Did I put too much paste?

skubany

Weaksauce
Joined
Feb 4, 2008
Messages
87
Right after HS installation:
CPU: Intel Devil's Canyon i7-4790K (no OC other than what the MB does by default, manual RAM settings to match advertised specs)
4.4GHz, 1.312 vCore, CPU power: 116W (max)
MB: Asus Maximums VII Hero
Cooling: Corsair H80i 120mm, Push 25xx RPM max/Pull 22xx RPM max (limited to avoid annoying whine sound), stock fans, adaptive fan speed
Corsair Pump RPM: 2209
Liquid T: 42.6C (max)
Ambient T: 25.7C
Paste: AS 5

----- Update ----
~3 weeks after installation:
CPU: 4.4GHz, 1.312 vCore, CPU power: 118W (max)
Liquid T: 43.7C (max)
Ambient T: 26.4C

Looks stable at 75C (twice went to 76C (red color) briefly) in AI Suite 3 after 24 minutes of stress testing. Fans started spinning at max again when temp hit 75C. Corsair CPU temp: 82.3C (max)
----

I took just under 18 minutes of loading all 8 threads using SP2004 by Johnny Lee to reach temperature of 82.1C (Corsair Link) / 75C (AI Suite 3).

A note about other fans in the system. Once AI Suite 3 went over 74C all my fans (connected to MB via headers) started spinning at max speed. I normally have them at comfortably audible rotation. 75C was also reported in red by AI Suite 3, while <=74C were in white.

I took about 16 minutes for the CPU to cross over 80C (Corsair Link), it was between 76-80C during the first 16 minutes.

Once the temp crossed over 80C it seems like it was going higher and higher slowly, as if the cooling was not keeping up with the CPU anymore. I stopped the stress test.

Did I put too much paste?
Are my temps pretty much what should be expected by using mid-level cooler? I know there is a twice as big radiator that can be top mounted which probably would fair better.

T0 (original application, updated picture)
http://i.imgur.com/c5myLMi.jpg

T1
http://i.imgur.com/E6TBUDM.jpg
Right after HS installation:
CPU: 79C (Asus), 84.8C (Corsair) after 17 minutes.
CPU: 4.4GHz, 1.312 vCore, CPU power: 117W (max) Once or twice 1.328 vCore.
Liquid T: 41.8C (max)
Ambient T: 26.7C
Stayed in 79C for a while until the end of the test.

T3 (formerly T2) [1/3/2016]
http://i.imgur.com/Id1V9yG.png
Right after HS installation:
Appears to have a little more paste than T1.
CPU: 78C (Asus), 83.7C (Corsair) after 21 minutes.
CPU: 4.4GHz, 1.312 vCore, CPU power: 117W (max) Once or twice 1.328 vCore.
Liquid T: 42.2C (max)
Ambient T: 26.7C
Once or twice hit 79C briefly.
After little over 3 weeks:
CPU: 79C (Asus), 85.2C (Corsair) after 22 minutes.
Liquid T: 43.3C (max)
Ambient T: 26.3 - 26.4C

T4 [1/31/2016]
http://i.imgur.com/TxS9Hfe.png
Right after HS installation:
CPU: 80C (Asus), 85C (Corsair) after 21 minutes.
CPU: 4.4GHz, 1.312 vCore, CPU power: 120.8W (max)
Liquid T: 43.1C (max)
Ambient T: 26.6 - 26.7C

T5 [1/31/2016]
http://i.imgur.com/5iOPj2k.png
Right after HS installation:
CPU: 76C (Asus), 82.7C (Corsair) after 23 minutes.
CPU: 4.4GHz, 1.312 vCore, CPU power: 117W (max)
Liquid T: 42.3C (max)
Ambient T: 27C
Peak: 77C (briefly about 6 times)

Based on the above tests my conclusion is that very little paste is needed. In the final, 'X', application I could have probably used less paste. Maybe even 2x as less. No matter how much paste, after HS is applied, the paste must cover the entire CPU.

In all cases where CPU was not covered entirely I had higher temps. I expect the 'X' application to cover the CPU entirely. 'X' is the paste application I'm running right now.

So far it seems like the T0 application had the lowest temps, though I'm expecting the 'X' application to have similarly low temps. That is 75C steady with 3-4 brief crossing to 76C.
 
Last edited:
you want a ultra thin layer of paste properly feathered out to the edge
 
apply_1.jpg


Like that.
 
What paste did you use?

Did you remove the block after installing it to make sure that the paste spread evenly and covered the whole heatspreader?

The pea size dot in the center generally doesn't work form the dozens of times I have tried it.

What works best from what I have found with newer CPUs is to put a blob in the middle and then some more to make an X coming out from center.

Then I reapply and tighten down the cooler.

Then I remove the cooler and apply a tiny bit more around any places that it didn't spread to and then reinstall the cooler.

Been doing it like this for a few years now and it works great.

As long as it is tightened down properly, any excess paste should squish out around the edges.
 
JSYoajQ.jpg


That looks like way too much, you shouldn't use so much paste that it would allow you to dip a French fry in it. Fortunately it will just blob over the sides and the difference between way too much and just right is often just a few degrees.
 
What paste did you use?

Did you remove the block after installing it to make sure that the paste spread evenly and covered the whole heatspreader?

The pea size dot in the center generally doesn't work form the dozens of times I have tried it.

What works best from what I have found with newer CPUs is to put a blob in the middle and then some more to make an X coming out from center.

Then I reapply and tighten down the cooler.

Then I remove the cooler and apply a tiny bit more around any places that it didn't spread to and then reinstall the cooler.

Been doing it like this for a few years now and it works great.

As long as it is tightened down properly, any excess paste should squish out around the edges.

Removing and replacing the cooler with the paste on it is a recipe for air bubbles.
 
There is a problem somewhere - there shouldn't be that big g of a delta between the liquid temp and the CPU. I have a similar setup (H100 and 4790k) and use don't think I've seen my CPU temp cross 50C while stress testing.

Could be paste, could be the tubing pinched, could be the pump failing, could be a bad lid on the chip, could be bad temps (either software showing the wrong numbers or the TCs are not calibrated correctly.)
 
I still don't care for the "drop in the center" method.

I've found it results in uneven spread.

I normally use a plastic spreader, make sure it's THIN and stop just short of the edges to give it a little room for expansion when I compress it.

More effort and a little messy on my hands/clothes, but I always have gotten good results with it.
 
OP, that's enough paste for at least 2 chips, maybe 3. Irregardless, your temps aren't too bad for a single 120mm rad. With better application you may see a drop of 3-5C, tops.

DC is a hot gal no matter what.
 
Definitely too much paste. I agree with Brian_B that there's a problem somewhere, but I doubt it is the paste. Maybe the block is not mounted flush across the whole CPU, or there's a flow issue. I think the water temps are in the expected range considering it is a single rad, but the core temps are high imo.
 
Friendly 2 cents of advice to OP: Your blob is a bit big, and since it's a conductive material if too much bleeds out it could short something. You don't have to cover the entire IHS, just the part over where the cores are. This link will show you what is under the IHS so that you can determine the correct area that needs coverage https://www.pugetsystems.com/labs/articles/Delidding-the-4790K-for-a-quick-look-at-the-TIM-573/

Because cores are now a long rectangle, a small line on top of them (vertical on IHS face) is all you need. Spreading causes air bubbles to form which can impede performance: This is an old video but you get the idea https://youtu.be/EyXLu1Ms-q4?t=111

Every AIO liquid cooler I've had shipped with fans that left a lot to be desired. Everybody's ears are different, but I've enjoyed using Noctua fans for my radiators due to their increased static pressure and SSO bearings. For your size radiator I'd recommend trying the NF-F12, which should allow you to run a more aggressive curve while maintaining low noise levels.

As far as mounting, make sure the cold plate is flush against the cpu and pressure is even across the four corners, also make sure the support bracket and cold plate are facing the correct direction (incorrect direction on either causes 'lean'). Finally, make sure there are no kinks or sharp curves in the hoses so that the liquid can flow freely. Beyond that, you could probably try undervolting the CPU slightly and seeing if it remains stable at stock speeds, but in the long run if you're under 90c at full load you should be fine under normal everyday conditions. Even right now your deltas don't look terrible considering how warm the room is. Personally I can't sit in my leather chair above 24-25c, but everybody has their own comfort levels.
 
Let's address the responses.

Once I did paste application testing with a different (air cooling) heatsink and CPU. I found that "pea sized" dot was not sufficient to spread the paste across the entire CPU under heatsink pressure.

I do agree that it's probably best to apply paste, put on HS, remove it to see how it looks, wipe it and reapply a final time adjusting the amount as needed. That way one knows how much is needed for a CPU/HS combination, which may be slightly different between CPUs/HSs because of mounting hardware solution, etc.

I sort of agree that even if there is too much the pressure of the HS should squeeze out the extras over the sides, but there still may be a thicker layer left than needed.

I will reapply the paste again. The 2-5C different of a proper layer of paste would actually be worthwhile to go through this exercise, if only not to trigger the MB from spinning all my fans at max speed. I will post back with pictures and results once done.

I dislike spreading of the paste ahead of time because of the potential for air bubbles.

It's good to know that maybe my cooler is not functioning correctly. The pump RPMs are reported as 2200 by Corsair Link software and 2330 by my MB, to which I connect the block's RPM wire. Are those proper pump RPMs for this cooler? Maybe I'll buy another one, on some deal, to compare. 120mm is the largest radiator my case can accommodate.

I've started using Noctua fans for my recent builds. How much improvement could I expect with Noctua versus stock? Even if a fan spins at the same RPM, one can be better at pushing air than another, how much better?
Plain NF-F12 (1500 RPM seems very low) or one of their Industrial NF-F12 versions?

As for room temperature, yeah, I'm pretty much naked here :) I'm used to this. I have a quality mesh chair, highly recommend. Much better than leather.

I used to have i5-4690K just a few weeks ago in my system (under the same cooler). I think under the same stress load it was around 50C max. Don't remember exactly, but I do remember the temps were quite low.

A final statement about reapplying. I would think removing the heatsink to check how the paste spread, adding extra in any area that is not covered and then reapplying the HS to the same paste would be a recipe for air bubbles/disaster. If HS comes loose, then old paste needs to be wiped entirely.

"horrorshow" that's sort of what I would expect to be the required amount. I'll try that next.
 
Tubes are not pinched. Curved approaching 80 deg in one place but that should be fine. The radiator is "upside down" though, with tubes at the top of it, but I don't believe that would make a difference. Am I wrong?
 
If possible the tubes should be at the bottom, but not because of effectiveness. Small amounts of air might be inside the closed loop and could potentially cause turbulence which will generate extra noise. I had this happen with an H50 and originally I thought it was the pump going out.

The difference between fans is static pressure and the fan's ability to overcome it. Moving air is easy, forcing it through tiny fins is where they really differ. I can't guarantee a performance improvement over stock fans, but it's certainly worth a shot. I've never needed to use Noctua's Industrial fans because mine never run at 100% anyways. Considering everything else, I would do this last if you're happy with the noise level of your current fans; mounting is far more important than the potentially marginal gains by switching to Noctua fans.

Pump RPMs look fine. The point of the SATA power cable is to prevent it from being connected to a PWM controller, so the pump is always running at max (which it should be). Based on my Google search it looks like 2200-2300 is normal.

The increase in temps is likely due to DC chips running at higher clocks and voltages than their Haswell counterpart. You're also going from 4 core to 4 core with HT. Granted, It shouldn't be a 40c difference between the chips, but once again as long as your stress testing it and it stays under 90c your daily use won't come anywhere close to a point of concern.
 
My take on this... Once running, the heat from the CPU will spread the paste a bit more than if you had just applied the paste for 15 seconds and then pulled the cooler off. Also you don't need to cover the entire heat plate, you just need to cover the area where the CPU cores actually are. If you put so much that it covers the entire IHS, I think you have worse contact over the core area of the IHS.
 
My take on this... Once running, the heat from the CPU will spread the paste a bit more than if you had just applied the paste for 15 seconds and then pulled the cooler off. Also you don't need to cover the entire heat plate, you just need to cover the area where the CPU cores actually are. If you put so much that it covers the entire IHS, I think you have worse contact over the core area of the IHS.

but the entire heatspreader absorbs and therefore dissipates heat from the cores. Therefore full contact with the isf would be ideal. That being said I don't think you could lose contact with the IHS over the cores by using too much paste [on the rest of the IHS] since the hsf will be spreading it out once pressure is applied.
 
One thing I forgot to add is a reminder about the break-in time for the TIM that you're using.

Taken from http://www.arcticsilver.com/as5.htm

"Due to the unique shape and sizes of the particles in Arctic Silver 5's conductive matrix, it will take a up to 200 hours and several thermal cycles to achieve maximum particle to particle thermal conduction and for the heatsink to CPU interface to reach maximum conductivity. (This period will be longer in a system without a fan on the heatsink or with a low speed fan on the heatsink.) On systems measuring actual internal core temperatures via the CPU's internal diode, the measured temperature will often drop 2C to 5C over this "break-in" period. This break-in will occur during the normal use of the computer as long as the computer is turned off from time to time and the interface is allowed to cool to room temperature."
 
but the entire heatspreader absorbs and therefore dissipates heat from the cores. Therefore full contact with the isf would be ideal. That being said I don't think you could lose contact with the IHS over the cores by using too much paste [on the rest of the IHS] since the hsf will be spreading it out once pressure is applied.

The heat spreader isn't there to absorb and release heat though, it's there to protect the CPU cores within the die. The best cooling for these modern Intel CPUs is to completely remove the IHS and put the cooler right on the CPU die. I think the best cooling while still using the IHS is the minimize the paste and make the best possible contact with the area of the IHS the covers the CPU die. When I had more patience for messing with this stuff (Q6600 and 2500K era) I usually found the best results using the "grain of rice" paste and putting it along with where the CPU die was under the IHS.
 
Last edited:
1.3v, pretty high, and result seems typical for a 120 rad. The difference on paste application don't go more than a few degree at most.

Drop your vcore and you will see a much bigger temp drop, unless you got real unlucky with the silicon.
 
I haven't OC it yet, letting MB work automatically.

How low can vCore go to maintain 4.4 GHz (max)?

I will be reapplying paste shortly (today).
 
Ok, updated first post with two additional applications of paste and updated initial picture to include pictures after HS removal.

Observations.

Looks like the more the paste covers the CPU the better the cooling.
Looks like the higher the pump liquid temp the better the heat transfer and cooler CPU.
Looks like applying paste is not really a science (little differences don't impact cooling greatly), a little too much doesn't hurt (will be pushed out the sides by pressure), too little does (will wait 3 weeks).

What do you think?

I'm looking at Corsair H80i GT, should I get it? "Extra thick radiator", might be a plus? I have a H80i from 2-3 years ago.

After three weeks I might do an 'X' shaped paste application and test again.

It looks like the original paste application was good enough and have to give a few weeks to determine if the current application is a bit too little.
 
With what I see in the image after removal, you still are using a tad too much.
 
Too much paste man. The pics posted earlier are examples of the amount needed. However, I think there is an issue elsewhere. Remount the block and see if that helps.
 
Your motherboard is ocing it. Typically you can get away below 1.25 for 4.4. In general, if a chip do 4.5 at 1.2,it is above average. I had a 4770k that can't do 4.2 at 1.2 though, sad story.

Thicker rad really don't help that much, if you like aio cooler, upgrade to 240 or 280.

But I would tune your vcore and grab a better fan first, before investing in unneeded parts.
 
for me, an old credit/gift card bent in half to make a 90* angle works great and smoothing it flat across the surface.
 
I've always went with the "grain of rice" right in the center and so far havent had any problems myself. Especially since the layout of the cores under the heat shield line up, and how it speads evenly in a circle.
 
I've always went with the "grain of rice" right in the center and so far havent had any problems myself. Especially since the layout of the cores under the heat shield line up, and how it speads evenly in a circle.
I agree.

With a quality cooler less is better.
TIM%20seat%20thickness%204_zps4ko3ixz7.png


TIM%20app%20HWS%201_zpsbjtzj7ra.jpg~original
 
It's not hard to determine for yourself whether dot or spread is better for you and the particular paste and and CPU and CPU cooler you are using.

Do 5 installations using the dot method, cleaning away all the paste between each installation. You will notice that some of your dot method installs run cooler than others. This is useful info! Keep track of the temps of each install.

Then do 5 installations using the spread method, again testing temps and cleaning away all the paste between each installation.

You should now have a sense of what a good installation is using both the dot method and the spread method. Whichever one is better, go with that. And when you do your final installation and test temp, you'll have an idea of whether you got a better or worse than average install, and be able to decide whether to try again for a cooler install.
 
What is the temp of air going into radiator? Not the room air temp, the radiator intake. I have yet to see a system with cooler / radiator intake air temp less than 2-3c warmer than room. Most are 7-15c warmer with many 12-22c warmer. Only exception is when radiator is drawing air into case from room.

I know using room ambient and CPU temp / water temp is most common way. But it's not the best way. Cooler intake / radiator intake air temp is the critical air temp, not the room.

While cooking in the kitchen we don't go into another room and look at thermometer to see how warm it is in the kitchen because the other room is most likely cooler than the kitchen. So why look at room thermometer to see what temps in computer are?


Low Cost Way to monitor air temperature different places inside of case:
  • A cheap indoor/outdoor thermometer with a piece of insulated wire and a plastic clothespin works great (aquarium / terarium / fridge remote sensor thermometers too)
  • Made up with floral wire and tape. (or single strand insulated electrical wire) If sensor is metal, wrap it with tape too. We don't want anything to short out with metal. ;)
    e65759c3-fbf9-414f-9f1f-f5941d540397_zps2fe90c77.jpg
  • Clip and position sensor where we want to check the temp. Make it easy to see what the air temp going into components actually is relative to room temp. ;)
  • When system is working air temps going into coolers will be 2-3c warmer than room.. up to about 5c is okay.
 
evilsofa, I think that is the best way to go about it and it's what I'm doing right now. Maybe next week I'll try an 'X' application, three weeks later maybe I'll try a spread application. I think different coolers and paste might require different CPU surface coverage percentage for optimal cooling. By optimal, I mean those 2-3 C differences, which in themselves aren't all that meaningful really for a day to day operation.

doyll, the way I look at room temp is for comparison purposes. Yes the temp in case will be different and the hotter it's inside the case the hotter the CPU will be. But if I run the test under the same conditions then the room temp is a predictive factor of CPU cooling efficiency. I've already saw that. Also, if I test a CPU in high ambient temp, say 26C, and it passed, I know it will be even cooler at 24C. The tests that I documented here were all done at high room temp, I've done one a few days ago at 21C, the CPU never went over 70C (Asus software temp). In 26C room the CPU heats up to 76-78C. It's almost +1 CPU temp per +1 room temp. Of course, as heat builds up in the case this will change.

My aim with these tests. See what application method works best with AS5 and my cooler. See what temp differences can be expected once AS5 cures (those ~3 weeks that the paste manufacturers mention to get optimal heat transfer).

By the way, my AS5 is a few years old now, at least 3/4. Could that affect it's efficiency?
 
Maybe you are not understanding the difference between room ambient and radiator intake air temps. Using the room ambient as a baseline gives us no idea how efficient the radiator is, not at all. What it gives us is only how efficient your system as it is now setup. Knowing the air temp gong into the radiators is the only way to monitor the efficience of radiator.

So while I understand how you are using room ambient, my point is if the cooler intake is significantly higher than it should be your results will be that same amount higher .. and you have no idea if this is because of radiator intake air temp .. or if it's because case airflow is not as good as it could be.

Lowering the radiator/cooler intake to within a few degrees of room if it is 15-20c warmer than room will lower your CPU temps by the same amount .. 10 lower if cooler intake air temp is lowered from 35c to 25c in a 20c room, or 15c if lowered from 40c to 25c in a 20c room. .. same as your CPU temps now change when room ambient changes.


Your slowly rising CPU temp is often caused by radiator/cooler intake air temp slowing rising the longer a stress load is running. . This is often caused by case not flowing the components heated exhaust out of case causing what doesn't leave cae to mix with the cool air coming into case to supply components, thus causing that air temp to rise, and keep raising the air temp going into radiator/cooler.
 
My aim with these tests. See what application method works best with AS5 and my cooler. See what temp differences can be expected once AS5 cures (those ~3 weeks that the paste manufacturers mention to get optimal heat transfer).

By the way, my AS5 is a few years old now, at least 3/4. Could that affect it's efficiency?

If I was you I would dump the AS5 altogether and get a new tube of MX4 thermal compound. AS5 is some old school shit and I don't believe it performs as well as the competition now. Also it spreads worse than other materials as you are discovering. I suspect you can get a near perfect application of MX4 in one attempt and no need for a 3 week cure time for an antiquated thermal material. At this point, you messing with the AS5 isn't accomplishing much of anything due to the excessive cure time and minor differences between application methods along with AS5's tendency to not spread well. Throw some better compound on there and be done with it.
 
Last edited:
You only need a small portion in the center. Too much and it acts like an insulation. Do note that the IHS (what you actually see when looking down on the processor) is not how large the actual core of the processor is under it. It's much, much smaller.

The TIM is only suppose to bridge the "air gap" from surfaces not being perfectly flat. You want as much metal to metal contact as possible.
 
My 4790k seems to like the thin line method down the center of the chip following the die.
 
Updated my experiment.

Added T3 (formerly T2) results after running for little over 3 weeks.
Added T4 and the current T5.
 
Interesting.
Just guessiing, but your 'X' seems to be begging for air entrapment .. all the areas encircled by TIM lines are natural air traps.

Additionally, I have found there is no reason at all to have TM coverage of entire IHS, especially on Intel chips.
Heat%20Transfer%20are%20fro%20IHS_zpsqhj1njj4.png
 
Back
Top