Dell Breaks the Latency Barrier in Quantum Computing with NVIDIA

Key takeaways:

• • Dell’s PowerEdge infrastructure hit sub-four-microsecond latency on Nvidia’s NVQLink platform
  • Independent partner testing validated the performance in real quantum workflows
  • The speed enables real-time error correction, dynamic circuits, and fault-tolerant quantum computing

Quantum computing has a timing problem. For quantum processors to deliver on their promise, they need classical computing systems that can respond in microseconds, not milliseconds (think 1,000 times faster). If they miss that window, qubits decohere, error correction fails and the quantum advantage disappears.

Dell, with NVIDIA, just proved that it can consistently hit that window.

We demonstrated and validated sub-four-microsecond average latency between Dell PowerEdge servers and field-programmable gate arrays (FPGAs) using the NVIDIA NVQLink platform. That’s fast enough to detect and correct quantum errors in real time, run dynamic circuits with classical logic inside gate-based quantum processors, and make fault-tolerant quantum computing actually feasible.

Why this matters now

Quantum computing has been stuck in a catch-22. To solve problems classical systems can’t handle, you need fault-tolerant quantum processors. To build fault-tolerant quantum processors, you need classical infrastructure fast enough to correct errors in real time. Without that infrastructure, quantum stays in the lab.

Sub-four-microsecond latency breaks that cycle. Organizations can now deploy quantum systems that actually work at scale. The same Dell infrastructure that serves as the real-time host for quantum operations can also run quantum emulation and machine learning workloads. That means enterprises can start exploring quantum use cases today without waiting for perfect quantum hardware, and then scale into production quantum as the technology matures.

The organizations moving first aren’t waiting for quantum to be ready. They’re building the infrastructure that makes quantum ready for them.

Why latency is Quantum Computing’s make-or-break challenge

Quantum processors are extraordinarily fragile. Qubits lose their quantum state in microseconds through a process called decoherence. To build useful quantum computers, you need to detect errors and apply corrections faster than decoherence happens, calibrate the system continuously, and make decisions about the next quantum operation based on measurement results from the current one.

All of this requires classical computing infrastructure that can keep pace with quantum timescales. Until now, this has been the bottleneck. Systems were either too slow or couldn’t integrate tightly enough with quantum control hardware.

The NVIDIA NVQLink platform changes that by creating a direct connection between GPU compute and the FPGAs that commonly control quantum processors. The architecture uses RoCE-based (RDMA over converged Ethernet) connections and integrates with the NVIDIA CUDA-Q software platform, bringing easy, open access to quantum control into the proven CUDA ecosystem that AI and HPC teams already know.

However, the platform is only as good as the infrastructure running it.

Dell’s role: Making Hybrid Quantum-Classical Computing real

Our XE9680, XE7745, R7715 and R770 AI server infrastructure played the role of real-time host (RTH), alongside the NVIDIA CUDA-Q Realtime API, in our lab’s NVQLink deployment. The sub-four-microsecond latency we’ve validated means this infrastructure can handle the three critical functions that make quantum processors practical:

Real-time calibration. Quantum systems drift. Our infrastructure can now detect those drifts and apply corrections continuously, reducing errors in quantum results and lowering the operational burden on quantum computing users.

Dynamic circuits. Classical logic, like if-statements and while-loops, can now execute inside gate-based quantum processors in ways that weren’t previously possible. This expands what quantum algorithms can do.

Quantum error correction. Our HPC infrastructure can detect and correct qubit decoherence as it happens. This is the pathway to fully fault-tolerant quantum computing.

Beyond the Real-Time Host role, Dell’s AI infrastructure brings additional capabilities to hybrid quantum environments. Our servers can perform quantum emulation, reducing the load on scarce quantum resources and improving throughput for the entire stack. We can also host custom machine learning models developed by Dell’s Applied Research team, adding value before and after quantum compute runs, not just during them.

Quantum machines: Partners in time

Dell partner Quantum Machines (QM) used our R7615 servers to connect to its pulse processing units, which work in tandem to control and calibrate three different QPUs across two different architectures. Using one of these Dell servers, QM demonstrated the ultra-low latency between its OPX1000 PPU and the server, effectively simulating real quantum error detection. The results validate a real-time classical compute control system leveraging GPU resources in a Dell server for a mission-critical operation.

What’s next

Over the coming months, we’ll validate the NVQLink platform across additional Dell offerings and test it in co-located deployments with real quantum control systems.

Quantum computing is moving from research to reality. The organizations that get there first will be the ones that solve the integration challenge between quantum and classical systems. Dell is solving it at the speed quantum computing demands.

Building quantum infrastructure? Let’s talk about what single-digit microsecond latency means for your roadmap.

^*Learn how Dell AI Factory with NVIDIA delivers a comprehensive and secure AI solution customizable for any business.