Start writing here...
Ah yes — Quantum Volume (QV) and benchmarking metrics are essential if you're trying to figure out whether a quantum computer is actually useful or just flashy qubits with no bite. 🧠⚙️ Let’s unpack what QV is all about and explore the other metrics that matter for evaluating real quantum performance.
🔹 What Is Quantum Volume?
Quantum Volume is a holistic benchmarking metric proposed by IBM to quantify the effective computational power of a quantum computer.
It’s not just about how many qubits you have — it’s about how well they work together.
✅ Quantum Volume Measures:
| Component | What It Tests |
|---|---|
| 🧮 Qubit count | More qubits = more complexity |
| ⚡ Gate fidelity | How accurately gates are applied |
| 🔗 Connectivity | Can qubits talk to the right neighbors? |
| ⏱ Circuit depth | How many gate layers can be applied before decoherence kills you? |
| 🧼 Noise resilience | Can the system tolerate noise and still give correct-ish answers? |
📏 QV = 2ⁿ, where n is the largest number of qubits that can run a random circuit of depth n with high fidelity (≥ 2/3 success probability).
🔹 Why Quantum Volume Matters
QV is useful because:
- It captures the practical power of a quantum device, not just theoretical specs.
- It can compare very different quantum systems (superconducting vs trapped ions, etc).
- It's one of the first attempts at a standardized quantum benchmark.
🔹 Quantum Volume Milestones
| Year | Company | QV Achieved | Notes |
|---|---|---|---|
| 2019 | IBM | 16 | First public benchmark |
| 2020 | IBM | 64 | Rapid improvement |
| 2021 | Honeywell (now Quantinuum) | 128 | Beating IBM briefly |
| 2022 | IBM | 128–512 | Pushed into higher QV territory |
| 2023–24 | Quantinuum | 2,048+ | Current leader (with error-mitigated trapped ion systems) |
⚠️ Note: These numbers keep growing, but they rely on random circuit success rate, so real-world performance may vary by task.
🔹 Other Quantum Benchmarking Metrics
📌 1. Fidelity Metrics
| Metric | Description |
|---|---|
| Gate fidelity | Accuracy of individual quantum gates (1-qubit and 2-qubit) |
| Readout fidelity | Accuracy of measuring qubit states |
| State fidelity | Overlap between actual and ideal quantum state |
Target: >99.9% for 1-qubit, >99% for 2-qubit gates
📌 2. Coherence Times (T₁, T₂)
| Metric | Meaning |
|---|---|
| T₁ (relaxation) | Time a qubit stays in |
| T₂ (decoherence) | Time a qubit retains quantum phase info |
Longer T₁ and T₂ = better performance, especially for deeper circuits
📌 3. Crosstalk & Connectivity
- Crosstalk: How much other qubits/gates interfere with each other
- Connectivity: How many qubits each qubit can interact with directly (important for mapping circuits efficiently)
📌 4. Benchmarking Protocols
| Protocol | Use Case |
|---|---|
| Randomized Benchmarking (RB) | Estimates average gate error |
| Cross-Entropy Benchmarking (XEB) | Used by Google in supremacy tests |
| Cycle Benchmarking | Tracks errors across circuit layers |
| Mirror Circuits / Heavy Output Generation | Used in Quantum Volume measurement |
📌 5. Application-Oriented Benchmarks
- AIQ Benchmarks (Application-Inspired Quantum benchmarks)
- SupermarQ: Benchmarks designed to reflect real application classes, like optimization or chemistry
- MLPerf-Q: Under development for quantum machine learning tasks
These are more task-specific, as opposed to QV’s general-purpose measurement.
🔹 TL;DR Summary
| Metric | What It Tells You | Good For |
|---|---|---|
| Quantum Volume | Overall quantum capability | General-purpose comparison |
| Gate Fidelity | Gate quality | Hardware calibration |
| T₁/T₂ Times | Qubit stability | Feasibility of deeper circuits |
| Connectivity | Circuit mapping efficiency | Algorithm execution |
| App-Specific Benchmarks | Real-world task performance | Vertical use cases (QML, QAOA, VQE) |
🧠 Big Picture
Quantum Volume is great for getting a sense of how robust a quantum system is, but application benchmarks will ultimately matter more as quantum computers move toward real-world use.
Wanna compare two specific quantum platforms by their QV or gate fidelities? Or see how QV relates to running something like VQE or QAOA?