10-06-2026, 11:06 AM
That matches what I was seeing. I don't think your conclusion is far off. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.
measure twice, benchmark three times