SDSignal Desk

Quantum processors get a standard plug into supercomputer schedulers, courtesy of Slurm 26.11

· Updated · Signal Desk newsroom

Analysis of reporting by NVIDIA Blog. How we work

Image: NVIDIA Blog

A new Slurm release lets quantum hardware makers describe their machines once in a simple file and slot into any supercomputing center’s scheduler. It’s plumbing, and plumbing is what quantum-HPC has been missing.

Key takeaways

  • Nvidia says nearly 30 supercomputing centers are deploying more than 50 quantum processors, a nearly sevenfold rise over five years.
  • Slurm 26.11, due later this month, lets QPU builders define a single interface to their hardware through a simple file.
  • Alice & Bob, IBM, IQM, Pasqal, Quantinuum and MQSC are adopting it; it supports the QRMI and QDMI interfaces.
  • Nvidia paired it with CUDA-Q updates, including a decoder it claims is 13,000x faster than the industry standard.
  • Fermilab says CUDA-Q let it finish studies in three weeks that would normally take five months.

The integration problem

Quantum computing’s practical path runs through supercomputing centers, where QPUs sit beside GPUs and CPUs and handle the pieces of a problem classical machines struggle with. Nvidia says those deployments have grown nearly sevenfold in five years, with nearly 30 centers now deploying more than 50 quantum processors. The friction is software. Until now, wiring a new QPU into a center meant bespoke integration work for every combination of hardware and site.

Slurm, the open-source workload manager many of those centers already use, is the obvious place to fix that. The new release, building on earlier work with IBM, Pasqal and others, lets QPU builders describe their hardware once in a simple file and reuse that integration everywhere. Centers can add quantum processors to existing scheduling and resource-management workflows instead of building custom solutions. Alice & Bob, IBM, IQM, Pasqal, Quantinuum and MQSC are adopting it, and it supports the QRMI and QDMI quantum-HPC interfaces. Slurm 26.11 is due later this month.

Why this matters more than another qubit record

We tend to cover quantum through hardware milestones. This announcement is a reminder that adoption depends on boring things like job schedulers. Computational chemists, nuclear physicists and materials scientists already submit work through Slurm. If a quantum processor shows up as just another schedulable resource, the barrier to trying hybrid algorithms drops sharply. That is how GPUs went mainstream in science, a history Nvidia is marking this week with CUDA’s 20th anniversary.

There is a strategic layer too. Slurm is community-built, but Nvidia positions this as the orchestration layer of its open quantum-GPU platform centered on CUDA-Q. Hardware makers keep their independence, while Nvidia’s software becomes the default way to write and run hybrid applications. That is a smart position for a company that doesn’t build QPUs.

The CUDA-Q updates

Alongside Slurm, Nvidia announced CUDA-Q releases at SC26. CUDA-Q QEC adds a decoder server and a TCBP-JOBS decoder Nvidia claims is 13,000x faster than the industry standard, without naming the baseline in the post. CUDA-Q adds fault-tolerant primitives; Fermilab reportedly finished studies in three weeks that would normally take five months. CUDA-Q Algorithms adds reusable components, including tools for chemistry and materials, and CUDA-Q Pulse helps hardware builders shape qubit control pulses with GPU acceleration.

Partners are building on it: Quantum Motion integrated its QUICHE chemistry framework, and Altera with Oxford Quantum Circuits built an FPGA controller linking qubit control to GPUs via NVQLink. At scale, QunaSys used 1,024 Hopper GPUs on ABCI-Q to simulate more than 300 million quantum operations and reach chemical accuracy for a four-hydrogen-atom model, and Nvidia researchers simulated 104 qubits under realistic noise on JUPITER. Those simulations underline an uncomfortable truth: GPUs are still doing much of the useful quantum research today.

Who it affects

Supercomputing centers
Can add QPUs to existing Slurm workflows without bespoke integration for each device.
QPU makers
One integration file reaches many sites, lowering deployment costs.
Domain scientists
Hybrid quantum-classical jobs become schedulable through familiar tools.
Nvidia
CUDA-Q becomes the default programming layer for hybrid workflows, even without its own QPU.

What to watch

  1. The Slurm 26.11 release later this month and which centers turn on QPU scheduling first.
  2. Additional QPU makers adopting the integration file approach.
  3. Independent verification of the 13,000x decoder claim and its baseline.
  4. Published science results from hybrid jobs scheduled through Slurm.

FAQ

What changed in Slurm?

Slurm 26.11 lets quantum processor makers define a single interface to their hardware through a simple file that works across supercomputing centers.

Who is adopting it?

Alice & Bob, IBM, IQM, Pasqal, Quantinuum and MQSC, according to Nvidia.

When is it available?

Nvidia says Slurm 26.11 will be available later this month.

How does Nvidia fit in?

Nvidia frames Slurm as the orchestration layer for its CUDA-Q hybrid quantum-classical platform, which also received updates at SC26.

Read the original

Continue at the source.

NVIDIA Blog

Companies: NVIDIA

More on NVIDIA

All NVIDIA coverage →