Prepared for a front-row seat to the subsequent scientific revolution?
That’s the thought behind Doudna — a groundbreaking supercomputer being constructed at Lawrence Berkeley Nationwide Laboratory. The system represents a significant nationwide funding in advancing U.S. high-performance computing management, guaranteeing U.S. researchers have entry to cutting-edge instruments to handle international challenges.
Also referred to as NERSC-10, Doudna is called for Nobel laureate and CRISPR pioneer Jennifer Doudna. The subsequent-generation system introduced at the moment at Lawrence Berkeley Nationwide Laboratory is designed not only for pace, however for impression.
“The Doudna system represents DOE’s dedication to advancing American management in science, AI, and high-performance computing,” stated U.S. Secretary of Vitality Chris Wright stated in an announcement.
Powered by Dell infrastructure with the NVIDIA Vera Rubin structure, and set to launch in 2026, Doudna is tailor-made for real-time discovery throughout the U.S. Division of Vitality’s most pressing scientific missions. It’s poised to catapult American researchers to the forefront of vital scientific breakthroughs, fostering innovation and securing the nation’s aggressive edge in key technological fields.
“Doudna is a time machine for science — compressing years of discovery into days,” stated Jensen Huang, founder and CEO of NVIDIA in an announcement. “Constructed along with DOE and powered by NVIDIA’s Vera Rubin platform, it’ll let scientists delve deeper and suppose larger to hunt the elemental truths of the universe.”
Designed to Speed up Breakthroughs
In contrast to conventional methods that function in silos, Doudna merges simulation, knowledge and AI right into a single seamless platform.
It’s engineered to empower over 11,000 researchers with nearly instantaneous responsiveness and built-in workflows, serving to scientists discover larger questions and attain solutions quicker than ever.
“We’re not simply constructing a quicker pc,” stated Nick Wright, superior applied sciences group lead and Doudna chief architect at NERSC. “We’re constructing a system that helps researchers suppose larger, and uncover sooner.”
Right here’s what Wright expects Doudna to advance:
- Fusion vitality: Breakthroughs in simulation that unlocks clear fusion vitality.
- Supplies science: AI fashions that design new lessons of superconducting supplies.
- Drug discovery acceleration: Ultrarapid workflow that helps biologists fold proteins quick sufficient to outpace a pandemic.
- Astronomy: Actual-time processing of knowledge from the Darkish Vitality Spectroscopic Instrument at Kitt Peak to assist scientists map the universe.
Doudna is anticipated to outperform its predecessor, Perlmutter, by greater than 10x in scientific output, all whereas utilizing simply 2-3x the facility.
This interprets to a 3-5x enhance in efficiency per watt, a results of improvements in chip design, dynamic load balancing and system-level efficiencies.
AI-Powered Discovery, at Scale
Doudna will energy AI-driven breakthroughs throughout high-impact scientific fields nationwide.
Highlights embrace:
- AI for protein design: David Baker, a 2024 Nobel laureate, used NERSC methods to help his work utilizing AI to foretell novel protein buildings, addressing challenges throughout scientific disciplines.
- AI for elementary physics: Researchers like Benjamin Nachman are utilizing AI to “unfold” detector distortions in particle physics knowledge and analyze proton knowledge from electron-proton colliders.
- AI for supplies science: A collaboration together with Berkeley Lab and Meta created “Open Molecules 2025,” an enormous dataset for utilizing AI to precisely mannequin advanced molecular chemical reactions. Researchers concerned additionally use NERSC for his or her AI fashions.
Actual-Time Science, Actual-World Affect
Doudna isn’t a standalone system. It’s an integral a part of scientific workflows. DOE’s ESnet will stream knowledge from telescopes, detectors and genome sequencers straight into the machine with low-latency, high-throughput NVIDIA Quantum-X800 InfiniBand networking.
This vital knowledge stream is prioritized by clever QoS mechanisms, guaranteeing it stays quick and uninterrupted, from enter to perception.
It will make the system extremely responsive. On the DIII-D nationwide fusion ignition facility, for instance, knowledge will stream control-room occasions straight into Doudna for rapid-response plasma modeling, so scientists could make changes in actual time.
“We used to consider the supercomputer as a passive participant within the nook,” Wright stated. “Now it’s a part of the whole workflow, related to experiments, telescopes, detectors.”
The Platform for What’s Subsequent: Unlocking Quantum and HPC Workflows
Doudna helps conventional HPC, cutting-edge AI, real-time streaming and even quantum workflows.
This consists of help for scalable quantum algorithm improvement and the co-design of future built-in quantum-HPC methods, utilizing platforms like NVIDIA CUDA-Q.
All of those workflows will run on the next-generation NVIDIA Vera Rubin platform, which can mix high-performance CPUs with coherent GPUs, that means all processors can entry and share knowledge on to help essentially the most demanding scientific workloads.
Researchers are already porting full pipelines utilizing frameworks like PyTorch, the NVIDIA Holoscan software program improvement equipment, NVIDIA TensorFlow, NVIDIA cuDNN and NVIDIA CUDA-Q, all optimized for the system’s Rubin GPUs and NVIDIA NVLink structure.
Over 20 analysis groups are already porting full workflows to Doudna via the NERSC Science Acceleration Program, tackling every part from local weather fashions to particle physics. This isn’t nearly uncooked compute, it’s about discovery, built-in from concept to perception.
Designed for Urgency
In 2024, AI-assisted science earned two Nobel Prizes. From local weather analysis to pandemic response, the subsequent breakthroughs gained’t look ahead to higher infrastructure.
With deployment slated for 2026, Doudna is positioned to guide a brand new period of accelerated science. DOE amenities throughout the nation, from Fermilab to the Joint Genome Institute, will depend on its capabilities to show at the moment’s questions into tomorrow’s breakthroughs.
“This isn’t a system for one area,” Wright stated. “It’s for discovery — throughout chemistry, physics and fields we haven’t imagined but.”
As NVIDIA founder and CEO Jensen Huang put it, Doudna is “a time machine for science.” It compresses years of discovery into days, and provides the world’s hardest issues the facility they’ve been ready for.