Qiskit/qiskit-ibm-runtime

IBM Kyiv reports double the usage time as other backends

Closed this issue · 1 comments

nonhermitian commented

Describe the bug
IBM Kyiv reports double the usage time as other backends for the exact same workloads.

Eg. on Kyiv the job csch754zx1qg008mazy0 is reporting 9 min and 59 sec worth of runtime usage

But one other backends, eg. Cusco (csce4s3tj7dg008ccsjg) and Osaka (csccga1x35wg0080ztag) the runtime for the exact same workload is 4min 37 sec and 5min 0 sec, respectively.

Steps to reproduce

Expected behavior
Given that all the systems tested have the same reported CLOPS (speed) numbers, I would expect only a small variation in the reported runtimes, not factors of 2.

Suggested solutions
Isolate why Kyiv is reporting twice the runtime, change the reported CLOPS value, or provide some indication to end users that executions on some seemingly similar backends will have markedly different runtimes and thus costs associated with it.

Additional Information

  • qiskit-ibm-runtime version:
  • Python version:
  • Operating system:
nonhermitian commented

The answer is that this is to be expected given the reset mechanism and the way the control electronics works on Kyiv. This is not an issue on later devices.