You don’t need to worry about specifying particular settings to optimize the GPU card for InfoWorks ICM Simulations. The card will simply go just as fast as it can for any given 2D simulation.
InfoWorks ICM provides confirmation that the GPU is active during the 2D Simulation. This can be found on the last line of the “Job Progress” tab.
The CUDA code in the 2D engine of InfoWorks ICM does not allow direct monitoring of the GPU from within the ICM Software itself, but there are two other routes you might like to explore if you are keen to monitor the GPU card performance during InfoWorks ICM 2D Simulations.
Your choice will depend on the GPU card you are utilising and the Operating System (OS) of your computer.
The first option is a utility within the latest drivers for the K Series Quadro and TESLA cards. This allows you to visualise the utilisation rate during a simulation. The example screenshot below is from a pure 2D ICM run I did with a Quadro K5200 GPU card. You’ll see the GPU card is running at 98% capacity.
Another option is to utilize a desktop gadget to monitor the GPU. Desktop gadgets are only officially supported in Windows 7, but if you are running Windows 8 or the new Windows 10, then there is a utility to re-enable the Gadget code in the newer operating systems (if you do this in Windows 8 or Windows 10 the Gadgets will remain unsupported of course, but once enabled they work just as they do in Windows 7). One gadget I particularly like is the GPU Meter available from www.addgadget.com. I find the “All CPU Meter” (available from the same site) is also a very valuable tool to monitor the performance of the motherboard CPU during regular 1D simulations.
If you can’t utilize the Desktop Gadgets to monitor GPU performance, then you could consider an application such as “GPU-Z” for monitoring, which is a good alternative. See http://www.techpowerup.com/gpuz/ for more details. The screenshot below shows the monitoring of the Quadro K1100M in a laptop PC during a 2D simulation.
Finally, do remember that the GPU is only utilized for the 2D calculations, with all the 1D mathematics being conducted on the motherboard CPU. For 1D/2D models you will only see the GPU being used for a percentage of the overall simulation time. The better (faster) GPU cards will always show a lesser percentage of the overall simulation activity in 1D/2D simulations because they can perform the 2D calculations they are assigned in a shorter timeframe than slower cards and the main CPU. The only true test of different GPU cards is with models that are pure 2D models. Innovyze can provide a standard GPU testing/comparison dataset for those who would like to contract and compare the performance of different NVidia GPU cards.