For most GPU-based rendering engines, the CPU only plays a fairly minor roll when it comes to how fast it takes to render an image. In fact, even the rendering engines that support both CPUs and GPUs (such as Iray) do not generally see much of a decrease in render times with a faster CPU. The CPU can make a difference if you only have a single video card, but as soon as you have two or more cards, the power of your CPU makes very little difference in render times.
However, while the GPU is the most critical factor when using a GPU-based rendering engine, a more powerful CPU is not always a waste of money. In fact, a faster CPU will still often let you open scenes faster and will assist in any multi-frame or animations. So while the actual time to render each frame may be almost entirely dependent on the video cards, your total work time can still be lower with a faster CPU.
Video Card (GPU)
As the name "GPU-based Rendering Engine" suggests, the video card is the most critical component in a rendering workstation that uses Iray, V-Ray RT, Octane Render, or any other GPU-based rendering engine.
The key thing to know with these rendering engines is that they are so efficient at utilizing multiple video cards that you will almost always get better performance for your money by focusing on getting as many video cards as possible before worrying about the power of each individual card. In fact, if you look at the articles listed below you will see that while three GTX 970's are about the same cost as a single GTX Titan-X, the three GTX 970 cards should be able to complete renders in about half the time.
The only time when spreading your budget across multiple cards is not a good idea is if your scenes are large enough that you need a larger amount of VRAM (memory built into the video card) than the cards can provide. Currently, the entire scene you are rendering needs to be loaded into each card individually so even if you have three GPUs with 4GB of VRAM, you actually only have 4GB of VRAM to work with (not the 12GB you would expect if you added the cards together). While the amount of VRAM you need will depend on your exact scene, a good rule of thumb is to have at least 1GB of VRAM for every 4.5 million polygons.
In general, we recommend using an NVIDIA Quadro video card for the primary card in a rendering workstation. Compared to the mainstream GeForce cards, they tend to be more reliable and have better driver support. In addition, many 3d Design packages (including those from AutoDesk and SOLIDWORKS) either recommend or require Quadro cards for their software. This level of support is usually not needed for the secondary cards (which only the rendering engine utilizes), however, so we typically recommend using GeForce cards for the accelerator cards as they will give you much higher performance than their Quadro equivalents.
While the exact amount of RAM you need is going to depend on your particular model/scene and render settings, GPU-based rendering engines typically do not need large amounts of system memory since you will most often be limited by the amount of VRAM your video card(s) have. Still, we generally recommend a minimum of 32GB for our rendering workstations.
Depending on what you are doing, however, RAM usage can quickly shoot up so both of our recommended systems are capable of much larger amounts of RAM. Our smaller, Dual GPU system can utilize up to 256GB of RAM while the larger, Quad GPU system is capable of up to 512GB.
Storage (Hard Drives)
With the falling costs associated with SSDs, we almost always recommend using an SSD for the primary drive that will host your OS and the installation of software. The high speed of SSDs allows your system to boot and launch applications many times faster than any traditional hard drive.
Since SSDs are still more expensive than traditional drives per GB, for long term storage we recommend using a traditional hard drive (or two if you need even more storage!). Using a SSD can be useful in some situations, but most of the time the high performance of an SSD is simply not required for a storage drive.