Thank you for helping us improve the quality of Unity Documentation. Although we cannot accept all submissions, we do read each suggested change from our users and will make updates where applicable.Close
For some reason your suggested change could not be submitted. Please try again in a few minutes. And thank you for taking the time to help us improve the quality of Unity Documentation.Close
Unity computes shadow map sizes this way:
First light’s “coverage box” on the screen is computed. This is what rectangle on the screen the light possibly illuminates:
At “High” shadow resolution, the size of the shadow map then is:
NextPowerOfTwo( pixel size * 1.9 ), but no more than
NextPowerOfTwo( pixel size ), but no more than
NextPowerOfTwo( pixel size * 0.5 ), but no more than
When graphics card has 512MB or more video memory, the upper shadow map limits are increased (4096 for Directional, 2048 for Spot, 1024 for Point lights).
At “Medium” shadow resolution, shadow map size is 2X smaller than at “High” resolution. And at “Low” resolution, it’s 4X smaller than at “High” resolution.
The seemingly low limit on Point lights is because they use cubemaps for shadows. That means six cubemap faces at this resolution must be in video memory. They are also quite expensive to render, as potential shadow casters must be rendered into up to six cubemap faces.
When running close to video memory limits, Unity will automatically drop shadow map resolution computed above.
Generally, memory for the screen (backbuffer, frontbuffer, depth buffer) and memory for render textures has to be in video memory, Unity will use both to determine allowed memory usage of shadow maps. When allocating a shadow map according to size computed above, it’s size will be reduced until it fits into
(TotalVideoMemory - ScreenMemory - RenderTextureMemory) / 3.
Assuming all regular textures, vertex data and other graphics objects can be swapped out of video memory, maximum VRAM that could be used by a shadow map would be
(TotalVideoMemory-ScreenMemory-RenderTextureMemory). But exact amounts of memory taken by screen and render textures can never be determined, and some objects can not be swapped out, and performance would be horrible if all textures would be constantly swapping in and out. So Unity does not allow a shadow map to exceed one third of “generally available” video memory, which works quite well in practice.