This document offers a comparison between the V-Ray and V-Ray GPU render engines.
Hardware Utilization | V-ray and V-ray GPU
Rendering with V-Ray uses the CPU device. Features like the V-Ray Denoiser and the V-Ray Lens Effects can utilize the GPU independently of the renderer, as they are applied as a separate post-process. To render on a CPU, your PC should meet the minimum System Requirements.
This engine takes advantage of the graphics card, but it relies on the CPU to feed it with tasks. This means that CPU performance also impacts GPU rendering.
Hybrid rendering (CPU + GPU)
Hybrid rendering is a mode of the V-Ray GPU engine that utilizes both devices. This way, the rendering performance benefits from all the available hardware on your workstation.
The advancement of GPU hardware increases performance at a lesser cost.
The more popular it becomes, the more users test the performance of the two V-Ray engines by simply switching from one engine to the other with the same scene and comparing the results. This approach is inaccurate. The two engines work differently and execute differently.
Therefore, it is recommended to use scenes that are optimized for the specific renderer. Some scenes are expected to render faster with GPUs compared to the CPU, while others will show comparable performance results.
Chaos offers V-Ray Benchmark – a performance measurement tool that assesses the hardware’s speed specifically when working with V-Ray. The benchmark includes two test scenes: one for V-Ray and another one for V-Ray GPU. It can quickly and easily evaluate a user’s machine capabilities when running V-Ray and V-Ray GPU. The results can be compared with other users’ data and help when looking at specific hardware.
A huge advantage of V-Ray GPU is that it utilizes multiple graphic cards installed on a single machine. Setting up additional GPUs is effortless and their pricing is reasonable, in contrast with CPUs installed on a multi-socket motherboard. This also scales the performance almost linearly with GPUs. For example, with two identical GPUs, a user may expect to see the render times to be twice as fast, compared to a single GPU.
The V-Ray engine runs on NUMA configurations (dual CPU machine). However, this may require a substantial upgrade to the whole PC. In contrast, a second GPU can be effortlessly added if your motherboard has compatible PCI slots.
Both render engines benefit from the latest features for performance improvements like Adaptive Lights and the Adaptive Dome Light.
The V-Ray GPU engine utilizes the RT Cores in NVIDIA RTX GPUs (Turing architecture). The RT Cores accelerate the ray-tracing calculations and can speed up the rendering by an average of 40%.
The time needed for rendering start is less for V-Ray, while the performance gain offered by V-Ray GPU compensates it in the rendering progress.
V-Ray supports more features than V-Ray GPU. The main reason for this is the difference in the engines’ codebase.
However, it’s important to outline that the GPU renderer is fully production-ready and is already equipped with all required features.
Faster Feedback, Faster Decisions
With the performance gain in the GPU used engine, you can expect to receive faster feedback from a project.
This allows you to make more decisions and be more creative while iterating for the same amount of time. This performance advantage also indirectly affects your overall rendering quality since more time is spent on creative choices, instead of rendering.
A major difference when comparing these engines is the utilized memory. V-Ray makes use of the CPU memory (RAM), while V-Ray GPU utilizes the GPU memory (VRAM). GPU technology is still far behind in terms of memory.
Common workstation configurations today have 32, or 64 GBs of RAM. At the same time, a common graphics card used for GPU rendering has 8 or 11 GBs of VRAM.
When using the CPU and RAM, it is possible to add more memory, given a free slot is available on the workstation’s motherboard.
You can also attach an additional unit to your GPU(s). However, this does not automatically mean doubling the cards combined VRAM. Only High-end NVIDIA GPUs offer this capability with a feature called NVlink that allows bridging the memory of 2 identical graphics cards.
V-Ray GPU supports NVlink connections, which allows almost doubling the amount of GPU memory on a workstation. The latest version of V-Ray comes with an initial implementation of out-of-core rendering. It transfers triangle meshes to the CPU memory when the GPU runs out of memory, allowing to render scenes with a lot of geometry.
More in-depth information can be found in the Hardware Recommendations article.
Network Rendering and Chaos Cloud
Both are available for network rendering using Backburner, Deadline, and other render managers.
V-Ray comes with its own Distributed Rendering feature that allows splitting a single render job within a single frame across many computers in a network.
Chaos Cloud is Chaos’s proprietary rendering service and it supports both types of engine.