As an administrator, you will be able to identify hardware requirements for machines running ICM.
Hardware requirements vary, depending on the components and types of simulations being run on a machine.
See the Autodesk support page System Requirements for InfoWorks ICM as a general guide.
Windows 10 or 11 laptop or desktop machine can run the ICM Workgroup Client. In a multi-user environment, you can dedicate specialized simulation machines.
Performance depends primarily on the clock speed of the processor. Memory and disc performance also contribute and become more important as the number of simulations running concurrently increases.
Can run multiple simulations concurrently on a single machine. This can be adjusted in the Agent Manager, from the Agent Options dialog.
In this graph of a single simulation run on a different number of cores, you see a reasonable improvement in speed from two to four cores, but less after that:
Individual simulations perform well with four cores or threads, meaning that 12-core CPUs can handle running three simulations concurrently.
Can contain hundreds of thousands or millions of elements.
Performance is improved with graphics card with good single-precision performance.
Memory bandwidth and size are also important—determines maximum size of the model.
NVidia CUDA enabled GPUs with a compute level of 3.5 or higher are currently supported for 2D calculations.
Models with 1D and 2D objects require performance CPU and GPU hardware.
GPU hardware can significantly decrease calculation time (from hours to minutes).
When simulating on the Cloud, hardware is selected from available machine catalogue, based on composition of the model and compatibility, to optimize performance.
Simulation results may require terabytes of data.
Solid state drives are recommended, especially for locations that will be directly accessed.
If centralized storage is used, it is recommended that the environment be easily scalable to cope with demand increases.
Vital for networks.
ICM can place a high burden on the network when transferring large models and results.
Ensuring connection stability and bandwidth amongst the components is essential to avoiding poor performance and user issues.
