Diamonds Are Forever, But It Shouldn’t Take That Long To Cut Them; GPUs Can Help

by Tony Kontzer

Few strive to combine the contrasting ideals of art and perfection as diamond cutters do.

But up until the 1990s, the focus on creativity and detail kept diamond processing at a snail’s pace, with the average diamond cutter able to produce just five finished pieces each day.

Then, cutters began using CPUs to assess diamonds to determine their color and clarity, and suggest the right cut. Almost immediately, they were able to increase their daily output to as many as 70 finished diamonds.

But, in what has become a $71 billion industry, that just wasn’t good enough.

“The challenge for this industry is either to reduce the amount of process time or increase the number of stones processed in that time,” said NVIDIA solutions architect Rupali Deshpande during a presentation Tuesday at GTC.

The answer to this challenge was to incorporate GPUs, which the industry has done to great effect.

Today, CUDA-powered GPUs are being used to:

  • Reconstruct 2D projections of diamonds in 3D in as little as 15 minutes.
  • Map imperfections present within a stone.
  • Speed up ray tracing processes that predict light reflection, ensuring a superior cut in a shorter timeframe.

As a result, diamond cutters aren’t just processing stones two to three times faster with GPUs than they were with CPUs. They also have improved the quality of diamonds, reduced their planning time requirements, and—of course—improved their bottom lines, said Deshpande.

And lest improved efficiency be perceived as a foe to the creativity and precision of diamond cutting, Deshpande said just the opposite is true. 

“The basic of this industry is the art,” she said. “Technology has just helped to elevate it.”