FreeLLMNVIDIA cuPhoton xBeam correlates 10 TB EHT dataset in 70 seconds on 64‑GPU system
cuPhoton’s GPU‑based xBeam processed a 10 TB black‑hole interferometry dataset in 70 seconds, a 460× speedup over CPU.
The FX correlator ingested VDIF frames, applied delay and fringe corrections, channelized signals via FFT, cross‑multiplied station pairs and wrote visibilities in DiFX SWIN or xbeam‑v1 format.
A 10 TB dataset from the Event Horizon Telescope was processed in 70 seconds, delivering a 460× speedup compared with an x86 CPU baseline that took 31,992 seconds (about 8.89 hours).
The team highlighted that AI agents can automate pipeline steps by parsing VDIF metadata, generating JSON job specs, launching xBeam jobs and monitoring execution.
What it means for you
The dramatic acceleration enables researchers to iterate on interferometry pipelines far faster, reducing analysis time from hours to minutes. For AI developers, the result showcases how GPU‑optimized software can unlock scientific workloads, encouraging similar optimizations for free or locally run models.
Sources
- Run GPU-Powered Black Hole Interferometry with NVIDIA cuPhoton, NVIDIA Developer (official), 13 Nov 2026
Written automatically from the sources above, then checked by our code: every number and name must appear in them, and no sentence may copy a source. Spotted a mistake? Report it.
Which AI can you use for free today?
Every free AI model, ranked by blind human votes, with the memory and speed it needs on your own GPU.