GPU engine

What is the GPU engine and why should we use it?

The GPU engine is an alternative to its CPU counterpart. It allows you to use your GPUs (graphics cards) to compute the renders. They are usually much much faster than the CPUs, especially when using more than one graphics card. In this version, many lines of code have changed in order to make the GPU engine even faster and to allow for rendering bigger images.

Not all the graphics cards will work; please check the next section to know about the details.

GPU engine hardware requirements

• Graphics cards have to be based on CUDA. The GPU engine is built upon CUDA computing platform, which is developed by Nvidia, so only Nvidia graphics cards will work. AMD or Intel cards won't work.

• Maxwell, Pascal, Volta, Turing and Ampere micro-architectures are supported (compute capability 5.0 and up). Kepler architecture is not supported but cards based on it could work.

• It's very important that graphics card drivers are up to date.

Here is an illustrative table with most of the supported graphics cards: Table of supported GPUs

Multi-GPU

Maxwell now allows you to use all the graphics cards on your computer to calculate the render.

In order to choose which graphics cards you want to use, you will find a gear icon next to Engine selector in Render Options, when the GPU engine is selected. The plugins have a similar button on their equivalent menu.

This button will open the Maxwell GPU Preferences panel.

It will set Maxwell's behavior on a particular computer for all Maxwell apps and plugins. These settings are not sent within a scene file or distributed across the network as each computer can have very different configurations.

In the case of FIRE (the interactive preview), Maxwell will only use one of the cards (the best one) and will be marked with [FIRE] tag.

Just mark the ones you want to use and leave unmarked the ones you want to spare.

This panel can also be opened from the Maxwell installation folder by running the “mxgpuprefs” application.

Will I get the same results using the GPU and the CPU engines?

Our aim is to make them render exactly the same result, however, the nature of the two engines is not the same.

Additionally, not all the features available in the CPU engine are supported yet in the GPU engine. We will keep working to add items to the list of supported features.

Here is a table summarizing the features currently supported or not supported in the GPU engine:

	Materials			Environment			Extensions			Channels			Channels			Camera
	Referenced MXM			Physical Sky			Maxwell Scatter			Alpha			Deep			Stereo lenses
	Normal layer mode			Constant Dome			Maxwell Sea			Z-buffer			UV			Simulens
	Additive layer mode			Image Based			Maxwell Grass			Shadow			Custom Alpha			Camera Response
	Layer opacity and masks			Rendering			Maxwell Hair			Material ID			Reflectance			Tone mapping
	Transmittance			Interactive render (FIRE)			Asset Reference			Object ID			Camera			Miscellany
	AGS			Production render			Subdivision modifier			Triangle ID			Thin Lens			Render Booleans
	Sub-surface scattering			Network Rendering			Maxwell Cloner			Motion Vector			Pinhole			Motion Blur *
	Dispersion			Multi-GPU			Maxwell Volumetrics			Roughness			Ortho			Region Render
	Procedural textures			Denoiser			Maxwell Particles			Fresnel			Fisheye			Hidden to Camera
	Displacement (pretess mode)			Multilight **			Maxwell Mesher			Normals			Spherical			Hidden to Refl/Refr
	Coatings			Maxwell Cloud						Position			Cylindrical			Hidden to GI

A tip regarding antialiasing (a little hack)

The antialiasing when using the GPU engine differs from the one used with the CPU engine. However, there is a bit hidden way of tuning it.

You can do it from Maxwell Render, while the render is running or after it has finished. Go to the Camera panel > Overlay Text section but do not activate it. In its text box, enter for example, "yes 1 1" (without quotes) and then refresh the render. The first value can be 0, 1, 2 or 3 and changes the type of antialiasing according to the table below; the second value can be between 0 and 2, but it can have decimals (it can be understood as the opacity or the intensity of the pixels of the antialiasing). 

Which card should I buy to use with the GPU engine?

Only Nvidia, no AMD or Intel, sorry.

As usual, the bigger the numbers, the better, but which numbers do really affect the render?

• The memory of the graphics is crucial as the whole scene has to fit in it to be able to render, so the higher the better. The amount of memory is not relevant for the speed of the render though; this will affect the size of the image you can render, the amount of geometry you can load and the number and size of the textures that can be loaded to render.

• The higher the number of CUDA cores and their speed, the better. This will determine the rendering speed for that particular graphics card.

• The size of the memory interface and its bandwidth will also affect speed as it affects how fast the information travels between the memory and the cores. The higher the better.