Just some news about what's going on :
-OCCT PT 1.1.0 was successfully validated on a 8-core system : it proved unstable. It was a bi-quad-core (2x4cores) Xeon Clovertown with 24GB of memory (yes, 24 GB). "RAM" crashed under 15min, "CPU & RAM" between 15 to 1h15min, CPU mode ran 16 hours. Here's a screenshot :
-OCCT is now used by chinese power supply companies to stress-test their products
I was really surprised to learn that it was THAT popular over there...
-On a "what's new" or "what's boiling" in my head i'd rather say, i have planned :
*A programmable test. You tell OCCT to run your own test, composed of "1h CPU&RAM, 30min RAm, 15minCPU mode" and OCCT will run it all alone and graph it.
*"memtest" mode where OCCT will try to test the memory left available by the OS; it won't decently compete with memtest86, but it's still a nice addition. It's roughly done (i.e. running), but i have *NO* idea on how efficient it is right now. It is multithreaded already 
now i have the interface and thread monitoring stuff to implement... well, still some work on that.
*"Oh great OCCT, unleash your power, kill that power supply now !" mode : CPU & GPU stress testing in order to have the maximum load on the computer. I do have a 3d test (DirectX) that is OK-ish (gets my 7900GTX only about 2°C under 3dmark2006), but it loads a CPU to 100%, and is less efficient than the *real* OCCT test. So i'm struggling to make a test that willload the GPu to 100% while leaving the CPU somewhat alone... dunno if i will be able to do it. there probably won't be any error checking on the GPU results, though.
*Better Core2duo support : i ordered a E4300 this morning, along with a Asus P5N-E SLI
So well, there's still alot of work to do, and i'm not quite done yet