Skip to content
New: a 2.5 million member pension buy-in, priced in 41 seconds on one machine.

Research

Measured, not claimed.

Every number on this site comes from a run you could repeat. The machine, the comparison and the source are written down beside it.

How it works

Every future, strung from today to the horizon.

A simulation follows thousands of possible futures from the same starting point. Henselian computes them all at once, so the answer arrives while you are still asking.

Slide the gate to compute the futures it passes.

A UK pension buy-in, priced member by member

1.5 trillion member-paths in 41 seconds

  • 2.5 million members on 600,000 paths of longevity, gilts and inflation, valued a year at a time.
  • One RTX 4090, end to end. The slowest of three runs took 44.4 seconds.
  • 137 times faster than the same method in hand-optimised C on all 32 threads of the same machine.
  • The graphics card and the reference agree to every printed digit on the deterministic case.
  • Synthetic scheme, illustrative mortality basis.

Nested stochastic variable annuities

28,000 times the SOA's reference

  • The Society of Actuaries' nested-stochastic valuation on the full 190,000-policy Gan and Valdez portfolio.
  • 0.0283 microseconds per policy-scenario on one RTX 4090, against 789.7 for the reference on the same machine.
  • 52 times faster than the best CPU code we wrote, fused Julia on all 32 threads.
  • Cross-checked against the Julia kernel to 2.3 parts in a million.

Measured, not promised

What took an afternoon now takes a minute.

to price a 2.5 million member pension buy-in across 600,000 possible futures, every member on every one.
41seconds
than carefully hand-tuned code running on every core of the same computer.
137times faster
than the Society of Actuaries' published reference, on its own benchmark.
28,000times faster

Measured on one RTX 4090 workstation. Every source is on the research page.

Check the numbers

Notes.

Short pieces on simulation for decisions. Each one lets you try the idea, and ends with what to take away.

We cannot prevent every storm. We can help you prepare before it arrives.

Bring us the simulation your organisation wishes it could run faster. We will show you what is possible, on a book shaped like yours.

Bring us a problem