The "two pairs" type architecture that the picture shows and you describe does explain what I'm seeing. I presume 0 & 1 are on one side and 2 & 3 are on the other.
Is there no way of just switching out or turning off CPU0 and CPU1 for ALL processes - I already do it for individual ones. (I've asked before here I think). Actually I know of at least one process than will not take anything but CPU0. If I change the affinity for GUI.EXE (presumably a GUI for Gigabyte's GSvr.exe) it just quietly slips back to CPU0, Situations like this, as well as no doubt parts of the XP Pro system software, would mean that anything interfering in this way (if it is even possible) would have to effectively map or renumber the cores and the maximum number of cores e.g. 2 & 3 of four cores would become 0 & 1 of two cores.
Actually, I just ran the sim for several hours with all my own processes ignoring CPU1 (only) and all worked OK. CPUs 2 & 3 warmed up only about 4 degrees C (CPU0 pretended it was cool with the whole affair) and, for the first time I noticed that one CPU wasn't taking a full load almost alone - the load was more or less evenly dispersed among the three. Is CPU0 still problematic you think? It seems like my main problem is CPU1.
Why did you say that 0 and 1 are getting stuck at 57C? Where did that number come from? Is that (TJmax0 - Dist0) or (100 - 43.0) towards the end of my spreadsheet, but then what about the Idle0 of 4.0?
Some time ago now, I did change carefully clean off and replace the gel with fresh gel after I read an excellent review/tutorial on the subject by Olin Coles at
http://benchmarkreviews.com/index.php?option=com_content&task=view&id=170&Itemid=38
I really liked his experimental and investigative approach and it gave me confidence to give it a try and err rather on the side of too little than too much gel. In the case of my processor and stock heat sink, one small drop dead centre seemed appropriate.
You said that "all of these sensors" stick at low temperatures. Do you mean that all the Intel Core2 processors sensors tend to (presumably) stick. Which affordable processors do not and take a balanced approach in their architecture. My grandson might be buying me a motherboard for my birthday, because the one I'm using is his hand-me-down and he felt it might have been damaged (anyway it's v1.0 of a Gigabyte board that they never made a later version of)! Also it has just 2GB of 2x333 (667) Mhz memory which may be a bottleneck for a 2.33GHz Quad processor. However, you seem to be suggesting that I should throw away my processor and get another anyway. It all sounds expensive!
I have tested the memory with free DOS type utilities, but would welcome advise of what type of (ideally free and Windows) utilities I could use. Wouldn't Prime95 or LinX just test the processor's stability? I think I know about the processor now - it passes Intel's "Processor Diagnostic Tool" tests, but they do not seem that stringent anyway.
Actually, the reason I am now BACK on Trainz 2004 is that it DOES work (most of the time) and, no doubt, because of my PC's problems, my Trainz 2010 has stopped working and would not reinstall (I tried five times). It is very database intensive and my PC keeps messing up the installation and patching procedure (which runs overnight!).
I've thought about getting Trainz 12, just coming out now, but want to ensure my PC is working right first.
Thanks anyway.
Peter.