April 25, 2025

[ad_1]

Simulating how bulk and granular supplies work together with tools, containers, and each other is a vital functionality for industrial, manufacturing, and life science organizations. The bigger these simulations, the extra correct they grow to be, reducing the time and expense firms should spend iterating designs and prototypes. Altair and Google Cloud just lately collaborated to see how massive a simulation they might produce utilizing Altair® EDEM™ on a single Google Cloud digital machine. The outcomes have been groundbreaking.

GPUs energy EDEM simulations

Altair EDEM is a high-performance software program software for bulk and granular materials simulation. Powered by discrete ingredient methodology (DEM) calculations, EDEM rapidly and precisely simulates and analyzes the conduct of mined ores, soils, fibers, grains, tablets, powders, and extra.

Over time, industries have been demanding even larger scale from EDEM, leading to quite a few breakthroughs round mannequin measurement. 20 years in the past, the higher restrict was 200,000 particles — a simulation that took upwards of 10 days to create. That rapidly rose to 1 million particles, then 10 million, then 20 million. The holy grail right this moment is to realize a simulation containing 1 billion particles in just a few days.

As you would possibly anticipate, these simulations require an enormous diploma of computational energy, so the utilization of graphics processing models (GPUs) has significantly improved the velocity and effectivity of simulations. GPUs are particularly designed to deal with parallel processing duties, making them extremely environment friendly for dealing with the big quantities of information and sophisticated calculations concerned in EDEM simulations.

The experiment: Altair EDEM on Google Cloud

Designed as a desktop software, EDEM has at all times been restricted to shared-memory architectures, which exist on a single host. This solely permits processors to entry it straight, versus distributed reminiscence, which might scale throughout a number of hosts. With out in depth rebuilding, EDEM can’t benefit from distributed reminiscence’s scalability and suppleness. Nevertheless, multi-GPU programs supply a chance to extend computational energy with no need emigrate towards distributed reminiscence programming fashions.

There have been two targets for the collaboration between Altair and Google Cloud: simulate the biggest system doable, containing one billion particles, and collect information to construct estimates for mapping a given hardware sort to a doable simulation scale. In Could 2023, Google Cloud introduced the provision of its A3 digital machines (VMs) with NVIDIA H100 GPUs. The A3 VMs mix NVIDIA H100 Tensor Core GPUs with trendy CPU, in addition to providing improved host reminiscence and main community upgrades, which made this scale of simulation doable.

Altair and Google Cloud ran two simulation eventualities on a single A3 VM with eight NVIDIA H100 GPUs, every with 80 GB of GPU reminiscence and a complete of three.6 TB/s bisectional bandwidth. The system additionally included a 4th Technology Intel® Xeon® Scalable processor and a pair of TB of host reminiscence.

The take a look at state of affairs chosen was the filling take a look at. It is a sensible simulation used to verify the impression of particle conduct in a practical industrial setting — on this case, particles dropped from a transferring plate right into a container. Two kinds of particles have been used on this simulation: single-sphere and multi-sphere.

Breakthrough outcomes

[ad_2]

Source link