Understand the parameters that make an ideal rendering system
Rendering is one of the most important functions when it comes to processing advanced graphical media. Simply put, it is the process of generating realistic images from basic 2D or 3D models. Rendering makes use of a scene file consisting of multiple 3D or 3D models, which contains all relevant information in a predefined format such as the geometry, texture, viewpoint, and lightning descriptions, to name a few. The scene file is submitted to the rendering software for processing, and the final result is a digital image file. Traditionally, rendering was been performed on CPUs, but many applications can now take advantage of GPUs to render far faster than a CPU alone. Some rendering applications only work on CPU or GPU, so it is important to know how it will work so you have clear understanding on the time to results and their quality. There is also a growing number of applications that allow rendering on the CPU and GPU together, offering more performance and flexibility. Let’s compare them.
CPU rendering applications are better at intricate projects, as they can smoothly handle a set of more diverse tasks that require different computing operations, as opposed to GPUs which are designed to process huge chunks of data by executing
them in the same way over and over again.
CPU rendering relies on system memory which can be scaled much higher than GPU memory - this combined with the added intimacy capability leads to a higher quality render in many cases.
The downside is that CPUs are much slower at rendering than GPUs so to really scale performance you need a workstation with lots of CPUs cores or ideally multiple servers with dual CPUs running together in a cluster, which can get very costly and complex to manage.
GPUs are ideal for rendering as they have thousands of cores working parallel. A render that will take a CPU hours may be complete in minutes on a GPU. However, there may be some tasks where some intricate detail may be lost.
Although a separate component from the CPU and featuring dedicated memory, GPU cards can be scaled and performance is delivered through the many cores and rapid throughput.
As scene complexity and resolution increases it’s often far more cost effective to use a workstation with four GPUs than multiple CPU servers in a cluster. However, for the largest of products a datacentre NVIDIA EGX server with eight GPUs, or multiple servers linked together in a cluster is also possible.
CPU & GPU Rendering
Applications that can be set to use either CPU or GPU offer ultimate flexibility as both CPU-based renderers and GPU-based renderers are perfect for their specific tasks.
In fact, GPU only serves to enhance your current CPU system and allows you to significantly accelerate image rendering. Your GPU can take on the resource-intensive 3D visualisation elements, as your CPU executes the remaining tasks.
In large server deployments multi-GPU servers can be used with two CPUs and large system memory to deliver the ultimate performance whichever approach you choose.
Can't see your favourite application?
Please contact our team of Pro Graphics experts who can help you pick the perfect rendering solution.