Task 16101313 has Spikes=4, Autocorr=26; and
Task 16101312 has Spikes=2, Autocorr=3.

[I know the 3rd task hasn't returned yes, so I wouldn't know who's "'more' right" at the moment...]

---

...[I know the 3rd task hasn't returned yes, so I wouldn't know who's "'more' right" at the moment...]

Yours is 'more right'. There's nothing in your host task history that would seem to indicate any particular CPU unreliability, and the app is far more mature than the ARM one there (7.26) . Usually (but not always) more signals especially leading to overflow is a warning sign.

Could be in for a wait to confirm with the new Wingman. If you wanted to, one way to add more certainty would be to get Lunatic's bench setup on the Windows host, and drop in the task for a reference CPU app run, and other available and compatible GPU & CPU apps for comparison.

In my view having controlled lab conditions trumps all the above guessing.

---

Chaos: When the present determines the future, but the approximate present does not approximately determine the future.
Edward Lorenz