A study by AWS, Nvidia, and others proposes a model to assess quantum processor needs in relation to supercomputing resources.
A collaborative study by researchers from Amazon Web Services (AWS), Nvidia, Lawrence Berkeley National Laboratory, and NASA has introduced a model aimed at assessing the relationship between quantum processors and classical supercomputers. This framework is designed to help determine the optimal placement of quantum hardware in proximity to high-performance computing resources versus utilizing cloud access.
Task-Specific Requirements
The researchers have categorized the interaction between quantum and classical systems into two distinct levels. The first level focuses on real-time management tasks, such as calibrating qubits and processing quantum error correction data. These tasks require a direct connection with minimal latency, as they need to be executed within microsecond intervals.
The second level includes hybrid algorithms where both classical and quantum processors alternate in performing computations. While latency can influence the overall duration of these tasks, it does not necessarily impede the execution of the algorithm, allowing some processes to utilize cloud access.
Limitations of Remote Access
The authors analyzed the execution time of hybrid algorithms, noting that if a significant portion of the computation involves classical processing, reducing network latency may not substantially enhance performance. Conversely, algorithms that demand frequent data exchanges necessitate a fast connection.
For instance, quantum diagonalization based on sampling (SQD), tested on an IBM Heron processor with 77 qubits, demonstrated that the majority of processing time was dedicated to classical computations, making remote access viable. In contrast, the Markov chain Monte Carlo method with quantum acceleration (QE-MCMC), tested with a 10-qubit IBM device, required numerous sequential exchanges, leading to communication costs that were 1000 times greater than computation time, although it did not necessitate a full supercomputer.
Future Demands for Error Correction
Looking ahead, as the field progresses toward fault-tolerant quantum computing and logical qubits, the reliance on classical infrastructure is expected to increase. Continuous measurement processing and error correction will become critical, with the speed of classical components likely dictating the operation pace of quantum systems.
The authors predict that large-scale fault-tolerant quantum setups will necessitate closer integration with high-performance computing systems. They emphasize that infrastructure requirements will differ based on various factors, including algorithm type, hardware characteristics, and data exchange volume. The model should be regularly reassessed to adapt to evolving quantum processors and computational techniques.
In July, Nvidia also introduced code for an AI module aimed at preprocessing quantum error signals, which, when combined with a classical decoder, improved the efficiency of logical failure rates and processing speed in simulations.
