Issue link: https://resources.randsim.com/i/1504494
4 ©2023 ANSYS, Inc. / Confidential GPU Solver: Sliding interfaces Model rotating and stationary components • An efficient and robust sliding mesh algorithm • Enhanced robustness of dramatically changing intersections • Test case: F1 140M rotating wheels, URANS SST 277 399 100% 72% 0% 20% 40% 60% 80% 100% 0 50 100 150 200 250 300 350 400 450 8 16 Parallel efficiency GPUs F1 Racecar Sliding Mesh 140M GPU Scalability 10 20 36 100% 97% 88% 0% 20% 40% 60% 80% 100% 0 50 100 150 200 250 300 350 400 450 4 8 16 MIUPS Nodes (44 cores each) F1 Racecar Sliding Mesh 140M CPU Scalability Intel Platinum 8168 dual-socket cluster Nvidia A100 40GB 8-GPU per node 1 GPU ~ 16 CPU nodes (704 cores) With 8 GPUs, solve 1000 steps in 40 minutes!