Heatsink paste degridation

[dawgtuff]

A friend of mine said to repaste the CPU/GPU HS paste every 1.5/2 years depending on use or OC'ing. He said the paste degrades and dries out to a clay like substance and doesn't transfer heat as well. Is this true?

 

[silverpig]

You want it to dry out. Most of that stuff is just silver particles in some kind of solvent to make it liquidy. The solvent sucks at conducting heat, the silver is good. You put it on and the solvent dries, leaving just the silver.

I suppose your silver could oxidize, making it less conductive.

Of course this could be moot if there are other kinds of non-silver based pastes out there (ceramic?).

 

[Forumpanda]

I would still advice not to mess with what works, reseating a 2 year old heat sink can cause its own set of problems imo.

However cleaning it out of dust is probably a very good idea!

And as the previous poster pointed out, I think it depends on the type of paste, and I don't think any great studies have been done to prove it either way.

I would say that most of the 'I redid my paste and now my temps are lower' stories ceoms from the fact that people at the same time also clean out dust in their computer and thus improve airflow in general.

 

[Modelworks]

Paste does not go bad once installed. I have opened equipment that is 30 years old and the paste is still working fine.

 

[Colt45]

^that. I think it's usually zinc oxide in some silicone based oil, for the plain stuff.

 

[SunSamurai]

Originally posted by: Modelworks
Paste does not go bad once installed. I have opened equipment that is 30 years old and the paste is still working fine.

paste today != 30 year old paste
hardware today != 30 year old hardware




...just say'n.

 

[Modelworks]

Originally posted by: aeternitas

Originally posted by: Modelworks
Paste does not go bad once installed. I have opened equipment that is 30 years old and the paste is still working fine.

paste today != 30 year old paste
hardware today != 30 year old hardware




...just say'n.

Not sure what you are trying to say, but there is no reason to replace paste on a heatsink unless you remove it.

 

[funkymatt]

Originally posted by: Modelworks

Originally posted by: aeternitas

Originally posted by: Modelworks
Paste does not go bad once installed. I have opened equipment that is 30 years old and the paste is still working fine.

paste today != 30 year old paste
hardware today != 30 year old hardware




...just say'n.

Not sure what you are trying to say, but there is no reason to replace paste on a heatsink unless you remove it.

this is not 100% true. I have worked on several socket A systems over the years that were overheating due to "old" thermal conductive paste. As the OP said, it gets dried out. I suspect what is happening is the drying out/hardening is causing air pockets and then the processor overheats.

 

[Modelworks]

Originally posted by: funkymatt


this is not 100% true. I have worked on several socket A systems over the years that were overheating due to "old" thermal conductive paste. As the OP said, it gets dried out. I suspect what is happening is the drying out/hardening is causing air pockets and then the processor overheats.

The whole reason it is semi-fluid is so you can apply it. They even make paste that are designed to dry before attaching the heatsink. What provides the heat transfer is not the liquid but what is in it, for white paste, that is usually zinc oxide + oil. Oil conducts poorly. If you find a cpu overheating and have to replace the paste then it is because it was not put on right in the first place.

Like I said, I have opened 30 year old equipment that used paste on transistors that got way hotter than any cpu and the paste is fine. There are several IEEE papers on it if you have a subscription.

 

[Absolution75]

If it dries out and the heatsink is nudged - I would guess that this then moves the heatsink off of the original setting, allowing those microscopic pores to accept air and thus doesn't conduct heat as well (though. . . . it wouldn't make that much difference). Most really old thermal paste I've seen is like hard glue and is generally fine.

 

[NeoPTLD]

Originally posted by: Modelworks
Paste does not go bad once installed. I have opened equipment that is 30 years old and the paste is still working fine.

Have you tested the before and after thermal conductance to validate your claim?

 

[Modelworks]

Originally posted by: NeoPTLD

Originally posted by: Modelworks
Paste does not go bad once installed. I have opened equipment that is 30 years old and the paste is still working fine.

Have you tested the before and after thermal conductance to validate your claim?

The equipment was still functional and had no heat related problems. Seen this many times.
There are several IEEE studies that show, if it is applied right, that the paste will last however long it takes for zinc oxide to break down. There really isn't much there to go bad, it is basically a form of metal in a semi liquid form so you can apply it, If it dries out the zinc is still there doing its job. Zinc takes a very long time to corrode.There is even gear from the 1950's that use compound and mica insulators that still works fine with the original paste.

When was the last time you opened up your home stereo to replace the compound ? Or car radio ? Both operate at much higher temps than a cpu and the paste goes right on working for years.

 

[NeoPTLD]

The heat density (W/cm^2) is so high on CPUs and such that silver based pastes are being used. As you can see, surface modification makes all the difference in the way they work. Arctic Silver and conductive paint are both very high in silver content, yet one is nearly an insulator while the other is a very good conductor.

You just proved my point. These legacy devices are not even pushing the limit of thermal resistance. Computer CPUs have a high heat density and a requirement to hold a LOWER temperature than older equipment. This demands a low thermal resistance path from the chip to the air.

When the paste dries out and compound turns to dust, it will no longer fill the gap as the device/heatsink changes dimensions with thermal cycles.

 

[Modelworks]

Originally posted by: NeoPTLD
The heat density (W/cm^2) is so high on CPUs and such that silver based pastes are being used. As you can see, surface modification makes all the difference in the way they work. Arctic Silver and conductive paint are both very high in silver content, yet one is nearly an insulator while the other is a very good conductor.

You just proved my point. These legacy devices are not even pushing the limit of thermal resistance. Computer CPUs have a high heat density and a requirement to hold a LOWER temperature than older equipment. This demands a low thermal resistance path from the chip to the air.

What does arctic silver have to do with it ? Are they the ones spouting the gibberish that people should 'update' the paste on their cpu every so often. Hmmm increased sales ? Companies like Arctic Silver became popular mainly because of hype, they created a demand for a product that wasn't really needed. Do you see anyone like Intel buying their stuff in bulk ? You will not because it is not needed. Arctic Silver pitched that unless you used their product , the extra temp of 5C would shorten its life, and it will, it might lose a couple months off the 10 year life span, but who is going to care 10 years from now ? That is what is so annoying about overclockers. They think that because a cpu is above their body temp it is going to fry. Just read the forums with people who are worried because their cpu is 100F, its like some tech version of mass hysteria.

You obviously don't know much about older equipment. There are some transmitters that operate so close to the temperature failing point that if it is too sunny outside the transmitters could fail. If compound was so important and stopped working like you are suggesting they would be changing it on those devices every week.

CPU operate fine with current methods of not replacing the compound, most could even work fine with no compound at all , hotter, but the majority would still operate normally. I worked for Cray and Sandia from 1995-2000 and we never replaced the paste on ASIC that were handling extreme speeds for the time. They ran very hot, one processor we used ran at just 6C from the failure temp. I remember it well because there were 3 ASIC per module so when one failed we would say , 666, satan strikes again . Flouride based coolant was needed to keep the heatsinks operating. The topic of replacing the compound never came up during that time. If it would have made a difference we would have replaced it. It would have been like saying, lets replace the wiring because after 4 years, its all used up.

When the paste dries out and compound turns to dust, it will no longer fill the gap as the device/heatsink changes dimensions with thermal cycles.

If there is proper pressure applied to the heatsink/semiconductor interface the amount of thermal change in transfer is next to none. We will both be dead and gone before thermal compound on a device turns to dust.

 

[silverpig]

I concur with Modelworks here.