• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

VRM overheating, throttling i7 ivybridge

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Ruiner

Member
Joined
Jan 23, 2001
Setting: asrock z77 pro4m (4+2 phase), 3770k, delidded, scythe kotetsu (120mm crossflow design), 44x at default Vcore. my case doesn't have a 'window' on the backside of the cpu socket....solid metal.

It will throttle after a few minutes of priming. Core temps stabilize in the low 60's, and then the vrm heatsinks and particularly the adjacent chokes get fairly hot and it throttles back to default speed for a few seconds at irregular intervals.

a couple of threads elsewhere mention lowish quality mosfets on this particular board and having similar issues limiting overclocks.

Exploring options before taking things apart:
add small fans directly over vrms (challenging given space taken up by cpu hsf)

remove/replace vrm sinks with better paste (AS5?), unless pads are required because of uneven height

add small copper sinks (surplus in the junk drawer) to the chokes with epoxy. the chokes run much hotter, but IIRC are designed to do so

or

ignore it because it only seems to be an issue under prime or IBT


going to water won't help other than having much more room to aircool the mosfets.


I can pull out the old laser thermometer if anyone wants actual numbers.
 
VRMs likely have thermal pads on them...you will want to use whatever was there originally and not swap to paste or vice versa.

How are you running 44x all cores at default vcore is my question.......

I'd point a fan at it and see if that helps. If not, obviously getting those things cooler needs to happen in some way, be it additional sinks or something. The thermal pads likely won't help as much as active cooling with a fan. Otherwise, get a board more capable for an delidded CPU is my suggestion. While bottom of the barrel boards can handle an overclock, you are getting to the range where these start to show their mettle.....or lack there of.
 
Well there's always a way to attach small (40/60mm) fans to the VRM's area. Gotta be creative ;) Won't hurt to swap out TIM/Pads on those either. Aside of this, if it only happens during stressing and not gaming or other applications, I wouldn't worry about it.
 
Well there's always a way to attach small (40/60mm) fans to the VRM's area. Gotta be creative Won't hurt to swap out TIM/Pads on those either.

Was gonna say, setup looks like it already has a few good years of use so switching the pads might help more then not :)
 
Adding to this cauldron of VRM overheating reasons:

So it looks like you are really pushing that chip hard, but you don't quite have the board to really make the ends meet. For the most part the board behaves but you can't get the heat out, so you need to add fans and possibly replace pads. You should never have to replace pads unless you have taken the heatsinks off several times. Pads do not wear down by heat exchange unless high usage for several years (5+). But I'd say the source to your issue is that you are doing some decent ocing on a 4+2 phase system. That board doesn't look like it is designed to handle a lot of current pull from your CPU. With the de-lidding, you can do higher current pull at lower voltage because temps are nice. But that means the VRMs will be working extra. I'd say try to hard limit your core and mem current (if you have the option) or down clock a smidge to ease the rush current pulls from the VRMs.
 
Yeah it's a budget board, originally from a Microcenter combo.
The board will feed ~1.33v at load when set to 'auto' at that level. Since it's throttling it's not hard to claim I haven't really confirmed stability, but I've gamed on it hours at that speed without problems. 45x or higher will BSOD on boot unless I crank up offset voltage, but then it throttles faster (cpu temps still peak at only 70C, however).

I'll have to remove the mobo from the case to get to the mosfet HS pins in order to redo the TIM or pads, whichever dried up gunk happens to be under there. I'll attempt a fan first.

Thoughts on adding sinks to the chokes as well?
 
Something lower so that it doesn't display the VRM throttling you are seeing.

Consider setting an AVX offset as well. Those tests its throttling in are AVX based which use a lot more power. If you lower just the clocks for that test, perhaps this will curb throttling in normal use. ;)
 
Yeah it's a budget board, originally from a Microcenter combo.
The board will feed ~1.33v at load when set to 'auto' at that level. Since it's throttling it's not hard to claim I haven't really confirmed stability, but I've gamed on it hours at that speed without problems. 45x or higher will BSOD on boot unless I crank up offset voltage, but then it throttles faster (cpu temps still peak at only 70C, however).

I'll have to remove the mobo from the case to get to the mosfet HS pins in order to redo the TIM or pads, whichever dried up gunk happens to be under there. I'll attempt a fan first.

Thoughts on adding sinks to the chokes as well?

Thermal pads are the same all around. You won't find a better material than you have, and if you do it wouldn't even matter. Just make sure a lot of air is moving heat off for now.

Games do not have high demand all the time. If you look at your CPU load, you will see a few cores running between 10-20% at all times. The times when your CPU hurts is when you have spikes within usage. Scene changes or events can cause an inrush of requests to the CPU. This translates to a huge demand of current from your VRMs, which they physically can't provide. So you have to tell your CPU to limit its asking.

CPUs are smart these days in terms of what they can demand from their voltage regulators. There is a communication link on the low level side that monitors and handles requests. BIOS settings can modify the requests so that the CPU is limited in what it can ask.

Others would be best to ask how on what settings to use to create these limits. Like my title says, I shouldn't even be in this thread.
 
Did you use XTU to see if it is current limit throttling, power limit throttling or VRM thermal limit throttling?
 
Reported core temps under XTU stress test are a few degrees C lower than p95 , and it doesn't throttle like it does in prime. Reported package TDP in the mid 70's, lower than the 500 limit I have set in bios. XTU doesn't report thermal throttling either.

With p95 large fft running, peak package tdp climbs to about 89w, and speeds will transiently drop back down to 35x, but thermal throttling is still not reported in XTU.

I'll report back after getting some fans in place.
 
What about the other throttling wingman mentioned? You may need to resize that monitoring area to see it.
 
Consider setting an AVX offset as well. Those tests its throttling in are AVX based which use a lot more power. If you lower just the clocks for that test, perhaps this will curb throttling in normal use. ;)

I don't think AVX offset was added to consumer CPUs until Kaby Lake. Also don't forget that Ivy Bridge only has AVX, not AVX2 (FMA) which is the one that really turns the power up.

To see what is throttling, I'd suggest hwinfo64. It shows various limiters, some or more of which may be activating.
 
hwinfo-throttle.png

Maybe throttle isn't a correct term for all of these, but I don't recall XTU having so many options even if you can be bothered to dig through the settings to show more than the default set. The two showing on mine for example, is because I'm running the CPU stock so it wont go to higher clocks.
 
More detailed indeed, most not throttling though as you noted. :)

Xtu shows, thermal, power limit, current, and vrm limits by default.
 
I'm currently using hwinfo64 as well (non-pro, if relevant). It doesn't show a tag for throttling.
XTU shows thermal throttling (yes/no) and TDP readings (readings in watts for either package or core) but no mention of vrm limit throttling, as far as I can see, either maximized or in the wrench icon config settings.

edit: I'm using hwmonitor64
 
Back