Pretty mega contribution, pioneer! Well done!
To be perfectly honest, pioneer was correct in saying that I've touched and played with quite a large sample size of Socket 939 CPUs. I've got a small collection of these chips, some of which are still sealed in their boxes, in fact; I've also sold many to clients who wanted/needed a S939 platform.
For now, though, I'll tell you about the chip in my gaming rig. It's an Opteron 170 @ 2.8GHz (the CPU referred to in this thread's first post), mounted in an Asus A8N32-SLI Deluxe. The stepping is
LCB9E 0704TPMW; as far as this stepping is concerned, I've overclocked maybe as many as six or seven CPUs from this same batch (the only variation being in the final quartet of letters; the IMC versions and date of manufacture are the same, though), and many of them "read hot." What I mean is that they tend to return higher than expected core temperatures even at low-to-stock VCore.
This particular chip runs naked, i.e., sans its Integrated Heat Spreader. This wasn't the first CPU I had made topless, though, but by this time the operation only took maybe ten minutes at most. It's a safe operation if you're careful and if you know what you're doing.
It's a honey of a chip; the max I took it up to was 3.0GHz (300MHz x 10) @ 1.37ishV VCore; it booted into the OS, but froze almost immediately. Clearly this CPU needs a voltage bump if I'm ever to run it at higher than the 24/7 stable 2.8GHz I have it set to today. The max temperatures I've seen, which were observed in the summer when my bedroom transforms into a sauna, is maybe 57 degrees C on the hot core whilst running S&M v.1.90. My records indicate that the ambient temperature during the test was a toasty 97 degrees Fahrenheit...
I'm perfectly happy with my chip running at 2.8GHz. The RAM actually can run up to 280MHz (DDR560), but its ICs get really funky when running at higher than DDR520 (260MHz); at speeds higher than DDR520, the system becomes 3D unstable, so I had to back the RAM down by using a divider.
It's fully S&M v.1.90-stable. I've run the stability tests multiple times, and it's passed every single time as presently configured.