Nvidia announced Quantum-2 technology, an end-to-end networking platform, at its event today. In addition, the company announced that they broke the record for the fastest solution to the MaxCut problem with the cuQuantum SDK.
Nvidia Quantum-2 technology introduced today is defined as a 400 Gbps InfiniBand network platform (DPU) consisting of NVIDIA Quantum-2 switch, ConnectX-7 network adapter, BlueField-3 data processing unit.
Nvidia announces Quantum-2 technology
NVIDIA Quantum-2 has the essentials for demanding workloads. Powered by cloud-native technologies, it provides high performance with a transfer rate of 400 Gigabits per second to accommodate many users.
Gilad Shainer, SVP, NVIDIA Networking, said, “The requirements of today’s supercomputing centers and public clouds are converging. They must provide the highest possible performance for next-generation HPC, AI, and data analytics challenges, while also securely isolating workloads and responding to changing user traffic demands. Modern data centers will be able to have this vision with NVIDIA Quantum-2 InfiniBand. ” said.
With 400 Gbps, NVIDIA Quantum-2 InfiniBand offers double the network speed and triple the number of network ports. The technology, which accelerates performance by 3 times and reduces the need for data center fabric switches by 6 times, helps reduce data center power consumption.
NVIDIA Quantum-2 SHARPv3 Intra-Network Computing technology provides 32x more acceleration engines for AI applications compared to the previous generation. NVIDIA UFM Cyber-AI platform available for advanced InfiniBand fabric management in data centers, including predictive maintenance.
A nanosecond-precision timing system integrated into NVIDIA Quantum-2 will help reduce the overhead of waiting and idle times by synchronizing distributed applications such as database processing. This new feature allows cloud data centers to become part of the telecommunications network and host software-defined 5G radio services.
At the heart of the Quantum-2 platform is the new Quantum-2 InfiniBand switch. With 57 billion transistors on 7 nanometer silicon, it’s slightly larger than the NVIDIA A100 GPU with 54 billion transistors.
Nvidia breaks world record with cuQuantum
Transferring new developments about the world of quantum computing, Nvidia shared the record they broke with the audience. In addition, the company stated that it can develop and test quantum algorithms faster by simulating future quantum computers in today’s classical systems.
Using its self-developed cuQuantum SDK to speed up quantum circuit simulations on the GPU, the company has created the largest quantum algorithm simulation ever to solve the MaxCut problem.
MaxCut algorithms are used to design large computer networks, find the optimal chip layout with billions of silicon paths, and explore the field of statistical physics. MaxCut is known as one of the most preferred algorithms to show the advantage of using a quantum algorithm.
To solve the MaxCut problem, the cuTensorNet library from cuQuantum running on NVIDIA’s in-house supercomputer Selene was used. Using GPUs to simulate 1,688 qubits, a graph with 3,375 vertices was created. This operation was performed with 8 times more qubits than the previous largest quantum simulation.
21794
