Metashape can be run on Windows, Linux and macOS systems. There is no recommendation regarding the OS. The processing performance on different OS should be almost the same, we do not expect any benefit from using one OS over another. Some minor deviations may be observed however, depending on the OS specifics and resources access and utilization.
Metashape provided native support to Apple Silicon (M1, M2, M3, M4) including GPU acceleration.
Minimal configuration | Recommended configuration |
| • Windows 7 SP 1 or later (64 bit), Windows Server 2008 R2 or later (64 bit), macOS High Sierra or later, Debian/Ubuntu with GLIBC 2.19+ (64 bit) • Intel Core 2 Duo processor or equivalent • 8 GB of RAM | • Windows 7 SP 1 or later (64 bit), Windows Server 2008 R2 or later (64 bit), macOS Mojave or later, Debian/Ubuntu with GLIBC 2.19+ (64 bit) • Intel Core i7 or AMD Ryzen 7 processor (3.0+ GHz) • Discrete NVIDIA or AMD GPU (4+ GB VRAM) • 32 GB of RAM |
Minimal configuration is suitable for small projects (50-100 images).
It is very important to pay attention to the computer's cooling system. Since processing may take a long time and heat up computer components quite strongly, it is therefore very important that the computer has a good cooling system.
Confirm that system and CPU temperature under heavy load is reasonable (like less than 80-90 C degrees).
Storage
During processing, it is recommended to store Metashape project files (including source images) on an SSD.
If processing is performed on a cluster, then the project and images should be stored on a fast NAS connected to at least 1 Gbps network (preferably more).
RAM recommendations
RAM allows to process bigger number of images in the single chunk and/or to achieve higher reconstruction quality, since the model generation stage may be very resource demanding.
The amount of RAM does not affect the processing speed, but defines for some processing stages the maximum number of images that can be processed in the same project.
Approximate memory requirements for common workflow stages depending on the reconstruction parameters used and number of source images can be checked in the following article: Memory requirements for processing operations.
The amount of RAM: usually we recommend to install 32 GB or 64 GB RAM (for desktop systems), if the processing tasks requires high reconstruction quality or large amount of images.
The RAM frequency does not have a significant effect on the processing speed, but a high frequency may cause unstable system condition, that could lead to unexpected termination of the processing operation.
The memory modules must be from the same set and operate at a frequency no higher than the declared supported one (for some processors and motherboards, it may be necessary to further reduce the frequency for stable operation). If two identical sets are purchased (for example, 2x16GB + 2x16GB with a declared frequency of 3600 MHz), then stable operation is often achieved only by further reducing the frequency of the memory, because the more modules are there, the less stable they are.
You can check RAM for instability through the multithreaded Memtest86 and through all four stress tests in the Intel Extreme Tuning Utility.
How to decide on amount of RAM to be installed to the working computer?
For most common processing operations 32 GB RAM should be enough, but we could recommend to have 64 GB RAM to be on the safe side, in case extra large project should be processed, or at least keep the possibility of later upgrade by installing additional RAM modules.
CPU recommendations
CPU effects on the processing speed on every processing stage. Higher frequency and a bigger number of cores will result in faster processing.
Both multithreaded and single-threaded CPU performance are important, because at some intermediate stages only a few cores could be used for the short period of time.
Usually we recommend CPUs with 6 - 32 physical cores with the highest possible base frequency (at least 3.2 GHz) as optimal desktop solutions for Metashape processing.
In case the processing tasks require 128+ GB RAM we can suggest to use HEDT configurations. But same recommendations regarding reasonable number of cores and high base frequency also apply.
You can check the following benchmark page for some comparative information on different CPUs:
https://www.pugetsystems.com/labs/articles/Agisoft-Metashape-Performance-Intel-Core-10th-Gen-vs-AMD-Ryzen-3rd-Gen-1765/
How to compare CPUs?
We recommend choosing the option with the base frequency 3.0 GHz and higher and with at least 8 CPU cores.
Also, using the test results in Cinebench will help in the assessment (it is important to use the same version). In this case, it is important to consider the single-threaded and multithreaded result together.
Metashape supports processing on a CPU only without a GPU, but the stages that support GPU acceleration will take considerably longer time to complete, so we highly do not recommend using such computers for processing operations with GPU-support.
GPU recommendations
- image matching (first step of Align Photos stage),
- depth maps calculation (sub-step of Build Point Cloud stage and depth maps based Build Model / Tiled Model operations),
- depth maps based mesh, DEM, tiled model generation,
- texture blending (second step of Build Texture stage or corresponding sub-steps of Build Tiled Model stage).
Utilization of discrete GPU improves the speed of these processes significantly. Usually we recommend high-end gamer class cards. Professional class graphic cards are also supported, but are usually not reasonable in terms of performance/price ratio and due to the fact that Metashape does not utilize double-precision computations.
It is recommended to use only discrete graphics cards (NVIDIA, AMD, Intel). Metashape supports processing integrated graphics cards, but they are not designed for long and resource demanding operations, so it is recommended to avoid integrated graphics utilization, if at least one discrete GPU is available.
Different GPU models can be used at the same time. It is recommended to use video cards from the same manufacturer together and with no more than two times the difference in TFLOPS and with no more than one and a half times the difference in Memory Bandwidth (GB/s).
For multi-GPU systems we recommend to uncheck "use CPU" option in the GPU tab of Metashape Preferences window for better performance on GPU-supported stages. Also to uncheck integrated graphics.
Apple Silicon graphic processor option should be checked on.
How to compare GPU?
Compare graphics cards, considering the following parameters:
- Amount of VRAM - if you choose between 6 GB and 8 GB, we recommend selecting 8 GB (basically, the option with higher amount of VRAM is better, this will allow you to display large detailed models without problems).
- Compare the theoretical TFLOPs value for single precision computations (here higher value will also result in higher Metashape performance for GPU-supported tasks).