I tried to migrate an NT4 server using VMWare converter but it kept failing at 99% when finalising the clone. Due to time restrictions I didn't have time to investigate further so I had to resort to using Acronis True Image to create a .TIB file. Once I created the .TIB file I then used VMWare converter 4 to import this image file into vCentre.
After a few hours I noticed that the CPU was still alerting 100% CPU usage, I logged on the VM but it was using less then 5% CPU resources. I migrated the VM to another ESX host but I still encountered the same problem. I logged a call with VMWare who sent me the below email:
"We would like to suggest you to try changing monitor.idleLoopSpinUS value to 100 and if that does not make any difference, Increase to 2000.
Please make sure you unregister/register VM after modifying vmx file.
If issue would still occur, we suggest to create a new Windows NT VM from scratch and monitor."
I changed the VMX file but this had no affect, the CPU was still reporting 100%, rebuilding the VM wasn't an option as that would take far too long. I logged onto the phyiscal server where I took the image from and noticed it had 4 CPU's, this reminded me that Converter amends the HAL from a multi processor HAL to a uni processor HAL. As I used Acronis True Image to create the VM this had not been done. I logged onto the VM and noticed this was still using the multi processor HAL. I took a snapshot of the VM, copied HAL.dll and NTOSKRNL.exe from SP6a into SYSTEM32 and replaced the existing files. I then rebooted the VM, it booted up with the uni processor HAL and the performance graphs in vCentre started displaying the correct counters. All that was left to do was to remove the snapshot 
