# SiPearl Delivers First Rhea1 Processors to Bull to Power CPU Partition of European Exascale Supercomputer JUPITER

Source: TechNewsList (https://technewslist.com)
Canonical URL: https://technewslist.com/en/article/sipearl-rhea1-cpu-delivery-bull-jupiter-exascale-2026-09-25-morning
Section: Hardware (https://technewslist.com/en/hardware)
Author: TechNewsList
Language: en
Published: 2026-09-25T05:29:19.35+00:00
Updated: 2026-09-25T05:29:19.508391+00:00

> SiPearl shipped initial production samples of its eighty-core Rhea1 Arm server processor to Bull, beginning physical integration into the 1,300-node CPU partition of Europe's premier JUPITER exascale supercomputer.

## TL;DR
- SiPearl delivered initial samples of its 80-core Arm Rhea1 server CPU to high-performance computing vendor Bull.
- The processors are being integrated into BullSequana XH3000 blades for the European JUPITER exascale supercomputer.
- Rhea1 features 80 Arm Neoverse V1 cores, four HBM2E memory stacks, and four high-speed DDR5 memory channels.
- The 1,300-node CPU partition will execute complex, data-intensive scientific simulations at Forschungszentrum Jülich.

## Key points
- SiPearl announced the official silicon delivery milestone on September 24, 2026, advancing European semiconductor sovereignty.
- Bull began board-level validation and liquid-cooling thermal benchmarking on the BullSequana XH3000 platform.
- The dedicated CPU partition comprises 2,600 Rhea1 processors operating alongside NVIDIA GH200 GPU accelerators.
- Rhea1 is the most complex high-performance server processor ever designed and engineered in continental Europe.
- Development work has simultaneously commenced on Rhea2 to target France's future Alice Recoque exascale system.

## What happened

French fabless semiconductor designer SiPearl has officially delivered the first production samples of its flagship Rhea1 server processor to European supercomputer manufacturer Bull, an Eviden business unit. The delivery, announced on September 24, 2026, represents a historic watershed for European technological sovereignty. Rhea1 is the first high-performance general-purpose server processor designed entirely in continental Europe to be integrated into an operational exascale supercomputing platform.

Bull engineering teams at its Angers manufacturing and testing facility immediately commenced physical integration of the processors into the BullSequana XH3000 liquid-cooled platform. The Rhea1 silicon will populate the dedicated CPU-only cluster partition of the JUPITER supercomputer, currently undergoing deployment at the Jülich Supercomputing Centre within Forschungszentrum Jülich in Germany. When fully assembled, the dedicated CPU cluster will encompass 1,300 compute nodes powered by 2,600 Rhea1 processors, operating in parallel with the previously installed GPU-accelerated booster module.

## Why it matters

For decades, European scientific institutions, weather prediction centers, and industrial research laboratories were completely dependent on American and Asian semiconductor manufacturers for high-performance server microprocessors. In an era marked by shifting geopolitical alliances, supply chain export restrictions, and intensifying technological competition, European policymakers prioritized the development of sovereign compute silicon through the European Processor Initiative.

The physical arrival of Rhea1 silicon proves that European engineering consortiums can execute complex, multi-billion-transistor leading-edge server designs from architectural concept to working silicon. By dedicating 1,300 nodes of Europe's premier exascale system to a domestic CPU architecture, EuroHPC JU ensures that critical European scientific workloads—ranging from climate modeling and molecular dynamics to fusion energy simulation—can run on a sovereign hardware stack immune to external export restrictions.

![JUWELS high-performance supercomputing system installation at Forschungszentrum Jülich](https://rkhynbcsbnkkcwgexzwg.supabase.co/storage/v1/object/public/media/api/1790314148384-hpn8ys-sipearl-rhea1-cpu-delivery-bull-jupiter-exascale-2026-09-25-morning-inside-1-f6280b5f34.webp)

## Technical details

The Rhea1 processor represents a tour de force in modern high-performance microarchitecture. Fabricated on an advanced TSMC N6 process node, the monolithic die integrates 80 Arm Neoverse V1 cores arranged in a coherent high-bandwidth mesh topology. Each Neoverse V1 core incorporates dual 256-bit Scalable Vector Extension execution pipelines, enabling massive floating-point throughput for mathematical linear algebra routines without requiring specialized discrete graphics coprocessors.

Memory bandwidth—the traditional bottleneck for complex scientific codes—is addressed through a hybrid memory hierarchy. Rhea1 integrates four stacks of high-bandwidth memory (HBM2E) directly onto the multi-die substrate alongside four channels of standard DDR5 memory. This heterogeneous topology delivers over one terabyte per second of sustained memory bandwidth for cache-sensitive simulation kernels while preserving multi-terabyte system memory capacity for massive in-memory databases and graph analytics.

Thermal management in the BullSequana XH3000 chassis is handled via direct-to-chip liquid cooling utilizing 40-degree Celsius warm water. This cooling architecture eliminates the need for energy-intensive mechanical chillers, allowing the JUPITER installation to achieve world-leading power-usage effectiveness ratings while sustaining peak boost clock frequencies across all 80 cores under continuous full-chassis thermal loads.

![Forschungszentrum Jülich research center campus facility in Germany](https://rkhynbcsbnkkcwgexzwg.supabase.co/storage/v1/object/public/media/api/1790314152755-lhmyj0-sipearl-rhea1-cpu-delivery-bull-jupiter-exascale-2026-09-25-morning-inside-2-ef58069cb5.webp)

## Market / industry impact

The successful delivery of Rhea1 disrupts the entrenched global server silicon triopoly dominated by Intel, AMD, and NVIDIA. By validating Arm-based high-performance computing in Europe's flagship exascale facility, SiPearl establishes a viable commercial alternative for enterprise cloud providers, national laboratories, and telecommunications operators seeking high-efficiency server architectures tailored for memory-bound workloads.

The milestone also strengthens Bull and Eviden's standing in global supercomputing competitions. By fielding an integrated hardware portfolio featuring proprietary liquid-cooled enclosures, domestic server motherboards, and European-designed processors, Eviden can pitch turnkey sovereign high-performance computing installations to governments across the Middle East, Asia, and Latin America seeking non-aligned technological infrastructure.

Furthermore, SiPearl's silicon validation pipeline provides immediate momentum for downstream European semiconductor initiatives. The architectural innovations proven in Rhea1 are already being channeled into the design of Rhea2, a next-generation processor slated to power the CPU partition of the upcoming French *Alice Recoque* exascale supercomputer scheduled for deployment later in the decade.

## What to watch next

Over the next two quarters, engineers at Forschungszentrum Jülich will conduct exhaustive board-level validation, multi-node interconnect testing, and Linpack benchmarking on the Rhea1 cluster. The supercomputing community will monitor whether the dual-socket Rhea1 nodes achieve their projected performance-per-watt targets when executing unstructured mesh simulations and computational fluid dynamics applications that perform poorly on GPU architectures.

Software ecosystem enablement will represent another crucial benchmark. While Arm's enterprise server software stack has matured dramatically, compilers, scientific libraries, and MPI communication runtimes will require fine-tuning to fully exploit Rhea1's unique hybrid HBM2E and DDR5 memory controllers. Scientific consortia across Europe are preparing optimized code repositories to ensure immediate research productivity upon cluster commissioning.

Finally, industry observers will watch for SiPearl's commercial expansion announcements outside EuroHPC research facilities. If SiPearl secures deployment agreements with European automotive manufacturers, aerospace defense contractors, or sovereign AI cloud operators, the company will have completed the transition from a government-sponsored research project into an enduring commercial semiconductor powerhouse.

## Sources

* [SiPearl Corporate News](https://sipearl.com/en/press-releases/sipearl-delivers-rhea1-processor-to-bull-for-jupiter-supercomputer) - Official silicon manufacturer announcement detailing Rhea1 sample shipments, core architecture, and Bull collaboration.
* [Techzine Infrastructure](https://www.techzine.eu/news/infrastructure/123891/sipearl-delivers-first-rhea1-cpus-for-european-jupiter-supercomputer/) - Independent European technology analysis examining BullSequana blade integration, cluster node counts, and exascale benchmarks.
* [European Processor Initiative](https://www.european-processor-initiative.eu/news/rhea1-silicon-samples-delivered-to-bullsequana-platform/) - Technical consortium documentation specifying Neoverse V1 vector performance, HBM2E memory interfaces, and sovereign computing objectives.

Mentions: SiPearl, Bull, Eviden, Philippe Notton, JUPITER Supercomputer, Forschungszentrum Jülich

## Sources
- [SiPearl Corporate News](https://sipearl.com/en/press-releases/sipearl-delivers-rhea1-processor-to-bull-for-jupiter-supercomputer)
- [Techzine Infrastructure](https://www.techzine.eu/news/infrastructure/123891/sipearl-delivers-first-rhea1-cpus-for-european-jupiter-supercomputer/)
- [European Processor Initiative](https://www.european-processor-initiative.eu/news/rhea1-silicon-samples-delivered-to-bullsequana-platform/)