CLOPS and Speed Benchmarking
CLOPS measures how fast a quantum processor runs circuits, capturing throughput that fidelity-based metrics like quantum volume ignore.
Speed matters too
Quantum volume says how large a circuit a machine can run correctly, but not how quickly. For real workloads, especially variational algorithms that run thousands of circuit repetitions in a classical optimization loop, throughput is decisive. Circuit Layer Operations Per Second (CLOPS) is a benchmark that measures how many layers of quantum volume circuits a device executes per unit time.
What CLOPS captures
- Gate execution time on the hardware
- Readout and reset time between shots
- Classical latency of loading circuits and processing results
- Compilation and parameter-update overhead in variational loops
Why the full stack counts
A machine can have fast gates yet slow overall throughput if reset, readout, data transfer, or classical control introduce delays between circuits. CLOPS deliberately includes this end-to-end overhead, so it rewards tight integration of the control system and low-latency classical processing, not just the qubits themselves.
Modality differences
Superconducting qubits have nanosecond gates and can post high CLOPS, while trapped ions have microsecond gates and shuttling delays that lower throughput even where their fidelity is higher. This makes speed and fidelity a genuine trade-off across platforms, and why both numbers, not one, are needed to compare machines fairly.
Speed and quality are orthogonal axes. A useful quantum computer must be both large enough (quantum volume), accurate enough (gate fidelity), and fast enough (CLOPS) for the target workload, which is why the field now reports a suite of complementary benchmarks rather than a single score.