- Nvidia anticipates that the Isambard 3 will reach peak speeds of 2.7 petaflops when performing FP64 calculations. A petaflop is equivalent to one quadrillion computing operations per second.
- According to Nvidia, the system will support AI, life sciences, medicine, biotechnology, and astrophysics research initiatives. It can model big, complicated things like wind turbines and biological systems at the molecular level.
Nvidia Corp. recently announced three projects where its chips will be used to help scientists advance in scientific discovery.
In one of the initiatives, the University of Bristol will use Nvidia’s Grace CPU Superchip processors to power a supercomputer. The other two projects are concerned with quantum computing. The chipmaker discussed all three projects at the ISC 2023 conference, held recently in Hamburg.
The University of Bristol in the UK is in charge of a project to build a new supercomputer that is better for research tasks. The supercomputer will be called Isambard 3. Hewlett Packard Enterprise Co. will construct it using Nvidia silicon.
According to the chipmaker, Isambard 3 will include 384 Grace CPU Superchip processors. Each of them includes 144 CPU cores based on Arm Ltd. design. The cores support SVE, which is a technology in the Arm instruction set that makes it easier for CPUs to run tasks that use artificial intelligence.
Nvidia anticipates that the Isambard 3 will reach peak speeds of 2.7 petaflops when performing FP64 calculations. A petaflop is equivalent to one quadrillion computing operations per second. FP64 is a data format that supercomputers commonly use to organize the information they process.
Scientists intend to use Isambard 3 for a variety of purposes. According to Nvidia, the system will support AI, medicine, life sciences, biotechnology, and astrophysics research initiatives. It can model big, complicated things like wind turbines and biological systems at the molecular level.
Isambard 3 is set to launch in 2024. Once operational, the system is expected to consume less than 270 kilowatts of power, making it one of Europe’s most energy-efficient supercomputers. According to Nvidia, Isambard 3 is six times more energy-efficient than the previous-generation supercomputer at the University of Bristol.
Ian Buck, Vice President of hyperscale and high-performance computing at Nvidia, said, “As climate change becomes an increasingly existential problem, it’s vital for computing to embrace energy-efficient technologies. NVIDIA is working alongside the Arm Neoverse ecosystem to provide a path forward for the creation of more energy-efficient supercomputing centers, driving important breakthroughs in scientific and industrial research.”
New Quantum Computing Initiatives
Nvidia also detailed how its technology assists two quantum computing initiatives in the United Kingdom and Germany recently at ISC 2023.
Rolls-Royce Holdings plc is leading the first initiative in collaboration with Classiq Technologies Inc., a quantum computing startup. The project aims to help Rolls-Royce develop more efficient jet engines. The company hopes to use quantum computers to make the work of its engineers easier.
Rolls-Royce collaborated with Classiq to develop and simulate a quantum circuit using a set of Nvidia A100 graphics cards. That type of specialized algorithm could run on a quantum computer in the future. The algorithm developed by Rolls-Royce has 10 million “layers” and 39 qubits.
The quantum circuit, according to Nvidia, is designed to perform calculations related to computational fluid dynamics, a scientific field that focuses on simulating complex physical phenomena. The quantum circuit made by Rolls-Royce is said to be the largest one of its kind made for computational fluid dynamics.
Ian Buck commented, “Designing jet engines, which are one of the most complicated devices on earth, is expensive and computationally challenging. NVIDIA’s quantum computing platform gives Rolls-Royce a potential path to tackle these problems head on while accelerating its research and future development of more efficient jet engines.”
The second quantum computing initiative announced recently by Nvidia is a partnership with FZJ, a German research institute. FZJ is collaborating with Nvidia to create a new lab to house what the companies call a classical-quantum supercomputer. The system will incorporate both classical and quantum computing components.
Nvidia says that the machine will be made by ParTec AG, a company in Munich. In turn, FZJ researchers will simulate quantum processors on the system using Nvidia’s cuQuantum and CUDA Quantum software tools. FZJ already runs a traditional supercomputer with 3,744 Nvidia A100 graphics cards.