Now I noticed my VRM was warm to the touch upon shutting down after a prime95 test, of course (meant that they were working!), so i figured to add a small fan to the set-up. Here's the story in pictures:

[image loading]

40 MM fan (when hooked up, to my surprise, lights up blue/green/red like a christmas tree).

[image loading]

Ripped from this old, 2003 computer found in my dead uncle's garage.

[image loading]

Heatsink, fan on top, this on top of fan, screwed in.

This is how I got the idea to use the fan - there are no special brackets that held this 40mm fan onto the motherboard northbridge (***, why didnt my 2010 biostar a770e3 didnt have a fan on the mobo-northibrdge? I had to jerry rig one on...), it was straight screwed into the heatsink. I, too, would screw my fan directly into the heatsink then, having the heatsink fins hold it in place.

[image loading]

Theres the heatsink, but the rear 120mm hyper 212+ fan would be issues. If I put the fan on top of the mosfeat heatsink, i would have to move up the pull fan a cm or so. Maybe that'd raise cpu temps. maybe that'd lower it, who knows. And, maybe it would stick out too far that I woudnt be able to put my panel back on?

And then there were questions, shoudl I put this fan in a way where it's 'pulling' air through the heatsink, as in, instead of blowing down on it radially like a stock CPU heatsink, maybe I should set it up like a tower heatsink, and have the fan 'pull' air through the mosfet heatsink. Or, maybe with pull? Maybe I could sort of put the fan at an angle, on top of the heatsink, but not quite on top, sort of at an angle, to allow my pull fan to fit in with enough room (didnt give it enough room, so didnt do that).

I decided with blowing down, radially, onto the mosfet heatsink, in order to cool the chokes and capacitors as well as the mosfets and mosfet heatsink.
[image loading]

Notice how it glows! Note the best pic, it does look cool though:
[image loading]

Now maybe I screwed up my temp diodes when putting the fan in, i mean they are tight in there, and I had to remove the motherboard from it's standoffs to get the fan wires behind the motherboard, but I made sure to keep them in place (spent like an hour seating my 2 cpu diodes, god damn they still won't read the cpu temp no matter where i stick them, damn hyper 212+ being exactly the same size as an am3 cpu and direct contact copper pipes taking entire temp load so the top of the base isnt warm at all)...

but I think this fan has resulted in at least 15-20*C of cooling. My vrm temps aren't going above 56*C when the lowest they've ever been under prime95 is 65*C, and more than likely they should read 75*C during prime95 under the voltage I have right now.

Why buy a high end motherboard? Just buy the cheapest piece of **** you can find and slap mosfet heatsinks and a fan on it!

Belial88

United States5217 Posts

November 17 2012 01:52 GMT

#22

Okay so I literally tested everything, I was pulling my hair out, I couldnt even get 3.4 ghz right. I kept failing 6 hours in at 288fft length, and looking back I think when i tested my cpu at 'stock', i only ran it for 6 hours, thinking if it could do that itd be fine for 24 hours. stupid me, lesson learned, always prime95 test at stock just to make sure everything is okay -_-.

So I believe my IMC has some issues. I finally was able to get a stable overclock, on the first attempt, when I downclocked my RAM from 1333mhz to 1066mhz. It was an out of the blue attempt, because I know my ram was stable 2 years ago at 1348mhz at CL7, and I was running stock 1333mhz, CL9, with extra voltage (the same voltage i used for overclock).

I'm going to test the overclock again, 3.6ghz@1.472vcore, but with 1333mhz on single ram instead of 1066 on 2 ram sticks. If it passes the 6 hour mark that i always was failing at, then it means having 2 sticks of RAM strains the IMC, and if it fails, that might indicate the RAM. I've heard that running 2 sticks of RAM is more stressful than 1 stick of RAM.

Belial88

United States5217 Posts

December 05 2012 00:59 GMT

#23

Okay so I sort of stopped posting because I was having no luck with anything and it was kind of frustrating.

Every overclock setting I tried fail within 6-8 hours, with no correlation between voltage and frequency and the million irrelevant settings like boot priority, GPU controllers, et cetera. Just no matter what my voltage was or frequency, I mean 3.8 ghz at 1.4vcore lasted longer than my test at 3.4ghz at 1.5vcore and 1.3vcore, just nonsensical results. I kept failing at 6 hours at the same FFT length.

I finally found out, that downclocking my ram from 1333mhz to 1066mhz, fixed the issue. So, faulty IMC right? Well, I moved my 2 RAM sticks to my other 2 RAM slots, did some memtest86+ on both sticks on every motherboard slots, both paired and by themselves, and they never had any error. So a rather frustrating test. Anyways, fuck it I felt, I decided to figure out my max overclock on 1066mhz before I go back to figuring this out.

Well, I found I was 30+ hour prime95 stable on custom blend (set memory to 2500k memory so all availble ram when i have chrome up, i know, shouldnt run chrome when doing prime95, but i sometimes need to use pc) at 3.7ghz@1.504vcore (1.3375VID + .2v over voltage). I also found I was just as stable if I was at stock memory speeds using the other ram slots.

I decided to run the same clock/volts, but with stock ram settings, on the original, closest 2 RAM slots. And it fails within 6 hours!

So apparently the problem was never my chip, it was my RAM slots....

note that these same ram slots ran fine on my athlon ii system 2 years ago when I set up the overclock (or at least I think it did, I got from 3.2ghz stock to 3.41ghz/2.6ghz NB 1.472 vcore 1348mhzCL7 RAM, which is in line with what most people on got on an unlocked x3 rana 450).

So... did my ram slots go bad in the last 2 years? Was it always bad and maybe I could've overclocked my Athlon even further had I used different RAM slots? Or is there some sort of weird interaction between my Biostar motherboard and the Phenom ii i have that makes the first 2 dimm slots bad? Maybe there is some weird IMC flaw that causes this or something?

Most things I've read say the first 2 ram slots are better for stability in most people's cases.... but I guess in mine, it's the opposite that's true.

Anyone have any idea why that on the original 2 dimm slots, I can't overclock more than 3.4ghz without lowering RAM below stock, but on the further 2 dimm slots I can overclock past 3.7ghz on stock ram speed? Is there any way I can test or look for damage?

I have not tried above 3.7ghz yet because 3.8ghz requires at least a little more voltage, around 1.536vcore at least, and runs too hot at 60*C average (not 60*C for a second in 24 hours), so I stopped testing it until I have better cooling.I have an h50 and nh-d14 coming in tommorow.

Also, I will be putting up a benchmark test here and at OCN on H50 vs NH-d14 vs Hyper 212, as well as comparing Stock Paste vs PK-1 vs PK-2 vs PK-3 thermal pastes. I will be doing as close as possible to apples to apples comparison:

- For thermal tets, I will spread on both cpu and hsf (half rice grain on each), on an h50 with 2 x yate loons push/pull.
- For heatsink tests, I will test 2 x yate loons on push pull, h50 mounted outside the case to avoid any intake vs outtake issues, and NH-d14 with 2 x yate loons and the stock 140 mm fan in the middle (although 2 x yate loons 120mm only woudl be more fair, I think the 140mm fan in the middle is really a defining characteristic of dual tower heatsinks).