Neurowissenschaftliche Forschung auf Aurora Exascale
Nicola Ferrier, Senior Computer Scientist am Argonne National Laboratory, erklärt, wie Neurowissenschaftler mit dem Aurora Exascale Supercomputer unvorstellbare Mengen an Daten verarbeiten und komplexe Algorithmen erstellen werden, um Probleme zu lösen, die kein anderer bewältigen kann.
Transcript
[MUSIC PLAYING]
Having access to the first exascale system means that we're really going to be able to address problems that nobody else can address. Neuroscientists would like to understand the structure of the brain and the connections, and so they want to know what types of cells they are, what types of neurons, who's connected and how, so that they can then go on and do later studies of things like aging, development, learning, disease states.
In order to do that, you have to extract this information, not just from one brain, but from lots of brains. And so trying to build what's called the connectome of a brain, the amount of data is massive. If you want to do a centimeter cubed of a brain, which is a very tiny portion, this is petabytes of data. As we go to a larger part of the brain, we're going to be at exabytes of data.
The computer scientist in me says the possibility to handle the really large data sets that are being produced will be a very useful tool for all domains of science beyond just neuroscience. I'm really driven by the algorithms and the algorithm development. We're not going to be compute limited if we have an exascale machine, and so the sky's the limit in terms of development. And often as a, computer scientist, you do worry about whether you'll be able to run an algorithm at least in your lifetime. I'm thrilled to have the resources to address those problems.
[MUSIC PLAYING]
Exascale-Computing mit dem Aurora-Supercomputer
Der Aurora-Supercomputer, in dem Technik von Intel zum Einsatz kommt, wird die Konvergenz von High-Performance-Computing (HPC) und KI mit Exacscale-Computing beschleunigen.
Bahnbrechende Leistung für die Herausforderungen, denen Sie in der Praxis begegnen
Intels High-Performance-Computing-Plattform (HPC) kombiniert leistungsstarke Arbeits- und Datenspeicher-, Fabric- und Beschleunigungsfunktionalität, damit Sie Ihre größten Herausforderungen – von KI und Datenanalyse bis hin zu Simulation und Modellierung – bewältigen können.