So I noticed my CPU temps were increasing from what they used to be, from when I installed my CPU from about 3 years ago so I decided to reapply the thermal paste. Upon removing the heatsink something didn't look right with the paste.....

http://i.imgur.com/FxCiCMA.jpg
http://i.imgur.com/2kUFkZ1.jpg

the paste was hard and flaky/powdery and not its normal gooy consistency. When remove the paste with rubbing alcohol and cotton pad it sort of flaked off....

http://i.imgur.com/A5F5u20.jpg

The thermal paste I originally used was arctic silver, I have reapplied thermal paste many times in the past and never had anything like this before, any ideas why the paste has gone like this?7

** Resize your images or place in spoiler tags **

How did you apply it? blob, line or spread? AS5 is designed to be applied with line or blob and hand spreading can cause it to excessively dry out.

Yeah I used the clingfilm method. Blob in the middle of the heatsink and the put clingfilm over your middle finger then pull taught and spread paste to a thin layer over the heatsink

I really don't recommend that method with the popular arctic silver pastes unlike some other pastes they don't tend to fill the contours properly with that method - which looks like it has meant it has dried out excessively here judging by the images.

That's really odd as that's the method I've always used and not had a issue in the past, live and learn I guess.

So the line method, just run a line down the centre of the heatsink from top to bottom?

I've not seen it as bad as that before - line method works best if you know where the cores are under the IHS if in doubt the blob method works almost as well (probably won't even notice a difference in real world every day usage outside of a lab).

AS5 for instance really isn't designed for spreading though that is by far the worst of that kind of problem I've seen - might be the consistency of the paste in this case was lacking as well or something.

EDIT: For a 4770K then the best thing to do IMO is to add a line that follows the interface material as shown in this image:

http://images.anandtech.com/doci/8227/2 TIM.png

Stopping slightly short of the edges to allow for some spread without oozing out from under the heatsink.

I've used AS5 by spreading it on IHS with a plastic card for too long and never had an issue. Never used blob/line in the middle method for my rig or ones I've built for others. Been using AS5 for so long now, my Q6600 rig (built at launch) is still using AS5, repasted probably 3-4 years ago, still used for "stuff". When I built my i5 4690K rig I even used a tube of AS5 which must have been several years old, didn't note any separation/liquid from application on IHS.

I ended up using a kind of cross pattern from corner to corner.

Have been using x264 Stability Test for the past half hour now and temps haven't gone above 55c on any cores, before i was hitting like 78c so massive improvement!!

I also reapplied CLU to the dye as i had delidded the CPU back 3.5 years ago, or though the CLU still seemed fine thought i might as well redo it as i had the heatsink off!

Seems you sorted it .

I forgot to mention earlier if this had happened to me I'd have done an application of TIM fitted HS as usual then removed it to check contact. If all good redone application.

I suggest two things- first, stop using arctic silver pastes. There are better pastes both in terms of performance, as well as application and removability. There are also many pastes which are not capacitive and contain no metal.

I suggest Noctua NTH1 as the easiest paste to work with closely followed by MX-4. Both are affordable and tend to sell for less than AS5. They also have no curing time.

I also suggest that you look up proper application methods for your specific CPU. Ideally, you want to spread the paste very simply- just draw a line or a dot and let the weight and pressure of the HSF do the rest. Also, applying the paste in a way that is specific to your particular CPU is best. For example, modern mainstream INTEL CPUs have a die that is long and rectangular, so drawing a line along where the CPU die is, is ideal. With a square die CPU you want a nice blob in the middle of that die.

Very rarely do you want to spread the TIM yourself. This tends to introduce air bubbles and with many pastes, as has been mentioned, this will cause premature deterioration, like what has happened to you.

Just some thoughts.

Cheers.

The spreading method works well for very runny pastes, with a thick paste you might get air bubbles trapped in there if the cooler isnt applying enough pressure to push them out which would probably cause it to dry out like that.

Oh i just noticed sprinkles mentioned the same thing above...ignore.

Just put a blob in the middle and get on with it. Some people try to make science out of the most straight forward tasks.

In future when posting oversized images from imgur onto the forum put an "h" before the .jpg in the imgur URL. Imgur will then create a resized image that is 1024 pixels wide. "l" creates an image 640 pixels wide.

String said:

Just put a blob in the middle and get on with it. Some people try to make science out of the most straight forward tasks.

Click to expand...

Many traditional thermal pastes/adhesives were explicitly designed to be hand spread hence where many people get the habit from.

Arctic Silver actually explicitly stated on the original* documentation for several of their pastes that they should be applied using the line method for optimal results of failing that a blob and NOT to spread - while generally the difference is only really that relevant to a lab test it can produce sub-optimal results and excessive drying out, etc. though I've not seen anything like the OP on anything that hasn't been left for like 15-20 years :S

Winds me up when some reviewers do round ups of various pastes, says they are going to hand spread all of them for "fairness" and then marks AS down accordingly on performance - while not quite up there with the cream of the crop any more they still do very well when applied properly.


* They seem to have updated the documentation with less of an emphasis on it now so possibly newer variants are more tolerant to mixed application methods.

To me it looks as though the heatsink has been mounted with more pressure to one side and the other side has had enough of a cap to allow the compound to dry out.

With any heatsink compound I just put a blob in the centre of the CPU and apply the heatsink in a criss-cross pattern.

If its 8 cores or more, I would do 5 blobs. quarter blob nearer to each corner

silversurfer said:

If its 8 cores or more, I would do 5 blobs. quarter blob nearer to each corner

Click to expand...

Look at the image I posted further up - for many CPUs when using AS stuff the best results come from sticking a line that matches the interface of the IHS and die and let it naturally spread so you get optimal coverage where the most heat is passing through.

silversurfer said:

If its 8 cores or more, I would do 5 blobs. quarter blob nearer to each corner

Click to expand...

Because why? Depending on which two CPU generations we compare, you can have a dual or single core die that is the same size as, or larger than, an 8 core die. For example if we compare 65nm to 14nm. If we were still on original integrated circuit lithographies, a 22 core xeon chip would be the size of a bus.

String said:

Just put a blob in the middle and get on with it. Some people try to make science out of the most straight forward tasks.

Click to expand...

That works only on Intel CPUs not AMD ones though.
Because of the different shape the cap has.