A professional rendering workstation needs enough memory to hold the whole scene, its textures and caches, plus the applications around it. For GPU rendering, Puget Systems recommends system RAM of about twice the total GPU memory. For CPU rendering there is no fixed ratio; scene size decides, and 128GB or more is a sound starting point. Most professional machines land between 64GB and 512GB.
Starting points by workload
| Workload | Suggested system RAM | Source / basis |
|---|---|---|
| Blender, small projects | 16 to 32GB | Puget Systems |
| Blender, high-end users | 64 to 128GB | Puget Systems |
| GPU rendering (V-Ray GPU and similar) | About 2 x total GPU memory | Puget Systems |
| CPU rendering (V-Ray CPU, Corona-style) | 128GB or more if unsure; scene size decides | Puget Systems |
| Large scenes, simulation caches, many apps open | 256GB to 512GB | Workstation platforms |
GPU rendering: size from the VRAM
GPU renderers load the scene into graphics memory, but system memory stages it, holds the application and caches, and handles anything that spills over. That is why system RAM is sized from total GPU memory:
| Total GPU memory | System RAM at 2x | Practical kit |
|---|---|---|
| 32GB (one card) | 64GB | 64GB |
| 64GB (two 32GB cards) | 128GB | 128GB |
| 96GB (two 48GB cards) | 192GB | 256GB |
| 192GB (four 48GB cards) | 384GB | 512GB, workstation platform |
CPU rendering: size from the scene
CPU renderers hold the full scene in system memory, so the largest scene you render sets the number. Open your heaviest scene, watch peak memory during a render, then add headroom for the next project. Many cores also help, and a workstation platform keeps them fed with more memory channels; see capacity vs speed vs bandwidth.
Which platform to use
| Platform | Channels | Practical memory ceiling | Fits |
|---|---|---|---|
| Desktop (AM5, LGA1851) | 2 | 192 to 256GB; four modules run slower | Single-GPU artists, small to mid scenes |
| AMD TRX50 | 4 | Up to 1TB (512GB with 128GB modules) | Two to three GPUs, large scenes |
| Intel W790 | 4 or 8 | Up to 2TB or 4TB | Xeon W studios and OEM workstations |
| AMD WRX90 | 8 | Up to 2TB | Four or more GPUs, heavy CPU rendering |
On desktops, four modules run slower: AMD lists DDR5-5600 with two modules on Ryzen 9000 but DDR5-3600 with four. Workstation platforms use registered ECC memory with one module per channel, so capacity grows without that penalty. For help choosing, see TRX50 vs WRX90: how much RAM can you install.
Memory for rendering workstations
NEMIX RAM lists matched DDR5 ECC RDIMM kits for WRX90 (ASUS Pro WS WRX90E-SAGE SE, ASRock WRX90 WS EVO), TRX50 (ASUS Pro WS TRX50-SAGE WIFI, ASRock TRX50 WS), W790 (ASUS Pro WS W790-ACE and W790E-SAGE SE) and complete workstations such as the Dell Precision 7960 and Lenovo ThinkStation P7, plus DDR5 UDIMM and desktop memory for single-GPU machines. Browse DDR5 RDIMM and DDR4 RDIMM for older render nodes.
Frequently asked questions
How much RAM do I need for GPU rendering?
Puget Systems recommends system memory of about double the total GPU memory for GPU rendering in V-Ray. Two 32GB cards (64GB of VRAM) point to about 128GB of system RAM.
How much RAM do I need for CPU rendering?
There is no fixed ratio; scene size decides. Puget suggests 128GB or more is a sound starting point for CPU rendering in V-Ray if you are unsure.
How much RAM does Blender need?
Puget lists 16GB for small projects, 32GB for larger ones, and 64 to 128GB for most high-end users.
Does ECC matter for rendering?
It matters most for long, unattended renders and render nodes, where a memory error can crash a job hours in. Workstation platforms with RDIMM use ECC by design.
Why does my desktop limit RAM for rendering?
Desktop platforms have two memory channels and a lower capacity ceiling, and filling four slots lowers the speed (AMD lists DDR5-3600 for four modules on Ryzen 9000). Workstation platforms add channels and capacity.
Buying for more than one system
IT teams and integrators that maintain render nodes and artist workstations can standardize on one kit per model. Request a quote with your models and quantities, or send us your system model and we will confirm the right module before you order.
Steps and rules are summarized from the vendors' published documentation; follow the guide for your exact model.