 67 per cent render scale, the value many games default to for their performance presets, is about 45 per cent of the pixels — the GPU is doing well under half the shading work. Going the other way, a 200 per cent scale for downsampling is 400 per cent of the pixels, which is why supersampling is so expensive.

The calculator works in both directions. Give it a source resolution and a percentage and it returns the exact render target, rounded to whole pixels, along with the pixel count and the ratio. Give it a source and a target resolution instead and it returns the render-scale percentage that maps one onto the other, plus the same pixel comparison. On a 3840 x 2160 panel, rendering at 1920 x 1080 is a 50 per cent render scale and exactly 25 per cent of the pixels.

The source resolutions offered as examples are not invented. Recommendation ITU-R BT.709-6 specifies 1 920 samples per active line and 1 080 active lines with a 16:9 aspect ratio and 1:1 square pixels; Recommendation ITU-R BT.2020-2 specifies 3 840 x 2 160 and 7 680 x 4 320 with the same 16:9 shape, an orthogonal sampling lattice and the same 1:1 pixel aspect ratio. Square pixels are what make a single scale factor legitimate in the first place — if the pixel aspect ratio were not 1:1, scaling both axes by the same number would distort the picture.

One honest caveat. Pixel count is a proxy for shading and post-processing cost, not a frame-rate prediction. Vertex processing, draw-call submission, physics, animation and CPU-side simulation do not get cheaper when the render scale drops, so a workload that is CPU-bound will barely move. Treat the pixel ratio as the ceiling on what lowering the scale can buy you, not as a promise.

What is resolution scale calculator?

Resolution scale, also called render scale or dynamic resolution, is the practice of rendering a 3D scene into a buffer smaller or larger than the display's native resolution and then resampling the result to fit. It is expressed as a linear percentage applied to each axis, so a 50 per cent scale on a 3840 x 2160 output renders at 1920 x 1080 before upscaling.

The crucial property is that cost scales with area, not with the label. Because both axes shrink together, the pixel count — and with it the fragment-shading, lighting and post-processing work — changes with the square of the percentage. A 90 per cent scale is 81 per cent of the pixels, 75 per cent is 56.25 per cent, 67 per cent is about 45 per cent, and 50 per cent is 25 per cent. Above 100 per cent the same relationship makes supersampling expensive: 141 per cent is roughly double the pixels and 200 per cent is quadruple.

How to use this calculator.

1. 01Choose whether you know the scale percentage or the target resolution.
2. 02Enter the source resolution — your panel or output raster, in whole pixels.
3. 03In the first mode, enter the render-scale percentage. 100 is native, below 100 is upscaling from a smaller buffer, above 100 is downsampling from a larger one.
4. 04In the second mode, enter the target width and height instead; the percentage is derived from the width, and a mismatched target shape is flagged in the result.
5. 05Read the render target for the actual pixel dimensions, and the pixels-vs-source figure for the shading workload the change implies.

The formula.

Scaling a render target is a uniform multiplication on both axes: W′ = W × s and H′ = H × s, where s is the linear scale as a fraction. A render target has to be a whole number of pixels, so each result is rounded half up and floored at one pixel — a renderer cannot allocate a fractional or empty buffer. Because that rounding perturbs the requested percentage, the calculator reports the effective scale back from the rounded raster rather than repeating the number you typed: 3840 at 67 per cent gives 2572.8, which becomes 2573 and an effective 67.005 per cent.

The headline output is the pixel ratio, (W′ × H′) ÷ (W × H) × 100. For an exact uniform scale this collapses to s² × 100, which is the whole story of why render scale is such an effective performance control. A 25 per cent reduction in each linear dimension removes 43.75 per cent of the pixels; a 33 per cent reduction removes 55 per cent of them. The relationship is symmetric above 100 per cent, where a 2x linear scale means 4x the pixels.

In target mode the percentage is derived from the widths, W′ ÷ W × 100, and the pixel ratio is computed from the actual pixel counts. If the target's aspect ratio does not match the source's, the two axis scales differ, and the calculator says so explicitly rather than pretending the operation was uniform — scaling 3840 x 2160 to 1920 x 1200 shrinks the width by 50 per cent but the height by only 55.56 per cent, which would stretch the picture unless it is letterboxed or cropped.

All of this assumes square pixels, which the relevant standards specify: ITU-R BT.709-6 item 2.5 and ITU-R BT.2020-2 Table 1 both give a pixel aspect ratio of 1:1 for the HD and UHD rasters. With non-square pixels a single scalar scale factor would change the picture's shape, and the anamorphic case needs a different tool.

A worked example.

Frequently asked questions.

References& sources.

1. [1]Recommendation ITU-R BT.709-6 (06/2015), Parameter values for the HDTV standards for production and international programme exchange — item 2.1 aspect ratio 16:9, 2.2 samples per active line 1 920, 2.4 active lines per picture 1 080, 2.5 pixel aspect ratio 1:1 (square pixels).
2. [2]Recommendation ITU-R BT.2020-2 (10/2015), Parameter values for ultra-high definition television systems for production and international programme exchange — Table 1, Picture spatial characteristics: pixel count 7 680 x 4 320 and 3 840 x 2 160, picture aspect ratio 16:9, orthogonal sampling lattice, pixel aspect ratio 1:1 (square pixels).
3. [3]W3C, CSS Values and Units Module Level 4, Absolute Lengths — the reference pixel defined as 'the visual angle of one pixel on a device with a device pixel density of 96dpi', the basis for distinguishing a CSS pixel from a device pixel.