Massive-Scale Simulations of 2D Ising and Blume-Capel Models on Rack-Scale Multi-GPU Systems

Abstract

We present high-performance implementations of the two-dimensional Ising and Blume-Capel models for large-scale, multi-GPU simulations. Our approach takes full advantage of the NVIDIA GB200 NVL72 system, which features up to $72$ GPUs interconnected via high-bandwidth NVLink, enabling direct GPU-to-GPU memory access across multiple nodes. By utilizing Fabric Memory and an optimized Monte Carlo kernel for the Ising model, our implementation supports simulations of systems with linear sizes up to $L=2^{23}$, corresponding to approximately $70$ trillion spins. This allows for a peak processing rate of nearly $1.15\times 10^5$ lattice updates per nanosecond-setting a new performance benchmark for Ising model simulations. Additionally, we introduce a custom protocol for computing correlation functions, which strikes an optimal balance between computational efficiency and statistical accuracy. This protocol enables large-scale simulations without incurring prohibitive runtime costs. Benchmark results show near-perfect strong and weak scaling up to $64$ GPUs, demonstrating the effectiveness of our approach for large-scale statistical physics simulations.

18 pages, 4 tables, 5 figures, version accepted for publication in Computer Physics Communications

References in corpus (30)

• Theory of Phase Ordering Kinetics
• Worm algorithms for classical statistical models
• Tricritical universality in a two-dimensional spin fluid
• Multi-GPU Accelerated Multi-Spin Monte Carlo Simulations of the 2D Ising Model
• Pre-Cooling Strategy Allows Exponentially Faster Heating
• Exact results for curvature-driven coarsening in two dimensions
• The Cluster Processor: New Results
• Multicritical Points and Crossover Mediating the Strong Violation of Universality: Wang-Landau Determinations in the Random-Bond $d=2$ Blume-Capel model
• Strong Violation of Critical Phenomena Universality: Wang-Landau Study of the 2d Blume-Capel Model under Bond Randomness
• Performance potential for simulating spin models on GPU
• Slave Boson Theory of Orbital Differentiation with Crystal Field Effects: Application to UO${}_2$
• Janus II: a new generation application-driven computer for spin-system simulations
• First-order phase transition and tricritical scaling behavior of the Blume-Capel model: a Wang-Landau sampling approach
• Simulating spin systems on IANUS, an FPGA-based computer
• Simulating spin models on GPU
• GPU accelerated population annealing algorithm
• Scaling and universality in the phase diagram of the 2D Blume-Capel model
• Aging rate of spin glasses from simulations matches experiments
• Parallel multicanonical study of the three-dimensional Blume-Capel model
• SUE: A Special Purpose Computer for Spin Glass Models
• Shortcuts of freely relaxing systems using equilibrium physical observables
• Testing statics-dynamics equivalence at the spin-glass transition in three dimensions
• A Performance Study of the 2D Ising Model on GPUs
• FPGA Hardware Acceleration of Monte Carlo Simulations for the Ising Model
• Connected component identification and cluster update on GPU
• Slow growth of magnetic domains helps fast evolution routes for out-of-equilibrium dynamics
• Wang-Landau study of the triangular Blume-Capel ferromagnet
• Optimized GPU simulation of continuous-spin glass models
• A new test for random number generators: Schwinger-Dyson equations for the Ising model
• The QISG suite: high-performance codes for studying Quantum Ising Spin Glasses

Cited by in corpus (1)

• Universal exotic dynamics in critical mesoscopic systems: Simulating the square root of Avogadro's number of spins