Ocing is lengthy here atm, tring to breack the 2.5GHz wall my be has, but the freeze occures only after ~3hrs.
Welcome to the "core freezes" club!



Me: 2665MHz froze after a few hours. Guess how it froze?

I had nothing but the installation of WMP9 open (idling) when I was ticking file association options -> it was idling 118W, and then, first the screen background applications froze with the mouse pointer still moving, so I looked over to the power meter and it suddenly jumped to 199-200W AC and then the mouse pointer eventually froze. It gets stuck at 199-200W AC until I reboot.

This is the problem nearly all of us are facing -> we need to collaborate and hunt this down, as to why and how to alleviate it. Unfortunately many users will have to carefully test this as well as me, or we'll make no progress. Remember those settings of mine are still around 100MHz below what I have max P95/Memtest 20hr stable.

Can this be the idle clock speed fluctuation I talked about much earlier (Achim caught it too)?

I think it must be. When you oc and at higher MHz your system freezes, the same reaction typically occurs -> jump in power consumption.

I'll now have to open a monitoring tool and just stare at it for hours to see the various speeds when it locks up.

Quote Originally Posted by justapost View Post
What did your DMM probe read at 125°?
The MOSFETs at 75C CPU with 1.446V VCore measured 125C with the DMM.

Never had the right setup to measure the CPU temp.
There is no peltier effect between IHS and the cooler i assume.
If the sensor is calibrated at 60° and the scaling is too high, it would explain too low idle and too high load temps.
Had the DMM probe near the ihs for quite a while. The gap between probe and cpu temp readings increased with higher cpu temps from 10° to 20°, never went above ~55° cpu temp.
Hmm.. the temp even from the top of the IHS I measured 11C lower than when embedded within the IHS at just idle. Don't recall load temps.
In the server market they still have an advantage in power consumption due to not using fbdimm's. And scaling is also still better.
I heard that some OEM's received b3 barcelonas.
AMD's 45nm quad was first sampled December and partners already have working samples. For a company rumored so in death trouble, that's pretty good.
No i did not, gonna try it I assume I can not disable core0.
I'll explain more in a bit.