usatoday24
Supercomputing has never been just a matter of science. Since its origins, these colossal systems have represented national power, reflecting the technological, scientific and even military capacities of the countries that develop them. Now, The United States has presented His next big bet in this field: Doudna, a superorous who should see the light in 2026 and who promises to be more than ten times more powerful than Perlmutterthe current flagship of the Lawrence Berkeley National Laboratory.
A tribute to science that changed biology. The new system has been baptized in honor of Jennifer DoudnaBerkeley’s biochemistry professor and one of the scientists who promoted CRISPR technology, Recognized with the Nobel Prize in Chemistry in 2020. It is not a minor gesture: the choice of the name symbolizes the fusion between biomedical research, artificial intelligence (AI) and computational power. Three axes that, together, aim to define the scientific advances of the next decades.
What is exactly Doudna and what will you do. Doudna will be the next super -tider of the National Energy Research Scientific Computing Center (Nersc), a center of the United States Department of Energy located in Berkeley. Its design is in charge of Dell Technologies and will use the new platform Vera Rubin de Nvidiawhich will integrate ARM processors of general purpose. At a technical level, it is designed to execute large -scale hybrid work loads: high fidelity scientific simulations, training models of artificial intelligence, real -time data analysis and quantum algorithms.
The machine will not be limited to executing tasks faster: it is designed to integrate into scientific workflows where the data comes from telescopes and laboratories. The objective is clear: to transform the way in which science is done, allowing to adjust experiments almost instantly and accelerate processes that took weeks or months.
A project that wants to make a difference. Those responsible for the project are explicit: Doudna is presented as a key piece in the American strategy to lead the development of AI. Chris Wright, Secretary of Energy, It came to compare it with the Manhattan projectensuring that this supercomputer will be fundamental to win the Global AI race. Beyond the headline, the fields of application that are being prepared are ambitious.
Nersc has already identified more than twenty scientific teams that are adapting their workflows for this system. The key tools are already on the table: from Frameworks such as Pytorch, Tensorflow and Cuda-Q to the Holoscan development kit, all optimized for the coherent architecture of Rubin and the NVLink interconnection.
Why is it important at the national (and world) level. In a context of growing technological competition, especially with China, these types of systems represent more than a scientific resource. They are strategic infrastructure. Japan, with his escape; China, with its most recent Sunway Oceanlite; And now the United States, I already had Frontieramong others, it is reinforcing its muscle in this area with Doudna.
The election of Dell against HPE also breaks with the usual dynamics of the great contracts of the Department of Energy, which until now had favored the latter in their three most recent exaescala systems.
A jump in energy efficiency. In addition to gross power, one of Doudna’s great achievements will be its efficiency. According to NvidiaThe system will offer between three and five times more perlmutter performance. This is possible thanks to improvements in chips design, dynamic load balancing and new optimizations at the system level.
When will it come and what is known about deployment. Douda’s launch is scheduled for 2026. We know that it will be built with servers Poweredge and advanced liquid cooling technologies of Dell, in addition to high -speed connectivity through the network Nvidia Quantum-X800 Infiniband. It will be located at the Berkeley Lab facilities and connected directly to the rest of the Energy Department centers through the scientific network ESNET.
What has not been revealed, for now, is the official budget. Unlike the super -terrorist The Captain, which cost 600 million dollarsthe Department of Energy has not advanced any figure on the total investment in Doudna. Nor has it specified how many nodes or how much exact memory the system will have in its final configuration. But it has been made clear that it is designed to climb and adapt over time.
Images | Nvidia (1, 2) | Lawrence Berkeley National Laboratory
GIPHY App Key not set. Please check settings