Recommended Hardware for Redshift:
Like most applications, the system requirements for Redshift can be found on their official website. They also have a good FAQ, which primarily focuses on GPU compatibility and performance. Since their information is split into two sections like that, and since it may tend a bit toward the minimum requirements rather than best specs, we have put together our own hardware recommendations below.
In Redshift, as well as most other GPU-based engines, the CPU does not play a direct role in the process of rendering scenes. It does have a small impact on the time spent loading a scene, but that is pretty minimal in the grand scheme of things. However, if you are also using the same system for modeling or animation - in a program like Cinema4D, Maya, or 3ds Max - then you will want a CPU with a high clock speed in order to ensure good performance in those applications. On the other hand, if you also use CPU based rendering engines then having a higher core count would benefit those programs... but it will not improve Redshift's rendering speed.
Aside from raw performance, it is also important to consider the number of PCI-Express lanes a CPU supports. This will govern how many video cards can be used, which has a big impact since video cards are the primary driver for rendering performance in Redshift.
- AMD Ryzen 7 5800X 8 Core - This is one of the highest clock speed CPUs available, and does very well with both Redshift and modeling / animation applications. It has enough PCI-Express lanes to support dual video cards, along with other standard hardware devices (NVMe drives, etc).
- AMD Threadripper PRO 3955WX 16 Core - AMD's Threadripper PRO processors have far more PCI-Express lanes than more consumer-oriented models. Combined with the right motherboard, this CPU can allow up to four video cards in a single tower workstation. The trade-off is that the platform as a whole is more expensive, but being able to fit three or four high-end GPUs in a single system is worthwhile in many workflows.
As mentioned above, the video card selection is the driving factor for performance in Redshift. The faster the better, and you can also use multiple GPUs to further speed up rendering.
There are two aspects of a video card that impact render capabilities: the raw speed of the GPU itself and the amount of memory on the card. Video memory will limit how large and complex of scenes can be rendered effectively, though Redshift does support "out of core" rendering which will allow system memory to be used if there is not enough dedicated GPU memory available... but that comes with a reduction in speed, so it is best to get video cards with enough RAM onboard if at all possible. GeForce cards tend to have good raw performance, with decent amounts of video memory, while Quadro cards come with larger amounts of VRAM but also cost more for the same level of raw performance.
- GeForce RTX 3080 Ti 12GB - A great choice if you want just one or two video cards and don't work with overly complex scenes. It is also available with blower-style coolers, enabling use of multiple cards if desired.
- GeForce RTX 3090 24GB - Our go-to recommendation for most GPU rendering customers, the RTX 3090 provides the best performance in Redshift while also having a tremendous 24GB of memory.
- NVIDIA RTX A5000 24GB - For those who want to stack several video cards in the same system, NVIDIA's professional GPUs are a solid option. The RTX A5000 is the top-end card that can be installed in a set of four in large tower chassis and still be within the limits of a 1600W power supply. If you need even more memory, the RTX A6000 has 48GB of VRAM but may be limited to three GPUs.
It is also important to remember NVIDIA's professional-grade video cards, as they can be a better choice than GeForce cards for some users. They do cost more, but for that increased price you get several benefits:
- Higher VRAM options - up to 48GB on the RTX A6000
- Better multi-GPU support - thanks to the use of blower-style cooling systems and more constrained power consumption
- ECC memory on higher-end models - for increased stability
Here is a chart showing performance of the latest NVIDIA RTX A-series cards compared to the previous generation:
Beyond the selection of which card to use is the question of how many. Redshift scales well across multiple video cards, but the cooling systems on most GeForce models are not designed with multiple GPUs in mind. For the best overall performance, variants with a single fan that exhausts heat out the back (commonly called "blower" cards) are ideal - and most NVIDIA "professional" cards use such cooling systems. Stacking a few of those will give fantastic rendering performance, though it does require a larger chassis, strong power supply, and plenty of airflow from the case fans.
Recent Redshift GPU articles:
While the exact amount of RAM you need is going to depend on your particular projects, for Redshift (and GPU rendering in general) we generally recommend 32GB. That should be plenty for rendering even very complex scenes, but our systems do support more for those who need it. If you run other software alongside Redshift, keep that in mind when selecting how much memory you get.
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 Redshift and other software. The high speed of SSDs allows your system to boot, launch applications, and load files many times faster than any traditional hard drive. If your budget allows, it is also a very good idea to have a second SSD that can be used to store your active projects to further decrease load and save times.
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.