Signaloid’s compute hardware joins CERN’s tech testbed

British computing company Signaloid is collaborating with CERN openlab to test its distribution-extended compute hardware (UxHw) within the Heterogeneous Architectures Testbed. The partnership arrives as CERN prepares for the High-Luminosity Large Hadron Collider (HiLumi LHC), anticipating a surge in computational demands that existing infrastructure may struggle to meet.

“Heterogeneous architectures are becoming essential for the HiLumi LHC and CERN openlab is pioneering a model for evaluating computing technologies such as Signaloid’s distribution-extended compute hardware technology,” says Maria Girone, CTO CERN openlab. Signaloid’s technology aims to accelerate Monte Carlo simulations, critical for particle physics, by performing calculations directly on probability distributions, potentially achieving speed-ups of up to 2,000×.

UxHw Technology Extends Heterogeneous Computing for Probability Distributions

Signaloid’s distribution-extended compute hardware (UxHw) is now undergoing evaluation within the CERN openlab Heterogeneous Architectures Testbed, a dedicated environment for assessing emerging processor technologies. This testing phase isn’t a distant research project; it’s directly tied to the operational demands of the upcoming High-Luminosity Large Hadron Collider (HiLumi LHC), which will dramatically increase data volumes and computational needs. The HiLumi LHC is projected to significantly exceed current computing resources, necessitating exploration of novel hardware solutions to maintain analysis capabilities.

The core innovation of UxHw lies in its ability to natively support computation on digital representations of continuous probability distributions, moving beyond the limitations of traditional CPU/GPU combinations. This approach is particularly relevant to the HiLumi LHC’s reliance on Monte Carlo simulations, a computationally intensive technique used to model particle interactions and predict experimental outcomes.

According to Prof. Phillip Stanley-Marbell, Founder and CEO of Signaloid, “The future of high-performance computing will not be defined by a single processor architecture, but by heterogeneous systems that combine specialised hardware for different classes of computation.” The joint project will specifically focus on identifying where distribution-extended computing can accelerate the Monte Carlo event generation pipeline, a critical step in processing LHC data. The collaboration with Signaloid aligns with CERN openlab’s broader strategy of proactively evaluating technologies before their full-scale deployment. This evaluation is not isolated; it reflects a global trend toward heterogeneous computing systems, evidenced by investments like the UK’s £750 million AI Research Resource (AIRR) heterogeneous supercomputer.

Because event generation relies heavily on multi-dimensional distributions, Signaloid’s UxHw technology offers the potential to significantly accelerate this software, helping to meet the forecasted computing budgets during HiLumi LHC operations starting in 2030. “We’re excited that CERN openlab is evaluating UxHw alongside other emerging computing technologies,” Stanley-Marbell added, “Experimental particle physics represents one of the most demanding and exciting environments in which to demonstrate its potential.” The project aims to determine the practical considerations involved in integrating this specialized hardware into existing high-energy physics software.

CERN openlab Evaluates UxHw for HiLumi LHC Monte Carlo Simulation

This testing phase focuses on a representative Monte Carlo event generation workflow, using the Pepper framework for proton-proton collisions, and will assess computational performance alongside numerical accuracy and integration effort. The project aims to pinpoint how distribution-extended computing technologies like UxHw can best integrate with existing CPU and GPU infrastructure at CERN. The evaluation isn’t simply a performance benchmark. It’s designed to understand the practical implications of adopting a new computing architecture within a complex scientific environment.

Stefan Roiser, Senior Computing Engineer at CERN, explained that the largest portion of LHC computing resources is currently dedicated to simulating particle collisions. He stated, “We will explore Signaloid’s technology in Monte Carlo event generation, the first step in the simulation chain expected to see substantial cost increases during CERN’s upcoming High Luminosity data-taking period.” This focus on the initial simulation stage highlights a strategic approach to mitigating future computational bottlenecks.

The HiLumi LHC’s escalating demands are driving CERN to explore diverse solutions beyond conventional computing paradigms. CERN’s Quantum Technology Initiative, launched in 2020, demonstrates this broader commitment to investigating potentially disruptive technologies. The company’s technology is being assessed for its ability to accelerate Monte Carlo simulations, a computationally intensive process vital for comparing experimental measurements with simulated particle collisions.

This evaluation will determine if distribution-extended computing can alleviate the anticipated computational burden as the HiLumi LHC generates increasingly complex datasets. CERN’s recent collaboration with CNPEM on 91km collider magnet technology, reported on September 26, highlights the institution’s dedication to advancing the underlying infrastructure supporting its experiments.

Signaloid’s UxHw Achieves 2,000x Speed-Ups Across Multiple Workloads

Signaloid’s distribution-extended compute hardware, UxHw, has demonstrated speed-ups of up to 2,000× across workloads including high-energy physics, regulatory risk modeling, chip design simulations and robotics, according to benchmarking against current server platforms. This performance leap stems from UxHw’s ability to directly calculate on probability distributions in a single pass, bypassing the need for repeated kernel executions common in Monte Carlo simulations.

The technology achieves this efficiency while requiring minimal alterations to existing software, an important factor for integration into established scientific workflows. The specific workflow simulates multiple gluon production, a computationally intensive process vital for analyzing data from the High-Luminosity Large Hadron Collider, slated to begin operations in 2030.

The potential for acceleration is particularly strong in event generation, where multi-dimensional distributions are heavily used. This evaluation aligns with a broader trend in high-performance computing, with organizations worldwide investing in heterogeneous systems combining CPUs, GPUs and specialized accelerators. Signaloid’s platforms aim to replace time-consuming Monte Carlo simulations with efficient deterministic computations, delivering equivalent probability distribution information.

The company offers flexible deployment options, including cloud compute, on-premises installations and edge hardware, designed to simplify adoption for organizations in finance, robotics, and manufacturing. This versatility, combined with the demonstrated speed-ups, positions Signaloid as a potential solution for addressing the escalating computational demands of scientific research and increasingly complex AI workloads.

More like this

Emerging TechnologyeMerge Americas hosts launch of Florida Quantum initiative

Artificial IntelligenceOffice of Science and Technology Policy says AI is now ‘Super Intelligence’

Emerging TechnologyCalifornia invests in fusion and quantum research.

Quantum SensorsOrganic photosensors tune signal amplitude with ionic control.