I will confirm: This is a show stopper BUG for me too:
SuperMicro SS6016TT-TF. 2 x 4 Xeon E5520 processors.
Hyper-V is not keeping time in Windows guest OSs (did not trn non-Windows guests).
Typical: Windows Clock stops for 10 seconds every other "second" count.
At this time, my project to put the Xeon E5520 into production is on hold as this timing problem prevents it from being used for anything that requires a time stamp of +/- 10 seconds accuracy. E.g. the only way to have unique time stamps is if you are only doing no more than 1 transaction per 10 seconds.
I hope Microsoft and INTEL expedite fixing this show stopper bug. Until then, deploying the Xeon E5520 to production would be a disaster for us.
Let me know if you find a resolution please.
Thank you for your observation of the issue on another platform.
I am not an authority on low level hardware myself. Although I have received feedback from Microsoft regarding this issue being a BIOS programming issue, I encourage you to seek support through your methods. So far, Hyper-V RC2 Cor Server of Windows 2k8 seems to be working to +/- 1 second. I hope they don't break it in the production release of R2. The following is the feedback I received:
"... this could have happened on any machine with NH CPUs. The problem is that the BIOS writers guide is not clear on how to handle this. During Hyper-V v1 testing we didn’t have NH machines available (not even pre-production at that time) so we couldn’t have found this issue. In R2 we added C3 support where we expect these timers to stop in that state so we don’t see the issue as readily there. Combine that with the chance that a BIOS writer interpreted the spec one way or another and the odds are pretty small that we see this. Unfortunately you just got unlucky and hit the “jackpot” – Win2k8 with NH CPUs and a (potentially) “broken“ BIOS.
Note that we are [Microsoft] still assessing this so don’t assume that what our hypothesis says today is the right answer – just that that is the path we are investigating currently.
Keep me posted please if you learn anything new.
Well! We found a fix to this problem!
You have to disable Intel C-State instructions into BIOS configuration! More about C-State: http://software.intel.com/en-us/blogs/2008/03/27/update-c-states-c-states-and-even-more-c-states/
Now all work fine!
Hope it helps!
I finally was able to make up a USB boot drive, install the latest SuperServer BIOS and then regression test Windows 2008 SP2 and Hyper-V in the new BIOS. The problem has gone away with the new BIOS. I did not disable the C-States in the new BIOS - they are enabled. Disabling the C-States in the old BIOS did not correct the problem. My conclusion is that the original (old) BIOS was the problem.
Thank you everyone for your attention to the issue and feedback.