Application note · Specialised mathematical computing

Application Note: Monte Carlo and Optimisation with CPU, MPI and GPU Execution

Choose threads, processes, MPI or CUDA by algorithm; preserve random streams and report efficiency, uncertainty and completed scenarios.

Monte Carlo is often embarrassingly parallel, while optimisation may mix independent candidates with coupled linear algebra. This note selects the simplest efficient execution layer and keeps stochastic evidence reproducible.

Correctness baseline

Implement a small deterministic or fixed-seed reference and verify statistical outputs, confidence intervals and constraints. Assign non-overlapping seed streams by scenario/worker and record the mapping. Reproducibility means equivalent statistics and traceable random inputs, not always bitwise identity across libraries.

Single-node scale

AxiKeystone provides 128 cores and 1.15TB RAM for process pools and in-memory datasets. Measure one worker, then increase concurrency while tracking memory bandwidth, serialisation and NUMA locality. Batch short tasks so scheduling overhead does not dominate.

GPU path

AxiFoundry provides a professional 24GB GPU for CUDA-enabled Python/Julia/MATLAB experiments. RAPIDS accelerates supported tabular pipelines. Keep arrays on device across steps and include transfer time in comparisons. Irregular branch-heavy simulations may remain better on CPU.

Two-node path

Use AxiVector when the dataset exceeds one node, the solver exposes MPI or independent work can be distributed with low coordination. Run fixed-problem and growing-problem tests over 100GbE RDMA. Do not accept more workers if efficiency collapses or licence cost outweighs saved time.

Campaign operation

Checkpoint optimiser state and completed scenario IDs. Use a scheduler for fair allocation and failure retry. Report wall time, scenarios/second, parallel efficiency, peak RAM/VRAM, convergence and uncertainty. Archive environment, code, inputs, seeds and selected outputs.

Primary technical references

References are provided for software architecture and implementation planning. Validate the versions, licences, support matrix and regulated-use requirements applicable to the final deployment.

Relevant Hyperion platforms

Hyperion X1140

AxiFoundry

A professional development and proof-of-concept node for CUDA, AI, data science and software engineering.

One 4U node; 96 AMD EPYC cores; 384GB ECC DDR5; one NVIDIA RTX PRO 4000 24GB; enterprise NVMe and remote management.

Hyperion X1140-C

AxiKeystone

A CPU- and memory-led HPC platform for simulation, genomics, analytics, compilation and workloads with large in-memory working sets.

One 4U node; 128 AMD EPYC cores; 1.15TB ECC DDR5; one NVIDIA RTX PRO 4000 24GB; enterprise U.2 NVMe.

Hyperion X2160

AxiVector

A tightly coupled two-node simulation and mathematical-computing platform for MPI, CFD, FEA and optimisation.

Two 4U nodes; 256 AMD EPYC cores; 2.3TB aggregate ECC DDR5; two 96GB GPUs; direct 100GbE RoCEv2 RDMA between nodes.

Configuration and quotation

Validate this workload on Hyperion

Final architecture and price depend on representative code and data, concurrency, storage, networking, site infrastructure, component availability and export compliance.

Request Formal Quotation