Skip to content
RESEARCH · QUANTUM SIMULATION

Better coordination helps eight GPUs test quantum error correction

AMD and IBM researchers report faster error-correction simulations by reorganizing preparation and results handling.

Published Source report Sources checked
Reading tools for text size, copying, printing and your reading list
My reading list

Adding more graphics processors does not automatically make quantum-computing research faster. A November 12 preprint from AMD and IBM tests how the ordinary computer coordinating those processors can hold them back.[1]

The team reorganized Pêlagos, software for simulating . It prepares work in parallel, gives each GPU its own persistent process and returns large results through shared memory.[1][2]

With eight AMD MI300X GPUs, the revised arrangement delivered about 2.64 times the throughput for the smaller test code and 1.48 times for the larger one. Both comparisons use the same eight-GPU hardware.[1]

The paper’s larger eight-versus-one figures answer a different question: total work grows with GPU count. They do not measure how much sooner a fixed job finishes.[1]

This is simulation, not real-time correction of a live quantum computer. Tests covered two codes, one noise setting and three measured repetitions after warm-up.[1]

For researchers, the useful lesson is to count preparation and results handling alongside processor speed.

Keep your reading trail

Evidence & context

2 linked sources

Research preprint; two simulated error-correction codes; no independent reproduction

How to read the evidence and dates

An announcement records what an organization stated. An experiment reports an observation. A simulation or resource estimate depends on its assumptions. QubitWire’s analysis connects these sources; the labels are not quality scores.

Published is the article’s first QubitWire publication. Updated marks a substantive content change. Sources checked records a review of the underlying evidence. An older paper can support a new explanation, and another press release about the same work is not independent replication.

QubitWire take

Preparation and result handling belong in the computing budget, not outside the performance comparison.

Sources

  1. Scaling Pêlagos: Multi-GPU Orchestration for Relay Belief-Propagation Decoding

    Preprint v1 submitted November 12, 2026 at 04:36:35 UTC. Sections III–IV and Table III checked. Matched-eight-GPU gains, warm-cache repetitions and scaled-work limitations retained; no independent reproduction.

  2. Pêlagos process-backend implementation notes

    Cited public source revision inspected. Persistent processes, shared-memory outputs and existing kernels documented. Initial public commit already includes this backend; not evidence of a newly shipped update.

Keep up with these topics

Explore the subjects covered in this story.

READ NEXT

Continue exploring

All Research stories
  1. More coverage of IBM

    Small quantum-training test cuts return trips to the cloud

    Three small IBM-processor experiments show why how work is submitted can matter, not just how many measurements run.

    Published · Source report · Sources checked · About 2 min read