Search is based on keyword.
Ex: "Procedures"
Do not search with natural language
Ex: "How do I write a new procedure?"
Contact Support
System and Hardware Requirements
Note: Mari increases its level of performance with newer, more advanced hardware configurations. However, Mari is tested and capable of operating on many older, prior-generation systems. For this reason we are listing below-minimum requirements, recommended, and on which tests have been performed. Your particular needs may vary from that of other users.
System Requirements
Officially Supported Operating Systems
• Mac OS X 10.10 (Yosemite) or higher
• Windows 7 64-bit or higher
• Linux 64-bit operating system (CentOS/RHEL 6)
Minimum Hardware Requirements
• Quad-core processor
• 10+GB disk space available for caching and temporary files
• At least 4GB RAM
• Display with 1680 x 1050 pixel resolution
• An NVIDIA or AMD* graphics card with the latest drivers
• 1GB of graphics memory
• OpenGL 3.2* or higher
*Displacement preview is currently only available on the cards and drivers that support OpenGL 4.0 or newer.
Recommended System Requirements
• 2.5+Ghz Quad-core processor
• 250+GB disk space available for caching and temporary files. SSD is preferable.
• 16GB RAM with additional virtual memory*
• Display with 1920 x 1080 pixel resolution
• An NVIDIA or AMD* graphics card with the latest drivers
• 2+GB of graphics memory
• OpenGL 4.4 or higher support
*The use of virtual memory improves stability and helps prevent data loss on large projects.
Recommended does not guarantee that it meets your particular needs.
Tested Workstation Hardware
The configurations listed below are those that Foundry have tested with Mari. Due to the constantly changing nature, and wide variety, of computer hardware available in the market, Foundry are unable to officially certify hardware. The below can be used as a recommendation and does not guarantee that it meets your particular needs.
Please download and install the latest graphics driver for the NVIDIA or AMD websites, and ensure that you are using 8.982.1 drivers or higher for AMD cards.
If you encounter any issues, please contact Customer Support directly through the Support Portal at the following address: https://support.foundry.com.
Tested Workstation Hardware |
|
---|---|
Apple |
• MacPro4,1 • MacPro6,1 • iMac13,2 • MacBookPro9,1 • MacBookPro10,1 |
HP |
• HP XW6600 • HP Z600 • HP Z620 • HP Z640 |
Tested GPU Hardware |
|
---|---|
AMD Prosumer Graphics Cards |
• AMD Radeon HD - D500 |
NVIDIA Prosumer Graphics Cards |
• NVIDIA Geforce GTX 660TI • NVIDIA Titan X • NVIDIA Geforce GT 650M • NVIDIA Geforce GT 675M |
NVIDIA Enterprise Graphics Cards |
• NVIDIA Quadro 4000 • NVIDIA Quadro K4200 • NVIDIA Quadro K5000 • NVIDIA Quadro M6000 • NVIDIA Quadro M6000 24GB |
What Hardware to Use for Mari
This section introduces which Mari features use which pieces of hardware. The following are hardware suggestions to improve Mari performance.
Article: Take a look at the Optimising your Mari projects for best performance article.
GPU - Speed and number of cores
The GPU is used for rendering and also for baking out results to textures. Faster GPUs can render a heavier scene at a better frame rate and shorten the waiting time for baking to textures, such as flattening layers, merging layers, and converting procedural to paintable.
GPU - Memory
The more GPU memory you have, the easier it is to paint more details in general. The two main features that use a lot of GPU memory are the following:
• Paint Buffer - If you have a lot of GPU memory, you can set the Buffer Size (paint buffer's resolution) higher and/or set the Color Depth (bit depth) of the paint buffer higher (16- or 32-bit instead of 8-bit). A higher resolution paint buffer allows Mari to put more details without having to repeatedly zoom, paint, and bake. If the paint buffer is set to a higher Color Depth, it prevents stepping especially for displacement maps.
Note: To configure the paint buffer, see Configuring the Paint Buffer
• Virtual Texture Atlas - Mari uses virtual texturing to contain a large texture data into GPU Memory. However, the virtual texture is limited. If Mari can't process all of the data in a timely manner, it starts using lower resolution mipmaps.
If you have a lot of GPU memory, you can set the Resolution (virtual texture size) to a high resolution so that Mari can render a really heavy scene with, for instance, a scene with lots of layers, UDIMs, and/or fragmented UV bits.
Note: To set the Resolution, see Mari Preferences Dialog
Article: Take a look at the Calculating your GPU memory usage for Virtual Textures in Mari article.
CPU
In general, a moderate quad-core processor should be 'good enough', but some non-GPU operations benefit from more cores or faster CPUs. The following are examples of non-GPU operations in Mari:
• Ambient occlusion calculation
• Whole patch bleed
• Tile level bleed after baking
• Changing bit-depth of a channel
• Changing resolutions of textures
RAM
4GB of RAM is sufficient to run Mari, but a RAM of 8GB is better for more stable operations especially if you're running other 3D applications. If you want to work with heavy scenes, it's recommended to have as much RAM as possible.
Ultimately, all data in Mari is cached to disk, so even if you don't have a lot of RAM, Mari still runs properly, but there is more disk reading. The areas where Mari uses a lot of RAM are:
• General application (UI, application logic, and so on), and
• The texture data loaded into the RAM from a disk stays in the RAM, but is removed from the RAM in Least Recently Used (LRU) cache.
Disk
A solid-state disk (SSD) is highly recommended. Lengthy operations are often bottle-necked by disk writing so having an SSD greatly helps reduce the time taken. Regardless of the data processed by CPU or GPU, the resulting data is ultimately written to disk.
If your project is light, for example, it uses a few UDIMs with up to 4k textures, an SSD may not make a big difference.
Note: Scratch space has not been tested yet. Mari manages its own data by keeping the recent data in RAM using a Least Recently Used (LRU) algorithm while all data is written to a disk.