This is the third instalment in a series of blogs I’ve written summarising run times for 2D simulations conducted on PC hardware and NVIDIA GPU cards. The first was in December 2012, I then followed up with a second in September 2016.
NVIDIA continue to evolve their GPU technology, producing faster and more powerful cards in rapid succession, often making it quite hard to keep up with the constant stream of new cards coming to the market! The other major change within the last 12-18 months has been a move away from physical hardware in preference to the adoption of Virtual (Cloud based) hardware platforms within corporate IT environments.
InfoWorks ICM is designed to leverage technological improvements as soon as new GPU cards and new hardware platforms come onto the market. The tables below show runtimes for two of our standard 2D testing models which were run on a high-end Workstation and a high-end physical Server equipped with two of the latest NVIDIA GPU cards, and a Cloud based system comprising of virtual Xeon processors and a virtual GPU card (NVIDIA GRID technology).
As with the previous runtime tests, the models contain 1.6 million and 3.3 million 2D elements and are 2D only datasets (i.e. no pipes or manholes). Each model is run for a 2hr simulation using a M5-60 design storm. Rainfall is applied directly to 100% of the mesh, so all elements in the 2D zone are wet throughout the simulation (this is important, because it’s the number of wet elements that most influences the runtime, not the overall number of 2D elements in the model).
My thanks go to John Baxter of AC&E Ltd, who kindly supplied the physical hardware and Cloud System we used to conduct the test runs with our standard Innovyze GPU Testing Database, the results of which are shown in the tables above. If you would like more details of the physical hardware or virtual environment employed to conduct these tests, you can contact John Baxter of AC&E using the details shown below.