How to configure your Arnold workstation
Arnold is a ray-tracing renderer program from Autodesk for calculating three-dimensional, computer-generated scenes and has become one of the most widely used rendering programs in professional film production worldwide in recent years. Over 300 leading studios and production houses use Arnold as their standard renderer. Autodesk Arnold is part of Autodesk Maya and Autodesk 3ds Max as an interactive renderer. From version 6 Arnold is able to use the CPU and GPU for its renderings. Plug-ins are available for the following programmes: Maya, 3ds Max, Houdini, Cinema 4D, Katana and Softimage.
The optimal CPU for Autodesk Arnold
Up to version 5, Arnold is a CPU-based rendering program. For these versions we recommend to pay attention to a balance between the number of their CPU cores and threads. The CPU renderer uses all available threads of the CPU. CPU cores refer to the cores that actually physically exist, while threads are often thought of as queues for the core in question. Multiple queues per core allow the CPU core to switch between different tasks. Intel Xeon processors typically have more threads than their Intel Core counterparts. We recommend focusing on the CPU up to version 5, as the graphics card cannot be used for its renderings. From version 6 Autodesk Arnold is able to use both CPU and GPU rendering. CPU rendering focuses on accuracy while GPU rendering is optimised for speed. The following processors are particularly suitable for Autodesk Arnold: Intel Xeon W-2155, 10 cores, 20 threads and max. 4.50 GHz, Intel Core i9 10900X, 10 cores, 20 threads and max. 4.50 Ghz. More than an alternative are the AMD processors, preferably Threadripper of the Pro series, e.g. model 3955WX with 16 CPU cores, 32 threads and max. 4.3 Ghz. For extremely fast CPU renderings, we recommend a business workstation with dual CPU. We recommend activating the function *Hyperthreading* in the BIOS for Intel processors. With active HT, Arnold is able to render up to 20% faster.
How much memory for the Arnold renderer 7.0
The memory requirements depend primarily on the complexity of your scenes to be rendered. Our hardware recommendations for programs such as Maya, Cinema 4D or 3ds Max serve as a basis. A typical rule of thumb is to provide at least 2GB RAM for each physical core. We always equip our business workstation with 4 identical memory modules in quad-channel mode to provide the optimal performance. For Intel Xeon processors we always use memory with the addition ECC. Modules with ECC function (Error Correction Code) are able to detect and correct memory errors automatically, system crashes are prevented. Pay attention to the number of memory banks in order to be able to upgrade the RAM without problems if required. Business Workstation have between 8-16 memory banks.
The right graphics card for Arnold 7.0
From version 6 you are able to switch between CPU and GPU rendering. Arnold is optimized for NVIDIA RTX series graphics cards with CUDA cores. Take advantage of the new technology to improve your rendering performance. With the introduction of Arnold 7.0, up to 33% GPU memory can be saved when rendering scenes with many geometry divisions. We can recommend the following graphics cards for optimal rendering results: NVIDIA Quadro RTX A4000 with 16 GB DDR6 VRAM, NVIDIA Quadro RTX A5000 with 8,192 Cuda cores, or an NVIDIA Quadro RTX A6000 with 48 GB VRAM. NVIDIA Geforce series graphics cards, the RTX 3070Ti, RTX3080Ti, and RTX3090 models especially demonstrate their power in Arnold 6 by combining RTX-accelerated rendering and GPU-accelerated AI denoising for fast, interactive rendering. The result is noise-free images with very high fidelity. RTX graphics cards are twice as fast compared to NVIDIA's Pascal series graphics cards due to faster GDDR 6 VRAM. A cost-effective alternative is the AMD Radeon Pro W6800 with 32GB GDDR6 VRAM.
The optimal hard drive configuration for Arnold 7.0
We recommend installing the operating system and your programmes on an M.2 NVME SSD (Non-Volatile Memory Express). NVME SSDs are connected directly to the mainboard via a PCI Express connection, make optimum use of the multi-core CPUs of modern business workstations and ensure optimum speed when reading and writing your data. The full speed advantage is achieved through the NVMe protocol for data transfer. With PCIe 2.0, 5GB/s can be transferred per lane. NVME SSD from Samsung models of the 970 Evo and Pro series can reach speeds of up to 3,500 MB/s via PCIe 3.0, provided their mainboard supports the connection of an M.2 SSD via PCIe 3.0. For their project data, we recommend a 2 or 4TB S-ATA SSD for cost reasons.