Benchmarking Stan under Dynamic Overclocking with ThrottleStop

This repository holds digital assets associated with the article "Benchmarking Stan under Dynamic Overclocking with ThrottleStop" [1]. That article discusses the impact of dynamic overclocking on benchmarking the Stan probabilistic programming language. The ThrottleStop CPU throttling utility was used to moderate overclocking for an example Bayesian inference task in astronomy.

---

Figure 1. Example runtimes for Stan's Bayesian inference with and without system overclocking. Inference of type Ia supernovae cosmological parameters was used as an example problem. Runtime variance was lower under overclocking. Hyperthreading was counterproductive. Execution configurations, moderated by ThrottleStop, were...

   Top (left) overclocking, physical cores, (right) no overclocking, physical cores.
   Bottom (left) overclocking, logical cores, (right) no overclocking, logical cores.

Adapted from [1].

---

Table of Contents

• Key Files
• Software Requirements
• Getting Started
• Acknowledgements
• References

Key Files

File	Notes
src/stan-overclocking-benchmarker.Rmd	R Markdown. Incorporates Stan's R interface, RStan.

Software Requirements

Software	Notes
Ghostscript 10.00.0	Available here. Free and fee-based options.
PowerShell	Free. Included with Windows 11.
R	Available here. Free.
RStudio	Available here. Free and fee-based options.
ThrottleStop	Available here. Free.    Recommended. The code was designed for use with ThrottleStop, but will run without it.
Windows 11	Supported operating system.

Ghostscript

The benchmarker assumes Ghostscript is available on the system. An invocation path should be set in the RStudio environment variables GS_CMD and R_GSCMD. I.e., the path to a relevant executable such as gswin64c.exe.

Note

The code was not designed for Ghostscript more recent than version 10.00.0.

Stan

Please ensure the R environment has the RStan package installed.

R and RStudio should be configured to enable execution of RStan code. For more information, see: The R Interface to Stan.

Getting Started

The file stan-overclocking-benchmarker.Rmd should be run from RStudio, with administrator privileges. E.g., invoke RStudio via the Windows "Run as administrator" option.

Acknowledgements

This work was supported by the Australian Research Council Training Centre in Data Analytics for Resources and Environments (project ICI9010031).

References

1. T. N. Stenborg, "Benchmarking Stan under Dynamic Overclocking with ThrottleStop", in Astron. Data Anal. Softw. Syst. XXXII, vol. 538, S. Gaudet, D. Bohlender, S. Gwyn, A. Hincks and P. Teuben, Eds., Astronomical Society of the Pacific, 2025, pp. 265–268, doi: 10.26624/GAQB5323.
   View PDF   View at publisher   SciX