Matter Matters: Understanding Anti-Matter With the Help of GPUs

by Tony Kontzer

Does antimatter fall to Earth the same way matter does? This question has long perplexed scientists due to antimatter’s rarity, and its tendency to annihilate upon impact with matter.

Thanks to advances in GPU technology, researchers at Switzerland’s University of Bern are getting closer to being able to analyze the impact of gravity on antimatter by studying the way particles fall when annihilation occurs.

Akitaka Ariga, of the university’s Albert Einstein Center for Fundamental Physics, told attendees during a session at NVIDIA’s GPU Technology Conference Wednesday that conclusive results are just a year or two away.

The path to observing the emissions of matter-antimatter collisions has been a long one. The photographic emulsion detectors used to capture collision images have been around since the 19th century, and have played critical roles in the discoveries of natural radioactivity in 1896, the antiproton in 1955 and the X particle 16 years later.

More recently, detector technology has advanced to the point where it can deliver the resolution needed to detect antimatter annihilation. Unfortunately, that capability has created processing challenges.

“The principle is simple, but reality is much more complicated,” said Ariga, who’s working on an experiment at CERN, the European Organization for Nuclear Research.

The amount of data collected by the high-resolution detection of annihilations clocks in at 210 MB per second. And with a lack of automated scanning tools necessitating human analysis, analyzing the data — which involves looking for 3D lines and groupings of antimatter particle grains emitted during annihilations — has been an insurmountable task.

GPUs have changed that.

By using a single GPU, Ariga and his team have been able to achieve a 25X improvement in image filtering, an 8X improvement in 3D grain recognition and a 16X gain in 3D tracking.

Still, Ariga called that progress “not good enough,” which is why his team opted to see what results they got with multiple GPUs, pushing the tracking gains to a factor of 60.

The moral, said Ariga, is that “high energy physics needs GPUs for very heavy computation.”