As quantum computing advances with the number of qubits and complexity, superficial integrations with classical resources are no longer enough. Deep hardware-level integration is mandatory for quantum computer operations, particularly quantum error correction and the mitigation of drifting analog parameters.

To meet this need, NVIDIA and Quantum Machines co-developed NVIDIA DGX Quantum, the first hardware-based solution to seamlessly integrate GPU/CPU accelerators with quantum control.

High Performance and Scalable System

NVIDIA DGX Quantum integrates the OPX1000 quantum controller with NVIDIA’s Grace Hopper (GH) server through a low-latency PCIe interface (OPNIC), enabling data transfer from the quantum controller to the server (GPU and CPU) and back, in a round-trip latency of under 4 μs, at least a 1,000x improvement in latency over previous state-of-the-art integrations.

A single NVIDIA DGX Quantum fits in a 5U 19” rack form factor and scales effortlessly in both quantum and classical domains:

  • Expand quantum control by adding additional OPX1000 controllers to support larger QPUs.
  • Increase classical compute by integrating more GH servers to meet advanced processing demands.

NVIDIA DGX Quantum is the only solution engineered for scaling from small experiments to full-scale quantum-accelerated supercomputing.

Check out this NVIDIA DGX Quantum tutorial in which reinforcement learning is used to calibrate parameters on a Rigetti superconducting processor. 

NVIDIA DGX Quantum Hardware

OPX1000

OPX1000 is the best-in-industry hybrid control platform. At its core, the unique Hybrid Processing Unit (HPU) delivers unmatched classical processing power close to the qubits, enabling real-time execution of demanding adaptive algorithms like quantum error correction at scale.

OPNIC

OPNIC is the low latency PCIe interface card that handles data transfer between the OPX1000 and the NVIDIA superchip. OPNIC enables multiple data streams of high-bandwidth between the different computational units of a DGX Quantum solution.

Grace Hopper

The NVIDIA GH server is powered by NVIDIA’s GH200 Grace Hopper superchip. This superchip integrates a high-performance GPU with a custom CPU. It boosts quantum control, ensuring the availability of the highest-performing classical resources within quantum circuits.

NVIDIA DGX Quantum Software

Optimized quantum-classical hardware requires optimized quantum-classical software. NVIDIA DGX Quantum includes NVIDIA’s CUDA-Q, an open-source quantum-classical programming platform, and Quantum Machines’ QUA, a powerful open-source quantum programming language.

CUDA-Q

CUDA-Q is a first-of-its-kind platform for integrating and programming quantum processing units (QPUs), GPUs, and CPUs in a single system, designed for hybrid workloads.

QUA

QUA unifies universal quantum operations in their ‘raw’ format – all the pulse-level operations used to control and measure qubits – with universal classical operations used in classical processing.

With NVIDIA DGX Quantum, classical acceleration routines can be called directly from the QUA pulse-level quantum program for execution on NVIDIA’s GH Superchip. Likewise, QUA routines can be called from CUDA-Q.

Key Features

  • Combines the best of quantum control – Quantum Machines’ OPX1000 Hybrid Control platform with the best acceleration solution – NVIDIA GH200
  • Data round trip latency of < 4 μs
  • Unified quantum – classical development platform with QUA and CUDA-Q
  • 5U rack-mounted system with hot-swappable components and redundancies

Benefits

Run the Most Advanced Hybrid Algorithms

The only solution designed to support QEC and ultra-fast calibration at scale.

Effortless Scalability

Add OPX1000 controllers as your QPU grows, and add GH200 servers to meet increasing computational demands.

HPC and Data Center Ready

HPC-QC integration from the hardware quantum control layer and up, designed for data center reliability with redundancies and hot swappable critical parts.

Compatible with All Qubit Technologies

OPX1000 controller supports superconducting qubits, spin qubits, neutral atoms, trapped ions, defect centers, photonic qubits, and more.