I can give you some notes I've made regarding the dumps. I don't see anything definitive, honestly. It just looks like general memory corruption which can be caused by buggy drivers or hardware issues. I don't see many 3rd party drivers loading nor do I recognize them as drivers I've seen cause problems on other systems.
** Notes **
[FONT=Verdana,Arial,Tahoma,Calibri,Geneva,sans-serif]The 0x3b bugcheck dump shows a thread owned by procexp64.exe (Process Explorer) threw an exception when it tried to read an invalid memory location. The function being called which threw the exception was dxgmms2!VidSchQueryProcessNodeStatistics. In fact, if I use .trap to set the context to the time when the exception actually happened, all of the cores look like they are executing the same function. I suspect Process Explorer was sampling all of the cores for information but something went wrong on core 1.
[/FONT]
[FONT=Verdana,Arial,Tahoma,Calibri,Geneva,sans-serif]The 0x18 bugcheck dump shows a thread owned by destiny2.exe tried to dereference an object while in the kernel which caused the system to detect an invalid value for the object being referenced in the context of what the function was trying to do; dereference the object. In this case it was an Event object and it looks like the reference count was already 0 prior to the attempted dereference.
[/FONT]
[FONT=Verdana,Arial,Tahoma,Calibri,Geneva,sans-serif]It's not obvious to me how they would be related. Kernel memory corruption is the only thing that comes to mind. Either a driver bug or hardware problem. I'd want to eliminate the possibility of a hardare problem by making sure the hardware isn't being pushed beyond its officially supported settings.
[/FONT]
[FONT=Verdana,Arial,Tahoma,Calibri,Geneva,sans-serif]The 2 process dumps for destiny2.exe suggest an object reference is invalid but the developer's code didn't check for that possibility. So, either lazy programming or whatever went wrong didn't seem like it could possibly go wrong to the developer.
*********
If you have any other dumps we can look at it might help. It often takes several to spot a pattern. I'd very much recommend removing the XMP as I've seen the tests you've used miss problems. In my mind it would be best to be sure the system is stable with officially supported settings for all parts.[/FONT]