Heatsink not working right?

[RickyJ]

So I've been running this i3-540 at 200x19 for the last 5 months at an arbitrary 1.30Vcore, never paid much attention to the load temperatures. Decided to crank it up a bit today to 4.2GHz and was hitting 75°C load temperatures, but when I touched the heatsink it didn't feel any warmer than ambient (19-20°C). Using a Gelid Tranquillo heatsink and the fins, heatpipes, and base all felt close to ambient. Feeling behind the CPU socket through the access panel in my case, it's very warm to the touch but not burning, I'd put it at 40°C while RealTemp and CPUID Hardware Monitor are both reporting 75°C CPU temps.

Pulling the heatsink off to check contact, I'm not too impressed with what I find. I used the pea method way back when.

I cleaned everything up, put a pea sized amount of MX-3 in the center of the processor, and reinstalled the heatsink so the fan points up towards. Temperatures didn't change and the heatsink still doesn't get warm. Dropping back down to 3.8GHz didn't drop the temps at all.

Dropping to bone stock speeds and voltages, temperatures are still too high.

Looking again at reviews of this heatsink, this shows that it can handle an i5-750 at 3.8GHz and 1.3V at average 53°C. Other reviews pitting it against an i7-920 also show it doing quite well as well. Definitely should be running cooler than a quad core, even if their testing conditions are different.

TechPowerUp got to 4.57GHz at 1.344V with 65°C load temps using a Scythe Katana 3.

I'll check heatsink contact again in the morning to see if I missed something. I was going to install the stock Intel heatsink to see if it works any better too.

Any other ideas?

 

[Spare-Flair]

Your CPU has four cores and the method of application for 4 cores is supposed to be a thicker line down the middle, not a pea drop (which is wrong as well, it should be more like a rice grain). I also see a lot of excess TIM at the edges of the heatsink which was squeezed out which means you possible applied too much in the first place.

What is the orientation of your heatsink? Blowing up or blowing back?

 

[RickyJ]

It's a dual core, though it does have two physical dies under the head spreader.

The heatsink was blowing back since I first built the machine, now blowing up into the 140mm blowhole in my case.

Working from home today, I'll redo my paste. I remember using rice grain sized drops on the i7's and such I've built for work, and had great contact pattern. Been so long since I've done it, getting rusty haha.

 

[RickyJ]

Using about this much paste:

Using stock Intel cooler at stock speeds/voltages:

Still using stock Intel cooler, crank it up to 1.25V and 3.8GHz for shits and giggles:

Not enough contact force? (pic after ~15 mins of LinX at stock, and a few mins at 3.8GHz)

Put the Gelid backplate back in, same amount of paste, tighten down and remove to check contact:

Close-up of non-squishness:

Clean and re-apply a tiny tiny bit less paste, temps still poo:

In the base of the Gelid cooler I can see marks from the corners of the CPU, is this one maybe more concave than others? I'll pop it out later and check if with a metal ruler, I don't really have anything super flat to lap the CPU with.

 

[RickyJ]

Watching RealTemp sure is interesting. Idle with everything else closed, 38/43°C on the cores. Click on Firefox to open it, temps jump up instantly to 50°C while loading, and back down to idle temps 1 second after Firefox finishes loading. Same story with IE. That seems to point towards poor thermal transfer if the temperatures can vary that quickly.

edit: Just clicking on Post Reply for this post caused the temps to jump up 3°C, then right back down.

 

[Skull_Angel]

odd problem you have going on here, I've seen it happen with some i7s that have bad ihs contact with the chip, but seeing as these i3s run really cool in comparison I wouldn't know if this is the case. it could be a bad temperature sensor on the cpu since it should be reading much lower than the core temps, but I wouldn't rule out bad ihs contact either.

 

[radaja]

are you just running one fan on the heatsink?or push/pull?

 

[RickyJ]

From everything I've seen, this chip should be running damn cool. i3-530's running 4GHz on stock Intel heatsink is fairly common, at least among the few that try it.

Pulled heatsink off again, same goopy story as before:

Using the best straight-edge I've got (compared to an x-acto knife blade to make sure), here's what it looks like with the ruler perpendicular to the writing on the chip:

Not bad, right? Check this ish out, single point of contact all along the chip with the ruler parallel to the writing:

WTF mate, right? This chip is bowed outwards somehow, I wouldn't call it convex since it's only along one straight axis. Now here's a grain of rice worth of MX-3 in the middle of the IHS, CPU placed upside down in the middle of the heatsink base, and me pressing as hard as I can until it won't spin easily (bottoming out basically):

Still poop. Now with the Gelid GC-2 that came with my heatsink (that I immediately tossed in my desk):

Now we're getting somewhere! Contact pads under CPU cleaned with alcohol, CPU re-installed, IHS cleaned, and GC-2 applied (flash makes it look like there's more):

Tighten down the Tranquillo heatsink, then remove it to check contact:

Still kosher, right?

Yeah that still isn't right at all. Heatsink still doesn't even feel warm during this, heatpipes and base are the same. Behind the CPU socket it still feels like 38-40°C, I'd expect it to feel like 50-60°C back there if the IHS was running 70°C since GTX285's certainly do when running under load.

 

[Spare-Flair]

Like I told you before, you aren't supposed to use a drop for CPUs that have 2-4 cores under the heat spreader. You use a vertical line. I don't know why you keep using a drop.

These are Arctic Silver's instructions for your i3. Should be the same for MX.

Core™ i7 Processor Extreme Editions Vertical Line
Core™ i7 Processors Vertical Line
Core™ i7 Mobile Processors Surface Spread
Core™ i5 Processors Vertical Line
Core™ i5 Mobile Processors Surface Spread
Core™ i3 Processors Vertical Line
Core™ i3 Mobile Processors Surface Spread
Core™2 Quad Extreme Processors Horizontal Line
Core™2 Quad Extreme Mobile Processors Surface Spread
Core™2 Quad Processors Horizontal Line
Core™2 Quad Mobile Processors Surface Spread
Core™2 Duo Processors Vertical Line
Core™2 Duo Mobile Processors Surface Spread

Yeah that still isn't right at all. Heatsink still doesn't even feel warm during this, heatpipes and base are the same. Behind the CPU socket it still feels like 38-40°C, I'd expect it to feel like 50-60°C back there if the IHS was running 70°C since GTX285's certainly do when running under load.

I don't know where you are getting the idea that you should be able to feel a lot of heat behind the CPU socket or the heatsink by just touching it with your fingers. The motherboard is insulating, the edges of the heatsink that you can touch with your fingers and even the heatpipes by the base will be only warm to the touch on most setups. Your fingers are the least accurate way of judging any of kind temperature. Get a laser thermometer if you really care about that but it doesn't matter. If the temps are that high, it means the heat is high and concentrated in the core and isn't being dissipated by the heatsink or the board or surrounding components anyway which is why it feels cool.

 

[RickyJ]

Every manufacturer has their own recommended method, though some still somehow recommend spreading completely. According to reviews, Arctic Cooling recommends the drop, but that's just their opinion.

Line of MX-3: (pic of line is 2nd application, spread pics are of first which wasn't as consistent)

Cleaned up and reapplied the same amount and the same length:

This is driving me nuts.

 

[Skull_Angel]

the method doesn't really make a difference so long as the coverage is good. this is exactly why you check the spread whenever you're working with new components.

 

[4LC4PON3]

would lapping make a difference you think? if your not getting complete contact wouldnt that be the next solution?

 

[RickyJ]

MX-3 has always had great coverage with the grain of rice method for me in the past for LGA775 and LGA1366 builds. I've never seen a heatsink spread it so poorly, even stock Intel heatsinks played nice with this paste before.

When I unbolt the Gelid heatsink it almost rotates on its own as if it's still floating on paste. When I pressed the CPU into the heatsink base by hand, it wouldn't rotate. This has been leading me towards thinking there's poor contact force somehow. I'll try the line method with the GC-2 in the morning, at least that stuff seems to spread nicely under the tension I'm getting. Since the stock heatsink is getting such poor distribution, I'm starting to consider the ridiculous idea that somehow this socket sits lower than it should? Never heard of this before, the idea is so foreign to me.

Other voltages for reference, though nothing should raise any alarms:
IMC: 1.20V
DRAM: 1.46V
PLL: auto in bios
PCH: 1.10V

 

[RickyJ]

Busted out my Kapton tape and went to town. Lapped the heatsink base from 320-2000 grit (no pics), using WD-40 sure made it easier than water.

320-grit in progress on the CPU:

Just as I suspected, there's a ridge down the center. Here's 2000-grit:

The sides are still a bit low, but the cores run along the center vertically. MX-3 in a thin line:

Still goopy, try a dot:

Still goopy. Reapply a line:

After I finished lapping, I put a drop of water in the center of the heatsink base and put the CPU on it. The water spread perfectly and got the CPU wet to all four corners. Going to pull heatsink off again and try it with the thinner GC-2, although I only have enough for one application.

 

[Spare-Flair]

Okay from all your troubles and your issue with goopyness, I suspect the issue is simply that your heatsink retention mechanism simply is not applying enough pressure on the CPU which varies with each board and the springs on the mechanism itself may have weakened. Try bending the retention legs a bit.

We know from the Venomous-X which has a pressure screw that pressure can mean huge differences in temperature.

 

[RickyJ]

Yeah that's what I've been suspecting. The springs on the bolts are pretty beefy, but the arms are flexing too much I think. Where the arms bolt to the base is slightly different on the two sides so one pair sits a bit more angled, looks like a manufacturing defect.

Going light with the GC-2, I verified that I'm getting great distribution and decent consistency with a thin line, but temps didn't drop at all.

I'm trying to think of how I can increase the contact pressure without having the heatsink retention arms just bend even more. I could space the backplate out a bit, or put washers between the bolt heads and springs. I still have the threaded rods and stuff from my old watercooling gear, but I'm not sure if the threads are right for the backplate, and then the fan wouldn't clear.

 

[RickyJ]

Figured out how to even out the heatsink mounting arms, aluminum foil shim. Also put a couple 1mm thick washers between the springs and bolt heads.

Stock setup:

Aluminum foil shim, folded for two layers so the arms match:

Springs shimmed, and arm shim installed:

Line of MX-3:

Pretty tough to get the threads started now, and there's a lot of force on them. See springs:

Heatsink doesn't spin as easily when I remove the bolts. Looks like the added pressure is helping:

Try a dot of MX-3:

I'm thinking that the thicker MX-3 bunches up when the heatsink is removed, lots of suction. Try a line of dots:

Well that's still not right.......

 

[Ocellaris]

try speading an even layer of thermal paste on the cpu and stick on the heatsink. use a credit card and spread an even layer across the cpu as thin as possible. in some of those pics you are using way too much paste and that is not helping anything. also maybe your new chip just runs hot, there are difference in each chip that can affect temps.
Posted via [H] Mobile Device

 

[Spare-Flair]

try speading an even layer of thermal paste on the cpu and stick on the heatsink. use a credit card and spread an even layer across the cpu as thin as possible. in some of those pics you are using way too much paste and that is not helping anything. also maybe your new chip just runs hot, there are difference in each chip that can affect temps.
Posted via [H] Mobile Device

He is definitely using way more paste than I would ever use myself.

I use Arctic Silver though.

Could just be a faulty heatsink though. He's mounted and unmounted the thing so many times. The only way to see if the temp readings are correct or not is to try a different heatsink. Years ago I had a Thermalright heatsink on my AMD XP that had one defective heatpipe that would always stay cool. I think the internal vapor had leaked out and evaporated.

 

[RickyJ]

I tried before with less paste, but it wasn't spreading far enough. I'm going to try it again now that I've increased contact pressure.

Even with the increased contact pressure, lapped heatsink and CPU, and line method, temps at stock speed and voltages haven't dropped (2°C probably from change in room temp):

 

[Spare-Flair]

Since you lapped the CPU, you should lap the heatsink as well. Have you tried that?

 

[RickyJ]

Was hidden in text, but I posted that I did though I don't have pics of the progress of it. The base wasn't as bad as the IHS, but seemed to be slightly bowed out along one axis through the center. It got the same 320/600/800/2000 treatment as the IHS, changing directions between stages.

Couple small dots of GC-2 (3rd smear is from me wiping off the syringe tip):

Seems to be pretty solid contact. Cleaned up and re-applied slightly less:

I remember running into this problem months ago after I first got this PC together, but had to redirect my time to other stuff and forgot about it.

 

[RickyJ]

At stock settings, temps still aren't any better either.

When I muster the patience later this evening, I'll use same same amount of GC-2 with the stock Intel heatsink. When I used the stock heatsink a couple days ago, it didn't require as much force to engage the pushpins as I've found with LGA775 and LGA1366 stock heatsinks.

 

[Ocellaris]

I tried before with less paste, but it wasn't spreading far enough. I'm going to try it again now that I've increased contact pressure.

Well some of your shots here are deep into the "way too much paste" territory. Warm up the paste and spread a thin layer with the edge of a credit card. Put the syringe in a plastic bag, boil some water in a cup, then toss the syringe-in-a-bag into the cup. That will make it really soften up.

Also why do you think your temps are so terrible?

 

[drescherjm]

Also why do you think your temps are so terrible?

Several times I read this thread and started to ask that very question before deleting it. The temps look fine to me. But I guess when one is trying to push a chip to its maximum reducing the temp may help..

 

[serpretetsky]

how hot is it where you are?
is this case open or closed?

 

[RickyJ]

I like the idea about heating up the paste, I'll have to give it a shot. Thin layer of paste is for smaller exposed cores, air bubbles can get caught with the larger surface area of an IHS, and you need a larger force to get the pressure to seat the heatsink properly.

Temps are terrible because they're high for no good reason. These dual cores are renowned for running damn cool, and reviews of this heatsink have it keeping overclocked quads at lower temps consistently.

 

[RickyJ]

how hot is it where you are?
is this case open or closed?

I'm in Canada, room temp is in the 18-20°C range. Initial testing was done with the case side panels on, but I've kept the case open to keep things easier. Beside the heatsink is a 120mm fan, and above it is a 140mm fan, so it's not like heat is getting trapped either.

Unplugging the case fans and stopping the CPU fan with my finger, the heatsink eventually heated up while running LinX at stock speeds. So there is thermal transfer going on, but the temperature differential between the temp sensor in the CPU and the heatsink itself is quite large. Just doesn't seem right how the CPU can be north of 70°C and any part of the heatsink is ambient temps when the fan is running, heatsink should be heat saturated to get those temps.

The only other heatpipe heatsinks I've used are Xigmatek 1283's on i7-920's, and the heatsinks heated up just fine under load (and core temps didn't break 65°C). My last personal machine was watercooled for 6 years with a custom loop, thermal transfer followed logic there too.

The only two possible explanations in my mind are:
1) Temp sensors aren't reporting properly.
2) Not enough heat (Watts/Joules) for the heatpipes to work correctly, or the heatpipes are defective.

 

[Silverghost]

Maybe stopping put such a huge insulating amount on? I use rice grain size then spread it manually too all for corners with my finger stretched through a plastic bag. The paste only has to fill the grooves left for sanding now.

Just because its a dual core doesn't not mean it will run cool. At 3.6 it will hit 90 C with a stock heat sink so 70C isn't so bad.

 

[RickyJ]

Somehow I don't think you meant this thin, spread it with my finger in saran wrap:

Heated up the MX-3 in a mug of boiling water as suggested, much easier to work with! Couple very small dots corresponding with the two separate dies under the IHS:

Spreads quite well, looks like the bunching up happens from the suction of pulling the heatsink off. Clean up and reapply a couple slightly smaller dots:

 

[serpretetsky]

First of all, i'm honestly not sure how bad those temps are. I saw that review you posted of your heatsink and i guess according to that guy your temps are pretty high, but i was always under the impression that it's normal for a stock core i3 under load to be in the 50's. Again, i dont really know

People keep telling you to try different paste applications, but i honestly dont think its going to make much difference for you with the amount of applications you have tried. I wouldn't bother lapping either guy either.

You've already pointed out contact pressure might not be enough so i'm sure you've attempted to increase that.

You said you've tried both the stock heatsink and yours, and both are fairly cool to the touch, and in both cases the CPU sensor is reporting very high temperatures

So, assuming that these temps are indeed TOO high, and that the sensor is accurate. We would have to conclude that either you have TWO heatsinks that are defective, something i would think is unlikely.

The other option, is we would have to conclude that you have A VERY HUNGRY CPU. Somehow you got a processor that is consuming more power than the average i3. I see that hardware program is giving you wattage for the cpu. I dont know how accurate that is, but is it inline with what other i3's consume?

Assuming the temps are too high, the other possible scenario is the sensor is not accurate. Now, i'm not sure this is likely either. Don't modern systems have multiple cpu sensors? Like in your pics with the hardware sensor, there's a temp for each core, and even ANOTHER cpu temp sensor (cputin).

Basically, i'm assuming you haven't mounted it incorrectly with every one of your attempts. I'm assuming the temps are indeed too high. You've tried two different heatsinks that you claim are overheating. I'm pretty sure there are multiple temp sensors, low chance ALL of them are bad. So maybe it's your processor.

 

[Korben Dallas]

I was reapplying some AS5 to my Athlon 64 yesterday and I noticed the same bunching or curling of the paste when I lifted the heatsink to check the spread pattern. I have never seen AS5 look like that before so I wondered if maybe the paste was old. After some deliberation I noticed there was some paste removing solution (AC's 2 step kit) was getting soaked up into the TINY slots at the base of the heatsink fins. When I would reseat the heatsink the leftover solution would leak out onto the paste/cpu making the thermal paste dilute with cleaning solution(which eventually cause the clumping/curling). After I noticed this, I blasted the tiny slots on the heatsink with compressed air and a significant amount of cleaner came dribbling out. After I blasted it with the can of air and and let it dry for a few minutes there was no more clumping to be seen.

I'm not sure if you follow all that but looking at your pics of the clumped up paste, you might be experiencing something similar.

 

[Silverghost]

Ok, maybe more than a grain of rice. That was pretty thin

I just read this. http://www.hardwarecanucks.com/foru...79-gelid-tranquillo-cpu-cooler-review-11.html

If the heat sink is cool, maybe the temp sensor is off. My thermal paste also doesn't clump like that when I take it off.

 

[RickyJ]

Haha, thought I'd give it a shot since I lapped them flat, wasn't expecting it to even post! From my searching, it looks like the i7's got decent thermal probes, but Intel cheaped out for the Clarkdales so they're only kinda reliable near TJmax, apparently.

To get a better idea, I dug out my old Cooler Master fan controller and temp monitor from my old watercooling setup. The probes are nice and proved to be quite accurate during testing (ie: body temp under tongue, room temp). Trimmed the excess plastic from one to get the actual sensor cell inside as close to the cores as possible, added a touch of MX-3 to help bridge any possible air gap, and secured it down with my Kapton tape. I know it's not perfect, but it's as good as I'm going to get without machining the IHS like Kyle here. Excuse the flash:

Also taped one to one of the two heatpipes that are directly over the CPU core (not the uncore):

Under standard (for this thread) 3800MHz at 1.3V, the probe taped to the IHS hits 34.5°C and the heatpipe hits 34.0°C. On my desk they were reporting a rather warm room temp of 24°C.

That HardwareCanucks review was one of the main ones I looked at before I bought this heatsink. I figured if it could keep an overclocked i7-920 at decent temps, it'd be awesome with a cool i3. Price was about $30cdn on sale, and fit+finish was much better than the Xigmatek heatsink I've used before.

I think the only way I'll get a better reading from the IHS is to use thermal adhesive, or mix super glue and thermal paste.

 

[Spare-Flair]

Those probes are terrible at giving you accurate core readings. The one on the heatpipe is fine but the one touching the edge of the headspreader will tell you nothing. Believe me, I have the exact same thing in my system.

My cores will go from 40° to 70° C and the thermistor touching the edge of the heatspread will tell me the temp changed from 38° to 44° C

The only way to get any accuracy from those is the [H]ardOCP way of milling a channel into the heatspreader.

 

[RickyJ]

Yeah I figured as much. Probe was giving me 38°C under full load at 4.4GHz, RealTemp was giving me 84°C. Heatpipe probe moved a bit to 37.5°C, heatsink was finally getting a bit warm.

With the bowed IHS and the shitty reported temps, I've been wondering if the IHS was making proper contact with the dies.

 

[Korben Dallas]

I guess the only logical step is to open her up...

 

[RickyJ]

Haha I've been hoping to avoid that. Engineering Samples of i3's have gone nude before, not sure if they used adhesive on production versions. Early P4's could go naked, but they switched from thermal paste to thermal adhesive later on.