When we reported on IBM’s development of the IBM Q System One we mentioned that a key design goal was to improve system performance through the use of minimization of vibrations, custom control electronics to provide cleaner signals, and more precise temperature controls. This week IBM provided some data that showed the results of this effort.
The full data is shown in the IBM Q System One rows on our Qubit Quality page. But the highlight includes significant improvements in the gate fidelities. The average one qubit gate error rate IBM reported was 0.041%, a 62% or greater improvement over previous designs, and the average two qubit gate error rate was 1.69%, a 25% or greater improvement over the previous designs. These results are similar to what IonQ has reported with their ion trap design.
Interestingly, the coherence numbers reported for the new design did not change significantly from their previous designs with an average T1 time of 73.9 microseconds and an average T2 time of 69.1 microseconds.
IBM has created a metric which they call Quantum Volume to provide an estimate of the overall power of a quantum computer. This metric combines into one number factors related to the number of qubits, connectivity of the qubits, and gate error rates. They’ve indicated that by this measurement, their first 5 qubit machine had a Quantum Volume of 4, the current 20 qubit machines in operation on the IBM Q Network have a Quantum Volume of 8, and the IBM Q System One which will be put into operation on the IBM Q Network soon has a Quantum Volume of 16. Their goal is to double this power metric every year through design improvements. None of the other hardware vendors have adopted this metric yet and it is not clear they will ever fully agree on the definition. Other factors such as gate speed, breadth of native gates supported, and other factors are also important. However this measure is valuable because it emphasizes that continued improvements need to be made in quantity, connectivity, and error rates to achieve machines that can demonstrate quantum advantage.
For more details, you can view IBM’s news release here, their research blog article here, and the our table of Qubit Quality which compares the qubit quality parameters of several different designs here.