Originally Posted by gr8sho
Nothing like empirical data.
So, we a sample size of one. I have contemplated running this way myself. Assuming thermal limits are contained, is there any concern for the board itself? VRMs and so on.
The VRMs will get VERY hot... just FYI Same with the backside of the socket. Universal VRM blocks are rather "cheap" though for the good they do @ ~$30USD opposed to a "complete" board solution (VRM+chipset,ect) with can run up to $140 and may cause VGA interference (chipset block) with larger GPUs. As stated earlier in this thread, simply swapping out the thermal pads on the OEM heatsink can make quite a bit of difference. Some have reported as much as a 12*C drop in VRM temps. However to measure these temps correctly you will have to physically probe 1-2 of the VRMs. I had some of the old probes from my "old" Crosshair 1 board (back then they were included) and installed them into the "opt temp 3" and "opt temp 4" pins on the board (opposed to a bay device) so that I could view/monitor them within the OS. I decided not to install a liquid block on my board since I'll be jumping the AMD ship until the new socket is released and why waste the $130 for 2-3 weeks of usage. (besides for the fun of it lol ) So luckily these boards still have some great resale value.
What I ended up doing for the time being was install a (though ugly) Antec Spot Cool unit, spaced approx. 3" from the VRM heat-sink. Simply by moving the fan's bendable arm away from the VRMs gained me a 14*C spike in temps from those 2 probes. When running wPrime 1024M benchmark I saw it increase to a total of 62*C. I replaced the fan back to it's original location and noticed a 14*C reduction in temps @ idle and a MASSIVE 28*C reduction while running wPrime (full LLC).
This was at a "mild" OC of 4.75Ghz @ 1.43vcore / 1.4NB. With that being said; I could only imagine it's impact @ a 1.6vcore.
Just something to think about
Forgot to mention about the backside of the CPU socket.... I had this rig up on a test bench unit (with an 8120 at the time) which had a "bamboo" mobo tray as a spacer between the steel tray and the board to help insulate when I do my DICE runs. Had the system running on that while I leak tested the cooling loops (5x rads + 3x pumps = a lot of possible leak points). Being that it was "open air" I had used a very large Thermalrite True HSF and decided to make the jump into the 5Ghz range (was winter) which was successfully benchmark stable.
However; when I went to remove the system from the test bench, I noticed a perfectly square burn mark left on the bamboo tray. Not just discoloration mind you, but actual "char" marks. Now with the 8320, I was playing around behind my mobo tray tieing down a few loose wires while the system was on @ idle and my shoulder brushed the backside of the cpu socket (1.43v mind you) for maybe a second and I swear it "nearly" burned a hole in my t-shirt. Definitely made me jump. lol
I remember back in the AM2 days we would cut out an opening on the backside of the case for 80/90mm fans but back then you were lucky to have maybe 1/4" of space between the back of the mobo and the panel. I have 5/8" worth of "clear" space back there and am still outputting that much heat. Not to mention the 5X rads, 16 fan, and 3 pump config which cool this system. So something to consider as well.
Anyway; I'd say "Go for it" , but expect to do some extra work to cool these areas.Edited by mrinfinit3 - 3/17/14 at 1:01am