制限

Eevee's goal is to be an interactive render engine. Some features may not be there yet or may be impossible to implement into Eevee's architecture without compromising performance.

Here is a rather exhaustive list of all the limitations you can expect while working with Eevee.

Cameras(カメラ)

現在、Perspective(透視投影)とOrthographic projections(平行投影)のみがサポートされています。

Lights(照明)

Only 128 active lights can be supported by Eevee in a scene.
Only 8 Shadowed sun lights can be supported at the same time.
As of now, lights can only have one color and do not support light node trees.

Light Probes(ライトプローブ)

Eevee only supports up to 128 active Reflection Cubemaps.
Eevee only supports up to 64 active Irradiance Volumes.
Eevee only supports up to 16 active Reflection Planes inside the view frustum.

Indirect Lighting(間接照明)

Volumetrics don't receive light from Irradiance Volumes but do receive world's diffuse lighting.
Eevee does not support "specular to diffuse" light bounces nor "specular to specular" light bounces.
All specular lighting is turned off during baking.

Shadows(影)

Only 128 active lights can be supported by Eevee in a scene.
Only 8 Shadowed sun lights can be supported at the same time.

Volumetrics(ボリューメトリック)

単一散乱のみがサポートされています。
Volumetrics(ボリューメトリック)は、カメラの "レイ" に対してのみレンダリングされます。それらは反射/屈折およびプローブには現れません。
Volumetrics don't receive light from Irradiance Volumes but do receive diffuse lighting from the world.
Volumetrics(ボリューメトリック)シャドウイングは、ボリューメトリックでのみ機能します。シーン内のソリッドオブジェクトに影を落とすことはありません。
Volumetrics(ボリューメトリック)シャドウイングは、ビュー錐台内のボリュームに対してのみ機能します。
Volumetric lighting do not respect the lights shapes. They are treated as point lights.

Depth of Field(被写界深度)

Alpha blended surfaces cannot be correctly handled by the post-processing blur, but will be correctly handled by the sample-based method. For this, you need to disable the post-process depth of field by setting the Max Size to 0.

Screen Space(スクリーンスペース) エフェクト

Eevee is not a ray tracing engine and cannot do ray-triangle intersection. Instead of this, Eevee uses the depth buffer as an approximated scene representation. This reduces the complexity of scene scale effects and enables a higher performance. However, only what is in inside the view can be considered when computing these effects. Also, since it only uses one layer of depth, only the front-most pixel distance is known.

これらの制限により、いくつかの問題が発生します:

スクリーンの境界に達すると、スクリーン空間のエフェクトは消えます。これは、 オーバースキャン 機能を使用して部分的に修正できます。
スクリーン空間のエフェクトには、詳細な情報(またはオブジェクトの幅)がありません。これが、ほとんどのエフェクトに、交差する可能性のあるピクセルを考慮する方法を制御するためのThickness(幅)パラメーターがある理由です。
ブレンドされたサーフェスは、これらのエフェクトによって考慮されません。これらは深度プレパスの一部ではなく、深度バッファーには表示されません。
Holdout(ホールドアウト)コレクションの一部がスクリーン空間のエフェクトでレンダリングされないオブジェクト。

Ambient Occlusion(アンビエントオクルージョン(AO))

Objects are treated as infinitely thick, producing overshadowing if the Distance is really large.

Screen Space Reflections

Only one glossy BSDF can emit screen space reflections.
The evaluated BSDF is currently arbitrarily chosen.
Screen Space Reflections will reflect transparent objects and objects using Screen Space Refraction but without accurate positioning due to the one layer depth buffer.

Screen Space Refraction

Only one refraction event is correctly modeled.
Only opaque and alpha hashed materials can be refracted.

Subsurface Scattering(SSS)

Only one BSSRDF can produce screen space subsurface scattering.
The evaluated BSSRDF is currently arbitrarily chosen.
A maximum of 254 different surfaces can use subsurface scattering.
Only scaling is adjustable per pixel. Individual RGB radii are adjustable in the socket default value.
Input radiance from each surfaces are not isolated during the blurring, leading to light leaking from surface to surface.

Motion Blur(モーションブラー)

Motion Blur is only available in final renders and is not shown in the 3D Viewport and thus Viewport Renders.

Materials(マテリアル)

Refractions

Refraction is faked by sampling the same reflection probe used by the Glossy BSDFs, but using the refracted view direction instead of the reflected view direction. Only the first refraction event is modeled correctly. An approximation of the second refraction event can be used for relatively thin objects using Refraction Depth. Using Screen Space refraction will refract what is visible inside the view, and use the nearest probe if there is no hit.

Screen Space Reflections and Ambient Occlusion are not compatible with Screen Space Refraction; they will be disabled on the surfaces that use it. Surfaces that use Screen Space Refraction will not appear in Screen Space Reflections at the right place. Surfaces that use Screen Space Refraction will not cast Ambient Occlusion onto other surfaces.

Volume Objects

Object volume shaders will affect the whole bounding box of the object. The shape of the volume must be adjusted using procedural texturing inside the shader.

Shader Nodes(シェーダーノード)

All BSDF's are using approximations to achieve realtime performance so there will always be small differences between Cycles and Eevee.
Some utility nodes are not yet compatible with Eevee (e.g. Sky Texture node).

参考

For a full list of unsupported nodes see Nodes Support.

メモリの管理

In Eevee, GPU Memory management is done by the GPU driver. In theory, only the needed textures and meshes (now referred as "the resources") for a single draw call (i.e. one object) needs to fit into the GPU memory.

したがって、シーンが非常に重い場合、ドライバーはすべてのオブジェクトが正しくレンダリングされるように、メモリの入れ替えをします。

実際には、GPUメモリを使いすぎると、GPUドライバーがクラッシュしたり、フリーズしたり、アプリケーションが強制終了したりする可能性があります。だから注意してください。

リソースがGPUメモリに収まるかどうか、および/または、GPUがリソースを正常にレンダリングするかどうかを見当する標準的な方法はありません。

CPU レンダリング

Being a rasterization engine, Eevee only uses the power of the GPU to render. There is no plan to support CPU (software) rendering as it would be very inefficient. CPU power is still needed to handle high scene complexity as the geometry must be prepared by the CPU before rendering each frame.

マルチGPUサポート

現在、複数の GPU システムはサポートしていません。

ヘッドレスレンダリング

There is currently no support for using Eevee on headless systems (i.e. without a Display Manager).