CPU vs GPU Rendering in 2026: Which One for Cinema 4D, Redshift and Octane?
Whether CPU or GPU rendering is faster depends on your renderer and on how big your scenes are. The main points:
- The renderer decides the hardware: Redshift, Cycles, V-Ray and Arnold can render on CPU or GPU. Octane needs an NVIDIA CUDA GPU. Corona is CPU-only.
- GPU wins when the scene fits: Redshift, Octane and Cycles are much faster on modern NVIDIA GPUs, and several GPUs share the work of each frame.
- VRAM is not pooled: Each GPU holds the whole scene, so the limit is the memory of one card: 24 GB on an RTX 4090, 32 GB on an RTX 5090. OTOY states this in its Octane FAQ.
- When CPU is still right: For CPU-only renderers such as Corona, or for scenes far larger than one card's VRAM, where Redshift CPU can use the system RAM.
- Fastest path for Cinema 4D animation: With Redshift or Octane, a multi-GPU server with 2x, 4x or 8x RTX 5090 or RTX 4090 is usually the fastest way to finish an animation, as long as each frame fits in one card's memory.
CPU or GPU is not one choice for every project. It depends on the renderer you use, how big your scenes are, and how much render time you can afford. This guide explains which renderers run on which hardware, when a CPU is still the right tool, and what real Redshift jobs take on 1, 2, 4 and 8 GPU servers.
CPU vs GPU rendering: the short answer
- Use GPU rendering when your renderer supports it and the scene fits in the memory of one graphics card. Redshift, Octane and Cycles are much faster on modern NVIDIA GPUs, and adding GPUs shortens each frame.
- Use CPU rendering when your renderer is CPU-only (such as Corona), or when the scene is far larger than the VRAM of one card and slows down too much out-of-core.
- For Cinema 4D with Redshift or Octane, a multi-GPU machine is usually the fastest way to finish animations.
How CPU and GPU rendering differ
A CPU has a few dozen powerful cores and uses the large system RAM of the machine, often 128 GB or more on a workstation. A GPU has thousands of smaller cores that trace many rays in parallel, but it can only use its own VRAM: 24 GB on an RTX 4090 and 32 GB on an RTX 5090.
Several GPUs in one machine render faster, but their memory is not added together. OTOY states in its Octane FAQ that multiple GPUs give a speed-up, but VRAM is not compounded. The same rule applies to Redshift: each GPU must hold the whole scene.
A note on sources about VRAM: Maxon’s support article states that Redshift does not combine the VRAM when using multiple GPUs. Maxon’s 3D rendering product page also lists “multi-GPU memory pooling” and says Redshift shares out-of-core memory across multiple cards. This article follows the support article: plan for every GPU holding the scene in its own VRAM. Check Maxon’s latest documentation for your Redshift version.
The figure gives the short version: the CPU has fewer, larger cores and large system RAM, the GPU has many small cores but only its own VRAM, and several GPUs share the work of each frame while every card still holds the full scene.
Which renderers run on CPU, GPU or both
The renderer decides the hardware. The table summarises what each developer documents.
CPU or GPU: what each renderer uses
Based on each developer's documentation. "Hybrid" means CPU and GPU render the same frame together.
| Renderer | CPU | GPU | Notes |
|---|---|---|---|
| RedshiftMaxon | Yes | Yes | GPU, CPU or hybrid (CPU + GPU on the same frame), set under Compute Devices. Maxon's docs say a GPU is usually many times faster than a CPU, so hybrid is off by default. |
| OctaneRenderOTOY | No | Yes | Requires an NVIDIA CUDA GPU. Multiple GPUs add speed, but VRAM is not combined. |
| CyclesBlender | Yes | Yes | CPU, or GPU through CUDA/OptiX (NVIDIA), HIP (AMD), oneAPI (Intel) and Metal (Apple). |
| V-RayChaos | Yes | Yes | V-Ray (CPU) and V-Ray GPU are separate render engines; V-Ray GPU can also use the CPU as an extra device. |
| ArnoldAutodesk | Yes | Yes | CPU by default; Arnold GPU runs on NVIDIA RTX GPUs. |
| CoronaChaos | Yes | No | CPU-only by design. A GPU can still be used for denoising. |
Sources: Maxon Redshift CPU and Cinema 4D Redshift preferences, OTOY Octane FAQ, Blender Manual GPU Rendering, and Chaos Corona: proudly CPU based.
When CPU rendering is still the right choice
- Your renderer is CPU-only. Corona renders only on the CPU, so more CPU cores are what make it faster.
- The scene is far larger than one card’s VRAM. Redshift and Octane can move data out-of-core into system RAM, but it renders slower. When a scene needs much more memory than 32 GB, Redshift CPU can render it with system RAM.
- You need exactly the same image on CPU and GPU machines. When Maxon introduced Redshift CPU, it said CPU and GPU output should be perceptually indistinguishable, and that the GPU still gives the best performance.
When GPU rendering wins
For Redshift, Octane and Cycles, GPUs are usually the faster and cheaper way to finish a job. Maxon writes in its Cinema 4D documentation that a GPU is usually many times faster than a CPU for Redshift rendering, which is why hybrid rendering is switched off by default.
The bigger gain comes from using several GPUs at once. Redshift and Octane render each frame on every GPU in the machine, so a 4x or 8x GPU server shortens every frame of an animation. The table below shows real Redshift jobs on iRender servers.
From one GPU to eight GPUs on real jobs
Render time from each video, priced at the list rate of the server used (per minute; 3i has a 1-hour minimum).
| Test (iRender video) | Server | Frames & render time | Cost at list price |
|---|---|---|---|
| Cloth simulation render1920x1080 | 3i1x RTX 5090 | 301 frames 1h 46m 30s · 21.3 s/frame | about $19.17 |
| Cinema 4D animation, 30 fps1920x1080 | 4S2x RTX 4090 | 300 frames about 2h 50m · 33.95 s/frame | about $42.44 |
| Cinema 4D animation1920x1080 | 5i4x RTX 5090 | 150 frames 1h 42m · 40.84 s/frame | about $64.66 |
| Fluid and particle motion paintingCinema 4D 2025.3.3 | 9S8x RTX 4090 | 360 frames 1h 18m 13s | $70.20 |
Watch the single-GPU test and the 4x RTX 5090 test from the table.
How to choose for your next project
- Check your renderer. If it is CPU-only, choose CPU. If it supports GPU, test GPU first.
- Check scene memory. If a frame fits in 24 GB (RTX 4090) or 32 GB (RTX 5090), GPU rendering will usually be fastest.
- Render a few test frames and multiply the seconds per frame by the number of frames to see the full job time.
- Pick the GPU count from your deadline. 2x for look-dev and short jobs, 4x for most animations, 8x or several servers in parallel for long sequences.
If your own workstation has one GPU, you can run the same Cinema 4D scene on a remote server with 2x, 4x or 8x RTX 5090 or RTX 4090. Cinema 4D, Redshift and Octane are pre-installed, multi-GPU servers are billed per minute, and the first top-up of $50 or more gets a 100% bonus. Plans and prices are on our Cinema 4D Redshift render farm page.
FAQ: CPU vs GPU rendering
Is GPU rendering always faster than CPU rendering?
For GPU renderers such as Redshift, Octane and Cycles, a modern GPU is usually much faster when the scene fits in VRAM. CPU-only renderers such as Corona cannot use the GPU for rendering at all.
Does Redshift work on CPU?
Yes. Redshift can render on the GPU, on the CPU, or on both in hybrid mode. Maxon recommends GPUs because they are usually many times faster.
Do multiple GPUs add up their VRAM?
No. Each GPU keeps its own copy of the scene, so the limit is the VRAM of one card: 24 GB on an RTX 4090 and 32 GB on an RTX 5090.
Can Octane render on the CPU?
No. OctaneRender needs an NVIDIA CUDA GPU to render.
How many GPUs do I need for Cinema 4D animation?
Start with a 4x GPU server for most animations, and use 8x or several servers in parallel when the deadline is short.
