GPU Cluster for Scientific Computing

1
GPU Cluster for Scientific Computing Zhe Fan,
Feng Qiu, Arie Kaufman, Suzanne
Yoakum-Stover Center for Visual Computing and
Department of Computer Science, Stony Brook
University http//www.cs.sunysb.edu/vislab/projec
ts/gpgpu/GPU_Cluster/GPU_Cluster.html
and Large-Scale Simulation

• Stony Brook Visual Computing Cluster
• GPU Cluster
• 35 nodes with nVIDIA GeForce FX 5800 Ultra
• Gigabit Ethernet
• 70 Pentium Xeon 2.4GHz CPUs
• 35 VolumePro 1000
• 9 HP Sepia-2A with ServerNet II

• Scale up LBM to the GPU Cluster
• Each GPU computes a sub-lattice
• Particles stream out of the sub-lattice
• Gather particle distributions in a texture
• Read out from GPU in a single operation
• Transfer through GigaE (MPI)
• Write into neighboring GPU nodes
• Network performance optimization
• Conduct network transfer while computing
• Schedule to reduce the likelihood of interruption
• Simplify the connection pattern

• Times Square Area of NYC

• 1.66 km x 1.13 km
• 91 blocks
• 851 buildings
• 480 x 400 x 80 lattice

• LBM on the GPU
• Application large-scale CFD simulations using
  Lattice
• Boltzmann Model (LBM)
• LBM Computation
• Particles stream along lattice links
• Particles collide when they meet at a site
• Map to GPU
• Pack 3D lattice states into a series of 2D
  textures
• Update the lattice with fragment programs

• GPU Cluster / CPU Cluster Speedup
• Each node computes an 80 x 80 x 80 sub-lattice
• GeForce FX 5800 Ultra / Pentium Xeon 2.4GHz

• Acknowledgements
• NSF CCR0306438
• Department of Homeland Security, Environment
  Measurement Lab
• HP
• Terarecon

• Dispersion Plume