V-Ray for 3ds Max is production-proven rendering software. Known for its versatility and ability to handle any type of project — from large dynamic scenes with thousands of lights to magnificent still lifes — it's the solution of choice for artists and designers in the 3D industry.
Known as the world's most comprehensive high-end visualization production 3D rendering software, super match. In today's article, Yunyun will take you to see how rendering speeds scale across multiple GPUs in Vray & 3Dsmax.
Test the settings
Since the new RTX card takes advantage of the RT core in Vray Next, we will test with the powerful RTX 3090 based on the Ampere architecture. It will prove to be the fastest gaming and creator GPU on the market, also because it will take up three PCI slots on your PC. To see how the increased number of graphics cards in the system affects Vray performance with the new GeForce 30 Series (ampere architecture), we will compare rendering times from one GPU RTX 3090 to eight GPU RTX 3090 in Vray 5 and 3Dsmax 2020. In particular, our project size is 4000p x 2200p, 1000 samples, and the noise limit is 0.01.
For our testbed configuration is the AMD Ryzen Threadripper Pro 3955WX. This provides the correct PCI-Express slot layout for up to 8 GPUs, and clock speeds run fast: 4.3GHz base and up to 4.7GHz turbo.
analyse
The new hybrid rendering mode (as of V-Ray 3.6) is only available for CUDA mode that requires NVIDIA graphics. Users who use AMD don't get embree acceleration because that's a reward from NVIDIA. When the engine is set to an RTX-enabled V-Ray GPU, the V-Ray GPU uses the RT cores in the NVIDIA RTX GPU. Rendering speed depends primarily on the number of CUDA cores (the more the better) and the clock speed of the GPU.
Test gpu performance for 3Ds Max and V-Ray on 1/2/4/6/8 GPU RTX 3090
The GPU-accelerated version of V-Ray is capable of taking advantage of multiple GPUs and is very effective. It won't be perfect "4 GPU speeds are 4x", but you can expect a significant performance boost on each graphics card. However, because V-Ray uses these cards for computing purposes, they do not need to be in SLI mode. Just looking at the results of the secondary cards, we can see that some of them will provide excellent rendering performance. Based on tests with the RTX 3090, it scales almost perfectly from 1 -> 2 -> 4 GPUs. In particular, the extension from the second video card is still considerable – in the range of 29% to 57% improvements. This reflects what we've seen in some other GPU rendering engines, such as OctaneRender. That's why we recommend using at least 2 GPUs for V-ray and 3Dsmax for optimal performance.
In particular, a single RTX 3090 takes 26 minutes and 39 seconds to render a still image. This may be an acceptable number for the local computer, but in this test, the dual RTX 3090 reduced the rendering time to 18 minutes and 55 seconds compared to 11 minutes and 38 seconds with 6 GPUs.
Since you added the 6th and 8th graphics cards, you may see that it does not increase linearly, because sometimes it depends on the model you are running. However, for large projects with tight deadlines (9 minutes and 40 seconds and 8 minutes and 2 seconds, respectively), the rendering time is still impressive. But it's still enough to justify their use in multi-GPU workstations and be very helpful for large projects.
NVlink performance
In addition to RT Cores, the new RTX card also supports NVLink, which allows V-Ray GPUs to share memory between the two GPUs; this has some impact on rendering speeds. In this benchmark, Chaosgroup aims to measure it. In order to enable NVLink, these cards need to be connected to a special NVLink connector (also known as an NVLink bridge). In order for NVLink to run on Windows, the GeForce RTX card must be placed in SLI mode from the NVIDIA control panel (the Quadro RTX card does not need to do this, nor does Linux, and is not recommended for older GPUs). If SLI mode is disabled, NVLink will not be activated.
Note that the available memory for GPU rendering is not fully doubled by NVLink; V-Ray GPUs need to copy some data on each GPU for performance reasons, and some memory needs to be reserved on each GPU as a scratchpad for calculations during rendering. Still, using NVLink allows us to render larger scenes than using each GPU alone.
summary
For truly realistic rendering, you need a ray tracer built for quality, optimized for speed, and ready to scale. V-Ray lets you work with lightning-fast interactive and heavyweight product renderings while making the most of your CPU and GPU hardware. Alternatively, you can choose a cloud rendering with high-performance cloud rendering.
With super distributed rendering technology, Yunyun can use massive node elastic expansion to load various rendering environments with one click to meet the completion of various rendering tasks. At the same time, the senior and strong R & D team product team, according to the software, plug-in updates, quickly complete the support service. And constantly research the rendering method that is more in line with the designer's work.