GPU Rendering

The Best Performing Graphics Cards for Rendering

Traditionally the only way to shorten render times was to buy faster CPUs or more systems and render over the network. As a result once you had finished creating the models, textures and animations for your scene the expensive graphics card in your graphics workstation was left largely idle.

However as graphics cards have become more powerful and frame buffer sizes have increased it's now possible to use them to accelerate rendering. GPUs are so good at rendering because unlike a CPU which features a small quantity of highly complex cores, a GPU has hundreds or thousands of comparatively simple cores that are perfect for breaking down the complex task of rendering into lots of small bite size chunks.

As a result many rendering engines now support GPU rendering as well as or instead of the CPU, with dramatic performance gains from anything from 6x to 15x depending on the renderer. For this reasons its crucial to understand what hardware your chosen render engine can take advantage of, especially as this will change over time as patches and new versions of the software are released as well as new graphics drivers.

We supply a wide range of GPU rendering systems, from dedicated rendering servers to Graphics Workstations that can also be used for modelling and animation.

GPU Rendering Performance

The most important characteristic to look out for in a graphics card for rendering is the number of cores, as this is the part of the GPU used to process all the calculations. Second, make sure you choose a graphics card that has sufficient RAM to fit all the geometry and textures inside. If the frame buffer is not large enough, then some assets will overflow into main system RAM which runs much slower than the RAM on the graphics card.

Render Graphics Cards

While its possible to slash render times by using GPUs its still important to have a fast CPU as this is heavily used during the load and setup processes. Other renderers, such as Thea Render also use the CPU during rendering, achieving an even more dramatic speed up than CPU or GPU rendering alone.

In this guide we'll look at how different hardware configurations perform in some of the most popular renderers. As renderers are coded very differently from each other it's important not to use this data to compare one renderer to another; however you can use it to see how effective upgrading your render systems can be.

Iray 2015 Rendering Performance

NVIDIA Iray is a physically based rendering engine that is available as a plugin for 3ds Max, Cinema 4D, Maya, Revit and Rhino. The graph below shows the rendering performance you can expect to see in 3ds Max when rendering using the NVIDIA Iray plugin on various CPUs and GPUs. The scores are shown as relative to a single Intel Xeon E5 CPU.

NVIDIA Iray

2x Quadro M6000

9.5

2x Quadro M5000

7.3

GeForce GTX Titan X

6.4

Quadro M6000

6.2

2x Quadro M4000

5.6

GeForce GTX 980

4.8

Quadro M5000

4.6

Tesla K40

4.5

Quadro M4000

3.5

Quadro K2200

2.3

Dual Xeon E5 CPUs

1.8

Single Xeon CPU

1.0

Octane Render Rendering Performance

OTOY Octane Render is an unbiased physically based renderer and uses NVIDIA CUDA to unlock the power of GPUs. The graph below shows the average Ms/s (mega samples per second) on a variety of different GPUs, with a higher sample rate meaning shorter render times.

Octane Render Performance

8x GeForce GTX Titan X

1065

8x Tesla K80

591

4x GeForce GTX Titan X

522

4x GeForce GTX 980 Ti

500

4x GeForce GTX 980

377

4x GeForce GTX 970

327

4x Tesla K80

298

2x GeForce GTX Titan X

252

2x GeForce GTX 980 Ti

250

2x Quadro M6000

234

2x GeForce GTX 980

185

2x GeForce GTX 970

161

2x Tesla K80

146

GeForce GTX Titan X

126

GeForce GTX 980 Ti

125

Quadro M6000

121

GeForce GTX 980

97

Quadro M5000

86

GeForce GTX 970

80

Tesla K80

73

Quadro M4000

55

FurryBall RT Rendering Performance

AAA Studio Furyball RT is GPU renderer that has unbiased and biased modes and is available as a plugin for 3ds Max, Cinema 3D and Maya. The graph below shows the performance you can expect on a variety of different GPUs relative to a single GeForce GTX Titan graphics card.

Furryball RT Performance

4x GeForce GTX 980 Ti

2.59

4x GeForce GTX Titan X

2.58

4x GeForce GTX 970

2.02

2x GeForce GTX 980 Ti

1.72

2x GeForce GTX Titan X

1.69

2x Quadro M5000

1.15

2x GeForce GTX 980

1.37

2x GeForce GTX 970

1.25

GeForce GTX 980 Ti

1.12

GeForce GTX Titan X

1.05

Quadro M6000

1.02

GeForce GTX Titan

1.00

GeForce GTX 980

0.86

GeForce GTX 970

0.75