Well, a portion of the issue is the disparity between claim and granted credit. So, when I turned HT off, I narrowed that gap significantly. But, I agree, I was using HT before mini, so I'm not proving much here.

Just exploring the thought that L2 cache may be one way mini might be running differently, and introducing such a credit difference. We know from past discussions that L2 cache, or lack thereof, seemed to be why many AMD CPUs don't seem to run as high a RAC as a similar Intel CPU. This was true before mini came on the scene, so I am just testing the extension of that theory.

We know the old app was L2 cache sensitive. This seems to say the cache was rather near full, or at least says L2 cache preferred exceeded the size on many AMD CPUs. So, if the L2 cache usage increased just slightly, it could cause lots of new waiting for memory access. And hence the lower credit.

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

Hi guys, just saw a blurb in another thread about hyperthreading and that it slows BOINC projects.