Strange problem using MDX - D3D device creation

Hi people, I'm having a strange issue with Managed DX. It appears that calculations involving (double) floating-point numbers are getting messed up after the d3d device is created. I'd be grateful if someone could try running the code found at http://onecrx.orcon.net.nz/TrialCode.zip (40KB) and checking the result stored into the variable 'a' both before and after the Device is created. You'll see the code has been seperated out by a few blank spaces so you can find it easily. (It's in the "Direct3D" region in the "InitialiseDevice" function) I have tried this on two computers now, one with a fresh install of VS2003 and the latest December update of the DirectX SDK. Both give me the same results. I also get the same thing happening on a couple of example D3D programs I have downloaded. My only guess as to what is happening is that the creation of the D3D device is overflowing the Floating-Point stack - I have heard of other cases of FP stack overflow causing similar issues. If someone else could confirm this problem, and perhaps offer a solution, it would be very much appreciated. Many Thanks, George Ricketts