First off, here are my current config. specs.
AMD 9370 (4.7 GHz at 1.51 volts),
AMD Reference R9 290 ( 0 modifications),
ASUS Crosshair Formula V-Z Motherboard with both eight-pin and 4-pin connected,
ToPower 1000W non modular power supply 80 plus gold cert,
Thermaltake Water 2.0 Extreme with 2 PWM Corsair SP 120 performance fans (Using software to control inside the OS),
Creative THX certified soundcard X-Fi HD Titanium,
Corsair 750D with all stock fans in place (Slightly negative air pressure),
RAID 1 WD 300GB 10000RPM Hard disk (MAIN),
Seagate 7200RPM 1TB Hard disk (SECONDARY),
Seagate Backup 2TB,
16GB RAM G. Skill 2133MHz 11-11-11-30 (Running at 1600 9-9-9-28),
Second latest BIOS,
All BIOS options unchanged with the exception of power saving features,
Now that the list is done (Tell me if I've missed anything important), my problem was... really weird. I had to by a new motherboard when my Gigabyte UD3 died and I bought the new AMD 9370 with it (it was the same price as the 8350 and I liked the guaranteed 4.4GHz clock). So, I build my machine all over again, get into setup, make sure my boot drives configured, yada yada yada. I eventually ran prime95 on stock CPU, RAM, GPU settings for stability purposes (I'm a heavy gamer and music editor). After twenty minutes i noticed some CPU throttling. We're talking about severe throttling, down in the 1400MHz range. It then goes back up and fluctuates frequently between the two numbers. I install core temp and HWmonitor, and I start looking for trends in the temperature of the CPU.
HWmonitor provides 2 temps. One I believe is the socket and I know the other is the core temp due to... well, core temp! The cores stay well below what I am aiming for (50 degrees centigrade. My absolute limit during stresses is 72 degrees centigrade, because in reality it won't ever get that hot). So there shouldn't be throttling correct? I'm pretty sure. Core temp says the chip's tjmax is 90 degrees centigrade. But it still throttles after 20 minutes of prime95. What the hell is going on? It has to be the other temp pin on the board. But how do I get to it?
I took off the windowed (motherboard front access) panel of my case and took a house fan too it. I thought the additional cooling would help cool off the VRMS and MOSFETS in the system where only air is being pulled indirectly over them. There wasn't a big difference. So I took the other panel off and aimed the fan at the back of the motherboard's CPU socket and ran the tests again.
There was a dramatic effect. First off the CPU socket temps leveled out at 70 degrees centigrade (this is not the core temp). Secondly, and most importantly, the CPU throttling disappeared and I was able to do my 24 hour stability tests without worry.
HOWEVER...
I wanted to overclock the CPU, and I'd rather not have that noisy fan in the back of my open, supposedly silent, dust free case. I could go custom loop but then I'd cool the hell out of the setup, and I'd rather not spend north of $700 to fix what might be a $25 problem. I didn't have this issue with my 8320 (With broken pins, that's why I replaced and upgraded) overclocked to 4.4 at 1.53 volts on my UD3 motherboard. Had I really been planning I would have used an Intel solution, but hindsight is 20/20 right?
Any questions, comments, thoughts, ideas, etc. on the subject? Honestly, the computer works, I just thought it was an intriguing fix to the problem and I can't find it anywhere else on the internet.