---
title: "Nvidia Introduces CUDA-X: EHT Telescope Signal Matching Now Takes 8 Minutes Instead of a Month"
description: "Nvidia has expanded its CUDA-X library suite: EHT telescope signal matching dropped from a month to 8 minutes, quantum-mechanical calculations accelerated by 22x, and cellular data analysis by 900x."
date: 2026-11-19T01:12:59.863Z
lang: en
url: https://xab.info/en/posts/nvidia-cuda-x-scientific-computing-acceleration
tags: [nvidia, cuda-x, gpu-computing, scientific-computing, event-horizon-telescope, fusion-energy, drug-discovery, molecular-modeling, materials-science, biology]
author: "Alexander Buran"
publisher: "XAB.info"
---

# Nvidia Introduces CUDA-X: EHT Telescope Signal Matching Now Takes 8 Minutes Instead of a Month

![Four panels with visualizations of Nvidia CUDA-X scientific computing including the EHT telescope and molecular structures](https://xab.info/media/2026/11/19/nvidia-cuda-x-nauchnye-vychisleniya-uskorenie/nvidia-cuda-x-nauchnye-vychisleniya-uskorenie-1.webp)

## Key Points

- The cuPhoton.xBeam module reduced EHT telescope radio signal matching time from over a month to 8 minutes
- The cuPotential library accelerates molecular modeling for drug discovery by 3 times
- Daviator GPU cores provide a 22x acceleration of quantum-mechanical calculations in VASP with a 75% cost reduction
- The open cuSpa library accelerates cell contour calculations by 900 times compared to CPUs
- A local AI agent based on Nvidia Nemotron can be connected to the plasma simulation process for real-time analysis

Nvidia has announced an expansion of its CUDA-X library suite, designed to accelerate scientific computing on GPUs. The announcement, made on November 18, 2026, spans several fundamental domains—from astrophysics and nuclear fusion to drug discovery and materials science. According to the company, the new modules allow operations that previously took weeks or months to be completed in a matter of minutes, while reducing computing costs by orders of magnitude.

### Astronomy: Black Holes in 8 Minutes

One of the most striking results was demonstrated by the cuPhoton.xBeam module. It is designed to match radio signals coming from the Event Horizon Telescope (EHT) project, a global network of observatories combined to observe supermassive black holes. EHT generates petabytes of data that must be combined and processed to reconstruct images. With cuPhoton.xBeam, this operation, which previously took over a month, was reduced to eight minutes. The performance difference is on the order of five to six magnitudes, opening up the possibility for more rapid data analysis and re-running calculations with modified parameters.

### Nuclear Fusion: Real-Time AI Agent

Within the same cuPhoton library, the xLens module has been implemented, allowing plasma simulation results to be analyzed directly during calculations rather than after their completion. Pacific Fusion uses this tool to gain insights into plasma behavior during the simulation process. A local AI agent based on Nvidia Nemotron can be connected to the computational process, analyzing intermediate results and helping to adjust the simulation in real time. A similar approach is used in research related to the development of targets and chambers for fusion reactors, where the ability to quickly react to parameter changes is critical for experiment optimization.

### Pharmacology and Molecular Modeling

In drug discovery, Nvidia is promoting the cuPotential library, created based on work by specialists from the Flatiron Institute and D. E. Shaw Research. It accelerates existing molecular modeling software by three times while simultaneously improving the stability and accuracy of calculations. With it, researchers can more quickly generate synthetic data to train generative AI models, discover previously unknown molecular binding sites, and calculate their interaction strength. Additional context is provided by the NVIDIA Nemotron for Science package, introduced in parallel with CUDA-X: it is built on the open Nemotron 3.5 Super VL model and includes a new set of chemical data and post-training recipes, expanding the capabilities of AI tools in chemistry and pharmacology.

### Materials Science: 22x Acceleration of Quantum-Mechanical Calculations

In materials science, Nvidia announced an acceleration of quantum-mechanical calculations by up to 22 times while reducing their cost by 75%. This result is delivered by Daviator GPU cores, designed for the VASP software package—one of the primary tools for modeling material properties at the atomic level. A key advantage of the new cores is that they accelerate calculations for systems of various sizes, including smaller models that were previously handled more efficiently by central processing units. This is important for preparing millions of atomic interaction calculations required for training and operating computational models in materials science.

### Experimental Physics: Data Processing Before Experiment Completion

A separate development is applied directly during experiments. At the Lawrence Berkeley National Laboratory, the Nvidia DAQIRI library connects a high-speed electron microscope detector to a GPU data processing system. Scientists are able to isolate useful signals, detect previously unknown areas of interest, and choose subsequent measurements even before the experiment concludes. The technology is also intended for studying materials with atomic resolution, allowing the distinction of individual chemical elements, which is critically important for nanotechnology and condensed matter physics.

### Biology: From Cell Contours to Protein Functions

In biological research, the open cuSpa library accelerated cell contour calculations by 900 times compared to CPUs. It allows for the construction of large spatial tissue graphs used in creating disease and cancer models. The rapids-singlecell tool, developed jointly with the scverse community, accelerates single-cell data analysis: Therna Biosciences reported a 200x processing speedup, and 10x Genomics integrated this technology into its Sentira platform. Additionally, the open GPU HMMER library accelerates the determination of protein functions from their amino acid sequences—in Basecamp Research tests, searches ran up to 84 times faster than the standard CPU version. All of these computations are used in preparing large biological datasets for training models applied in drug discovery.

## Fact-Check & Sources

- [Nvidia ускоряет научные открытия: компания представила целый набор инструментов для поиска лекарств, изучения чёрных дыр и моделирования термоядерного синтеза](https://www.ixbt.com/news/2026/11/18/nvidia-uskoriaet-naucnye-otkrytiia-kompaniia-predstavila-celyi-nabor-instrumentov-dlia-poiska-lekarstv-izuceniia-cernyx-dyr-i-modelirovaniia-termoiadernogo-sinteza.html) - Primary source. Contains all key facts: 8 minutes for EHT, 3x for cuPotential, 22x for VASP/Daviator, 900x for cuSpa, 200x for Therna Biosciences, 84x for HMMER, DAQIRI in Berkeley, xLens for Pacific Fusion.
- [От астрономии до генетики. NVIDIA представила открытую модель для обработки научных данных](https://techno.nv.ua/it-industry/nvidia-nemotron-for-science-novye-instrumenty-dlya-uskoreniya-nauchnyh-issledovaniy-50659948.html) - Confirms the parallel announcement of NVIDIA Nemotron for Science based on Nemotron 3.5 Super VL with an open chemistry dataset. Does not contradict the main text.

## FAQ

### What is CUDA-X and what is it used for?

CUDA-X is a suite of Nvidia libraries for accelerating scientific computing on GPUs. In the updated 2026 version, it covers astrophysics, nuclear fusion, pharmacology, materials science, experimental physics, and biology.

### How long does EHT telescope data processing take now?

Using the cuPhoton.xBeam module, matching radio signals from the Event Horizon Telescope takes 8 minutes instead of over a month as it did previously.

### How is the Nemotron AI agent used in plasma modeling?

A local AI agent based on Nvidia Nemotron connects to the plasma simulation computational process (xLens module) and analyzes intermediate results in real time, helping to adjust the simulation. This is applied in fusion reactor research, particularly at Pacific Fusion.

### What are Daviator GPU cores?

Daviator are specialized Nvidia GPU cores designed for the VASP software package. They accelerate quantum-mechanical calculations up to 22 times and reduce their cost by 75%, working efficiently for both large and small atomic systems.

### What biological tasks does the new version of CUDA-X accelerate?

The cuSpa library accelerates cell contour calculations by 900 times, rapids-singlecell (jointly with scverse) accelerates single-cell data analysis (up to 200x according to Therna Biosciences), and HMMER accelerates protein function determination from amino acid sequences (up to 84x faster in Basecamp Research tests).