Mosfet temperatures, cpu throttling, homework I wish I woulda done before buying...

A

Aaron_ATX

[H]ard|Gawd
Joined
Oct 15, 2003
Messages
1,804
After 6 weeks of owning my 2600k and gigabyte z68a-d3h-b3 combo, ill just man up and say it... I didn't spend enough time researching before I purchased. :eek:

I've been overclocking for years and years now and I guess I took my leet skills for granted.

Since I assembled this system i've been pushing the overclock, tweaking voltages, benching, priming, burning, gaming, and logging. Settled on 46x10 with around 1.38vcore at full load, with HT on.

Overall I was disappointed with the overclock, as the chip seems willing to boot into windows @ 50x100, but I could never get it stable under load...in the end I just chalked it up to a crappy motherboard, as far as oc'ing is concerned at least.

Well, believe I was right, but as of this afternoon I realize it was because of a potentially dangerous combination of ignorance on my part.

After noticing some bizarre behavior with programs like intel burntest and prime95 ~ cpu would ramp up to 46x100 and temps would shoot up to around 70-75c.....and then about every 15 to 20 seconds, the temps would all drop about 10-15c suddenly, and then a few seconds later be back around 70-75c. Seemed very odd to me.

This afternoon while benching I noticed that the temperature drop always occurred everytime my TDC got into the 125w range....it would stay there for a second or two, then drop to 75w or so, and then start climbing back up. Oddly, none of my monitoring programs showed my multiplier dropping during this...but it certainly could be noticed in Intel burn test, as the gflops and time to complete each cycle could vary considerably when this would happen. So I got to poking around on the net a bit more, and came across
this thread.

First of all, my motherboard is rated @ 95w max TDC. A stock clocked i7 2600k is rated at 95w. I had NO intention of maintaining stock clocks. I completely overlooked those "little" facts. All I saw was that I could change those limits in the bios and figured YEEHAW, put the voltage to it and see how fast it goes. In its current configuration I have seen my 2600k hit 130w before being throttled back. Now the throttling is where I became a little concerned when I realized what was happening. At those power consumption levels my mosfets are pegging out at what is apparently their maximum temperature ~ 66c (150f) , and apparently when they get to that temperature they geek out..had I continued to abuse them I would imagine it was only a matter of time before they melted right off the board.

As I am beginning to understand, my VRM's are on a 4+1+1 phase ~ and not only does that cause the horrible vdroop this board exibits, that style phase will heat up tremendously faster ~ and not to mention the factory "heatsink" draped across them looks like something from a lego kit, and feels like it is attached with unicorn turds.

So now I am not exactly sure where to go from here. First order of business is to try and get my mosfet temperatures under control, assuming that is even possible with the amount of wattage I am drawing with my cpu overclock. Other than that I guess I need to lower my oc. I suppose really my only legitimate option is to buy a motherboard that is actually designed for what I am trying to do, overclock the piss out of it.

If any of my information is incorrect, please help correct me. I am just coming to start to realize the pieces of the puzzle I was missing when it comes to overclocking a sandy bridge.

Hopefully this information catches someone before they nuke a motherboard doing dumb shit like I was.
 
If you are not going to get a new MoBo go on eBay and buy some Sekisui 5760 thermal tape and some small RAM or MOSFET heatsinks. It doesn't have to be particularly extravagant to keep the VRM section cool enough.

Get some of this, it's not a bad idea to get a little more than you need for the future: http://www.ebay.com/sch/i.html?_nkw...al+tape&_osacat=0&_trksid=p3286.c0.m270.l1313

These heatsinks are 6.5 mm wide: http://www.petrastechshop.com/enmomohekit.html

While these are 10mm wide: http://www.petrastechshop.com/enmomohekitp.html

Your motherboard only has one IC connected to each choke, so it is DrMOS. They are relatively large so you will probably need the 10mm one. If you get either of the above heatsinks you do not need thermal tape, it already comes pre-applied.
 
Last edited:
As an eBay Associate, HardForum may earn from qualifying purchases.
thanks for the links, I have boxes and boxes of old parts/heatsinks to look through first to see if I have something that will work.

I have yet to pull the factory sink off, I have a feeling it isn't secured correctly....sure feels loose.
 
I read your post and I saw your comment about 4+1+1 VRM but I still thought it is 6 phases. Since the other two phases go to the IMC and something else I'm not so sure that adding heatsinks is going to help. It's certainly worth a shot, you have nothing to lose, but sorry I got your hopes up, I don't want you to be disappointed if it doesn't help.

Here's an article I found and read maybe a month ago, about the physical components on a computer's voltage regulator section. The thread you linked in the first post had lots of good info, so I'm not sure how much you will learn, but hopefully you learn something from it: http://www.hardwaresecrets.com/article/616 I myself have forgotten most of what I read, I'll have to go back to it.
 
Good info there also, thanks. I'm noticing that only intel burn test will actually get my wattage draw into the 120w+ where the "throttling" if thats what it can be called is happening. Prime95 sits pretty at 103-105w.

Gaming with my 4.6ghz oc, and my overclocked 560ti (950/1900/2000) doesnt seem to get much higher than 77-80w.

I suppose if I just dont run a cpu burn style test I'll stay within recommended draw for this board.....but I really don't like that. Seems cheap. Looks like the only real alternative is to buy a solid overclocking board.

Damn, embarrassing :eek:
I can't ever recall buying a motherboard for myself where I missed the mark as bad as I did. Lesson learned.
 
How often are you planning to run IBT? It's just like the current limiter they put on GPUs to keep people from killing the card running Furmark - it's never going to draw that kind of power in real-world use, so why even worry about it?

What are you using to monitor the current draw? According to CoreTemp, mine only gets to about 108W under IBT loads.
 
You'll never see those kind of temps/watts with normal everyday use, only in IBT. Take the Moffset heatsinks off and put some good paste on them, plus make sure they are getting good air movement across them. If possible add more tension on the Moffsett heatsinks or replace with screws (unless glued or taped on). It should last you a long time after that.
 
I don't plan on running IBT at all, I was basically using it as a stability reference.

RealTemp and CoreTemp, and Hardware monitor show my wattage as exceeding the 125w range at full load @ 46x100 in IBT (hyperthreading on)

Does 130ish watts sound like more than I should be pulling at full load? Forceman, do you have your TDP set in the bios above the stock settings? I have mine at 125, (had it at 300 previously to prevent it from getting in the way of my overclock).

Spine ~ it isnt the brand name that is screwing me here, it was my lack of fully researching what my motherboard was capable of and how an overclocked i7 2600k fit into that picture. I've run plenty of gigabyte boards in the past, and asus, and abit, and epox, and intel, and dfi, and on and on and on.... the 1155 socket was brand new to me as far as overclocking goes and I just kinda half-assed my research.

fps4ever ~ yeah I am going to pull it tonight, it seemed like it was attached rather shitty....gotta love plastic clips and cheezy springs :rolleyes:
 
I can't recall off-hand what it is set at - I think the settings on my board were percentage based, and I set it to 150%, so whatever that works out to (145W?). I used IBT last night with Very High (4GB of RAM used) and manually forced 8 threads, and it topped out at 110W (Vcore is 1.38V). Maybe yours draws more due to the Z68 chipset (thinking integrated GPU or something).
 
Hmmm, odd.The onboard video isnt connected and it is as disabled as it will let me make it in my bios.

What would cause me to pull 20 more watts than your setup? I have basically the same vcore, but 200mhz less....ram voltage is 1.65 for my ddr3 1600...all my other settable voltages are stock or very close.... overheating mosfets = ohms greater = requiring more wattage to get the required juice? Not sure it works quite like that....
 
You must log in or register to reply here.
Back
Top