Beginning with the discharge of CUDA in 2006, NVIDIA has pushed developments in AI and accelerated computing — and the newest TOP500 checklist of the world’s strongest supercomputers highlights the fruits of the corporate’s achievements within the area.
This 12 months, 384 techniques on the TOP500 checklist are powered by NVIDIA applied sciences. Among the many 53 new to the checklist, 87% — 46 techniques — are accelerated. Of these accelerated techniques, 85% use NVIDIA Hopper GPUs, driving developments in areas like local weather forecasting, drug discovery and quantum simulation.
Accelerated computing is way more than floating level operations per second (FLOPS). It requires full-stack, application-specific optimization. At SC24 this week, NVIDIA introduced the discharge of cuPyNumeric, an NVIDIA CUDA-X library that permits over 5 million builders to seamlessly scale to highly effective computing clusters with out modifying their Python code.
On the convention, NVIDIA additionally revealed vital updates to the NVIDIA CUDA-Q growth platform, which empowers quantum researchers to simulate quantum units at a scale beforehand thought computationally not possible.
A New Period of Scientific Discovery With Blended Precision and AI
Blended-precision floating-point operations and AI have turn out to be the instruments of selection for researchers grappling with the complexities of recent science. They provide larger velocity, effectivity and flexibility than conventional strategies, with out compromising accuracy.
This shift isn’t simply theoretical — it’s already occurring. At SC24, two Gordon Bell finalist tasks revealed how utilizing AI and combined precision helped advance genomics and protein design.
In his paper titled “Utilizing Blended Precision for Genomics,” David Keyes, a professor at King Abdullah College of Science and Know-how, used 0.8 exaflops of combined precision to discover relationships between genomes and their generalized genotypes, after which to the prevalence of illnesses to which they’re topic.
Equally, Arvind Ramanathan, a computational biologist from the Argonne Nationwide Laboratory, harnessed 3 exaflops of AI efficiency on the NVIDIA Grace Hopper-powered Alps system to hurry up protein design.
To additional advance AI-driven drug discovery and the event of lifesaving therapies, researchers can use NVIDIA BioNeMo, highly effective instruments designed particularly for pharmaceutical purposes. Now in open supply, the BioNeMo Framework can speed up AI mannequin creation, customization and deployment for drug discovery and molecular design.
Throughout the TOP500, the widespread use of AI and mixed-precision floating-point operations displays a world shift in computing priorities. A complete of 249 exaflops of AI efficiency are actually out there to TOP500 techniques, supercharging improvements and discoveries throughout industries.
NVIDIA-accelerated TOP500 techniques excel throughout key metrics like AI and mix-precision system efficiency. With over 190 exaflops of AI efficiency and 17 exaflops of single-precision (FP32), NVIDIA’s accelerated computing platform is the brand new engine of scientific computing. NVIDIA additionally delivers 4 exaflops of double-precision (FP64) efficiency for sure scientific calculations that also require it.
Accelerated Computing Is Sustainable Computing
Because the demand for computational capability grows, so does the necessity for sustainability.
Within the Green500 checklist of the world’s most energy-efficient supercomputers, techniques with NVIDIA accelerated computing rank amongst eight of the highest 10. The JEDI system at EuroHPC/FZJ, for instance, achieves a staggering 72.7 gigaflops per watt, setting a benchmark for what’s attainable when efficiency and sustainability align.
For local weather forecasting, NVIDIA introduced at SC24 two new NVIDIA NIM microservices for NVIDIA Earth-2, a digital twin platform for simulating and visualizing climate and local weather situations. The CorrDiff NIM and FourCastNet NIM microservices can speed up local weather change modeling and simulation outcomes by as much as 500x.
In a world more and more aware of its environmental footprint, NVIDIA’s improvements in accelerated computing steadiness excessive efficiency with power effectivity to assist understand a brighter, extra sustainable future.
Watch the replay of NVIDIA’s particular tackle at SC24 and be taught extra concerning the firm’s information within the SC24 on-line press equipment.
See discover concerning software program product data.