Riverlane Cuts QEC Latency 10x Faster Than Google’s Results

Riverlane has achieved a mean latency of 16.32µs for quantum error correction on real quantum processing unit data, a result that outperforms the company’s own 20µs target and brings utility-scale quantum computing closer to reality. The performance, demonstrated using data from Google’s 2024 research, represents a major step toward fault-tolerant systems capable of executing trillions of reliable quantum operations. During a distance 5 rotated surface code quantum memory experiment, Deltaflow 2 delivered this latency, achieving speeds over 10 times faster than Google’s results. Riverlane reports that using Deltaflow to outperform Google’s published latency by a wide margin is an exciting result and a step toward fault tolerance at utility scale, though the company notes the comparison isn’t strictly like-for-like given differing research priorities. Deltaflow 2 Achieves 16.32µs Latency on Google’s QPU Data The implications extend beyond speed, addressing a critical bottleneck preventing quantum computers from tackling complex, real-world problems. This achievement centers on Deltaflow 2’s ability to process and decode quantum information with swiftness, delivering a sub-shot latency, the time to process a single data window, over ten times faster than results published by Google. Riverlane reports that its maximum sub-shot latency is more than 10 times better than Google’s, highlighting the stability of this performance across a million experimental rounds. This isn’t simply about faster processing; it’s about maintaining a continuous, reliable stream of error-corrected data essential for prolonged quantum computations. Riverlane utilized its QPU and control system emulator, coupled with QECi, to feed Google’s 2024 Willow experiment data into Deltaflow, allowing for a direct performance comparison without requiring a physical quantum processing unit connection. The resulting mean latency of 16.32µs contrasts sharply with Google’s reported 63µs ± 17µs, representing nearly a fourfold improvement. This speed is enabled by several key innovations within Deltaflow 2, including a proprietary Local Clustering Decoder (LCD) implemented on Field Programmable Gate Arrays (FPGAs). According to Riverlane, the LCD can reduce the number of physical qubits required for a logical qubit by a factor of four under a leakage-dominated noise model, while decoding in under 1µs per round. A unique windowing scheme allows for continuous, streaming decoding, preventing data backlogs and maintaining low latency. Riverlane states that these results emphasize the hardware-first approach that underpins this progress. The ability to achieve this level of performance on real QPU data, and to demonstrate stable sub-shot latencies, positions Deltaflow 2 as a crucial component in the pursuit of fault-tolerant, utility-scale quantum computing, enabling increasingly complex quantum algorithms and applications. Real-time QEC Critical for Utility-Scale Quantum Computing Quantum error correction remains the central challenge in realizing practical, “utility-scale” quantum computing, the threshold where these machines can reliably perform complex calculations beyond the capabilities of classical computers. Achieving this requires not merely correcting errors, but doing so in real-time, processing vast amounts of quantum data with minimal delay and resource consumption. Riverlane’s Deltaflow 2 system is demonstrating significant progress in this area, tackling the bottleneck of ultra-low latency that is crucial for executing trillions of reliable quantum operations. This result surpasses Riverlane’s internal 20µs target and brings the technology closer to the approximately 10µs latency considered essential for utility-scale applications. The system’s ability to maintain stable sub-shot latencies across one million rounds underscores its reliability for sustained operations. Several key innovations within Deltaflow 2 contribute to this performance, including a Local Clustering Decoder (LCD), a hardware solution implemented on FPGAs, designed for both speed and accuracy. Riverlane also implemented a proprietary windowing scheme, allowing for continuous, streaming decoding rather than waiting for complete computation, and a data routing system capable of handling input from diverse QPU and control systems. The importance of low latency extends beyond algorithmic speed; for example, consider a magic state teleportation experiment, a crucial operation for utility-scale quantum computing, where the time it takes the QEC system to issue a correction directly determines the logical clock speed of the system. Riverlane explains that the faster the system can issue this result to perform the S gate, the faster the quantum computer can execute operations per unit of time. Indeed, factoring a 2048-bit RSA integer would require significantly more time with increased latency, demonstrating that Deltaflow 2’s advancements are not just incremental improvements, but critical steps toward unlocking the full potential of quantum computation. For example, researchers have shown that factoring 2048-bit RSA integers could take 8 hours with a 10µs decoding response time, but a 100µs response time would slow it down by more than sixfold.

Local Clustering Decoder Enables Fast Decoding Performance Riverlane is pushing the boundaries of quantum error correction with Deltaflow 2, a real-time QEC system achieving increasingly critical performance benchmarks. This figure not only surpasses Riverlane’s internal 20µs target but also brings the prospect of “utility-scale” quantum computing, capable of trillions of reliable quantum operations, noticeably closer. The achievement centers on minimizing the delay between receiving qubit measurement data and issuing error corrections, a bottleneck that has long hampered progress toward fault-tolerant quantum computers. The system’s design prioritizes flexibility, accommodating data from various QPU and control system configurations without compromising performance. The comparison to Google’s 2024 work is direct, utilizing the same dataset to provide a clear performance benchmark, with Riverlane’s results showing a mean latency nearly four times lower than Google’s reported 63µs ± 17µs. While acknowledging that a direct comparison isn’t entirely straightforward, noting that latency wasn’t the primary focus of Google’s research, Riverlane emphasizes the significance of their improvement. The implications extend beyond mere speed; low latency is fundamental to achieving fast logical clock rates and enabling complex quantum operations like magic state teleportation, where the speed of error correction directly impacts the overall computational throughput. The company explains that this is a massive data management issue requiring a high-accuracy, low-latency QEC cycle. This performance isn’t merely incremental; it demonstrates a critical capability for maintaining continuous quantum operations, a feat previously hampered by processing bottlenecks. The core of this advancement lies in Deltaflow 2’s innovative approach to data handling, specifically its proprietary windowing scheme, which efficiently splits the decoding graph to enable continuous, streaming error correction and prevent data backlogs. Sub-shot Latency Stability Validated Over One Million Rounds The pursuit of fault-tolerant quantum computing often focuses on increasing qubit counts and coherence times, but latency is rapidly emerging as a critical bottleneck. While many assume scaling qubit numbers is the sole path to “utility-scale” systems, Riverlane has demonstrated that consistently fast error correction is equally vital, recently validating sub-shot latency stability over an impressive one million rounds of computation. The experiment involved processing raw readout data from Google’s Willow experiment, formatted using the QECi Specification and fed into Deltaflow via a QPU and control system emulator, and QECi. This demonstrates that Deltaflow 2 is not just a technical demonstration, but a crucial step toward building quantum computers capable of tackling real-world problems. Source: https://www.riverlane.com/news/riverlane-s-real-time-qec-system-performance Tags: