Anders Jensen, EuroHPC JU govt director, said: “I’m proud to announce the launch of the DARE mission which marks a major milestone for European digital sovereignty. This formidable initiative will drive innovation in each {hardware} and software program applied sciences and leverage the complete energy of HPC and AI to develop safe, environment friendly and European-led options for the longer term.”
Europe has traditionally relied on foreign-built processors and software program to drive HPC-AI techniques, leaving key industries weak to produce chain disruptions and exterior dependencies, EuroHPC JU mentioned. DARE seeks to alter that by creating totally European-designed processors, the important constructing blocks of future supercomputers. This may guarantee Europe can management its personal technological future, benefiting industries, researchers, and residents alike.
DARE is ready to create three new European designed and owned computing chips:
- Vector accelerator (VEC) to boost scientific and engineering simulations.
- AI Processing Unit (AIPU) designed for AI-powered purposes like language processing and knowledge evaluation.
- Normal-purpose processors (GPP) optimized for high-performance supercomputing duties.
These chips will energy next-generation European supercomputers, serving to researchers sort out essential challenges similar to drug discovery, local weather change modeling, and renewable power options.
Anders Jensen, EuroHPC JU Government Director
Osman Unsal, DARE Principal Investigator at BSC, said: “DARE is daring to begin from the highest of the technological complexity pile and produce European-designed processor chips for supercomputers, paving the way in which for Europe’s digital sovereignty.”
To make sure speedy progress, DARE will undertake a {hardware} (HW)/software program(SW) design strategy, utilizing real-world purposes to fine-tune efficiency and capabilities. This strategy will speed up innovation and help the rising demand for energy-efficient excessive efficiency computing.
Dwelling to MareNostrum 5, one among Europe’s strongest supercomputers, BSC has been a frontrunner in HPC analysis for years. With DARE, it can assist drive the event of Europe’s first totally homegrown supercomputing system.
A consortium of 38 European companions, led by BSC, is collaborating on the DARE mission to advance digital autonomy in HPC and AI. Openchip, Axelera AI, Codasip, imec and JÜLICH SUPERCOMPUTING CENTRE (JSC) at FORSCHUNGSZENTRUM JÜLICH will function technical leads, driving key improvements throughout the mission.
Because the mission coordinator, BSC performs a key position in advancing Europe’s supercomputing capabilities. Apart from being general coordinator, BSC can even lead roadmapping in addition to the VEC pathfinding efforts whereas taking part in HW and SW growth actions.
DARE is greater than only a analysis mission—it’s a strategic transfer in direction of technological sovereignty. By creating Europe’s personal processors and computing stack, the mission will pave the way in which for future generations of supercomputers which are designed, constructed, and optimized in Europe.
By the top of this part, DARE can have laid the inspiration for a European HPC ecosystem, guaranteeing that Europe stays on the forefront of scientific discovery, innovation, and digital autonomy, based on EuroHPC JU.
Digital Autonomy with RISC-V in Europe (DARE SGA1) is a large-scale European supercomputing mission that has obtained funding from the EuroHPC JU below grant settlement No 101202459.
DARE goals to develop prototype HPC and AI techniques primarily based on EU-designed and developed industry-standard chiplets, utilizing the newest silicon expertise nodes to realize the best efficiency and power effectivity. This three-year, 38 companions initiative, with a price range of €240 million and coordinated by BSC , will develop a complete supercomputing compute stack, that includes high-performance and energy-efficient processors designed and developed in Europe, enabled by an optimized software program stack.
