Jack: Await the results.

Informal: There was definitely a problem somewhere and narrowing it down I found it was AOD Assistant at bootup causing lockups. Since then I've not seen it.
But even still, if you choose >200x13 (2600MHz) with perfectly stable and stock RAM/NB/HT the system locks up there are then with 1.472V and 1.4VID. That's already topping 350W load. Basically while manipulating the PLL, it locks up. I recall Xbit labs saying 2800MHz was a breeze perfectly stable and 3000MHz needed just a few more volts for 100% stability by simply using multipliers. Well, there's the difference with different items. I've maxed out the VID/Voltage/HT/Multi, the system won't boot past 2550MHz and 2600MHz is screenable but will freeze very quickly. 2500MHz didn't freeze or anything, is quick short benchable but failed P95 after 4 minutes or so and 2500MHz was not even 3D benchable (Cinebench - I'm only intrested in all 4 cores yet).

Max on the same board, OS, AOD, BIOS is around 2914MHz with 9500. Even my last quick attempt just 2 minutes before packing it up reached this: http://valid.x86-secret.com/show_oc.php?id=288977

I've played with Phenom long enough to know the obvious trend in how to spot a bad oc'er, and this is one of them. The worst one so far. Max limit is upper 2.4 for stable operation. Voltage/Temps don't effect it really, stock gets you as far as 1.55V/1.5VID gets you and ambient 26C gets you as far as ambient 2C gets you.

Also AOD reads wrong values for temperatures, EVEREST is closest to correct I feel. AOD is basically showing me the ambient temps which are incorrect. EVEREST reads 10C higher idle at 1.18V/1.2VID which looks to be correct based on my previous experience.

I ran some Cinebench10, quickly since it's real world rendering and we've spotted some weird happenings around in the mutli-CPU rendering. Here are quick results and this time they have methodological accuracy as everything is kept constant but CPU speed/multi:

Setup
9600 BE @ 2000 (10x), 2100 (10.5x), 2200 (11x), 2300 (11.5x), 2400 (12x)
RAM @ 800 4-4-4-4 tRC 24 tRFC 75ns 1T
NB/HT @ 1800MHz
HT reference @ 200MHz
MSI K9A2 Platinum 1.13 BETA BIOS (no errata patch)

Format:
CPU Speed (%)
Single CPU Score (%)
Multiple CPU Score (%)
Multi-CPU Speedup

2000MHz (100%)
1698 CB-CPU (100%)
6404 CB-CPU (100%)
3.77x

2100MHz (105%)
1786 CB-CPU (105.2%)
6750 CB-CPU (105.4%)
3.78x

2200MHz (110%)
1868 CB-CPU (110%)
7060 CB-CPU (110.2%)
3.78x

2300MHz (115%)
1943 CB-CPU (114.4%)
7374 CB-CPU (115.1%)
3.80x

2400MHz (120%)
2015 CB-CPU (118.7%)
7765 CB-CPU (121.3%)
3.93x

That last one was odd. Usually I get 3.82x speedup but that was high. The trend I was looking for is there but I needed 2.5-2.8GHz benchable to find out more and that's not possible here.