We had previously reported on IBM’s Quantum Volume metric and their goal of achieving a doubling of this measure every year. This factor takes into account a number of factors including qubit count, qubit quality, qubit connectivity, crosstalk considerations and a number of other factors to provide a relative figure of merit for a quantum computer design. This metric is technology agnostic and could conceivable be used by other gate level quantum computer developers.
In March of 2019, they announced that they had increased this factor to 16 with the announcement of their IBM Q System One. Now, they have announced they have doubled this once again to 32 with a new 28 qubit design called Raleigh. This design combines the lattice structure of the 53 qubit design (the 28 qubit would appear to look roughly like half of the 53 qubit design) that they introduced last year with additional upgrades implemented in some of the latter versions of the 20 qubit design.
Although it might not seem obvious why a 28 qubit part would show a higher metric of their 53 qubit design, the answer is that the Quantum Volume metric assumes a square circuit of m qubits with a depth of m gates. And the limiting factor right now appears to be the gate depth that can be used before the errors become too great. For more on this, you can view IBM’s paper describing the quantum volume metric and measurement methodology here.
IBM has prepared a good blog article that describes their generation cycles of learning and plans to provide continued improvement. Among other things they will take some of the advancements created in this 28 qubit design and apply it to subsequent generations of the 53 qubit design. In addition, they have several other improvements ideas generated from their research that they intend to apply in the future to further improve the qubit quality.
One thing to mention is that the nature of the Quantum Volume metric is to treat equal importance to the width of the qubits and the gate depth so that both are equal to achieve essentially a square circuit configuration. However, it is not clear how many quantum algorithms are configured this way. Certain algorithms being researched for NISQ applications, such as QAOA, are called “short depth” algorithms where the number of qubits may be significantly larger than the gate depth to minimize the effect of decoherence. For more, you can view IBM’s latest blog describing this new development here.
January 8, 2020