GN: Der8auer Deep-Dives on Thermal Paste: Misconceptions, Curing, & More | LTX 2019

Snowdog

[H]F Junkie
Joined
Apr 22, 2006
Messages
11,262
Some insights into developing thermal paste with Der8auer. He did a lot dev testing for Thermal Grizzly and now owns part of the company:

 
This was an interesting watch. However, I wish Der8auer provided more information on which paste he sells that does not suffer from pump out. It sounded like he was recommending Hydronaut if you don't want to have to keep applying every year.
 
There are definitely things I didn't know about thermal paste in there. Honestly, I generally buy the cheapest and most plentiful pastes when I need them. With all the motherboard and CPU reviews I've been doing, I go through it like crazy. Premium pastes are hugely expensive comparatively, and I use vast quantities compared to the average enthusiast. I've often used the more premium stuff in my own rig, but I don't bother with the test bench. Those systems just need to work for a few days at most. It's not uncommon for me to swap processors or cooling solutions during testing. I also occasionally use it on video cards. I'm a bit of a clean freak, so I'll take apart a dusty cooler on a GPU and clean everything up until it looks new and then reapply some thermal paste and put everything back together.

A larger tube that a normal enthusiast might buy which will last for a couple of years or more I'll go through in a couple of months.
 
There are definitely things I didn't know about thermal paste in there.

Same here, and it has made me less keen on high end pastes, since I basically want to put it in there and forget it. According to this, the high end stuff is typically the thinner ones, prone to the pump out effect.

My computer is from ~2008, and somewhere around the 9 year mark the CPU fan failed and while replacing it and cleaning up, I checked the paste and noticed that it had dried to hard baked grey mud. I replaced it with the only thing I had around, a tube of generic hardware store white thermal paste, that I picked up when studying electronics tech in college in the early 1990's. So the cheapest of cheap pastes and the tube was over 20 years old at that. Seems to be still working fine on my overclocked C2Q.

My thermal paste might be older than some of the posters around here. :D
 
Even Der8auer stated that the difference between a low end and a high end thermal paste is only 3 or 4 degrees. Of course, LN2 overclocking is a different matter and you need something designed for that vs. ambient cooling via air or water. I could use the high end paste's myself as I typically replace motherboards, CPU's and cooling solutions relatively frequently for the most part. I ran a Core i7 5960X @ 4.4-4.5GHz for 4.5 years. However, in that time I ran three different AIO's and an external custom loop on it. I also changed the motherboard from the Rampage V Extreme to the Rampage V Edition 10. I also changed cases once during that period, so the system was literally rebuilt for that reason. As a result, I had to reapply thermal paste 6 times in 4.5 years? Anything that lasts a year is probably good enough for me. I'd take the extra couple of degrees in my own system as a result.
 
Thermal Grizzly - the paste they sell is known as one of the most overpriced worst paste ever.
 
Citation needed.
yPacjeY.jpg


here you go. overshilled thermal grizzly.
 
View attachment 177512

here you go. overshilled thermal grizzly.

So some cherry picked comments from random unhappy people on forums, and how far off mainstream did you have to go (and what language) to find something that looks like a review where it finished mid pack?

Typical results of 85 paste test. Kryonaut assumes it's normal position as the most effective non conductive paste:
aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9YL0gvNjk4NDUzL29yaWdpbmFsL2ltYWdlMDA4LnBuZw==.png


Sure it's expensive, but it's VERY effective.
 
Last edited:
1 site showing their results? not much of an argument either.

// and this one doesn't even have Tuniq, EGC,

Its either that bait and switch tactic, or just bad material batches from time to time. (but quite frequently.)
 
Last edited:
1 site showing their results? not much of an argument either.

First link in google search. And Toms is a widely recognizable site in English, Where did you get your graph?

It's one of the top if not the #1 most effective paste after liquid metal, in nearly every test at every reputable site, so here are some more after Toms:

27 pastes: Tied for first:
https://www.hwcooling.net/en/the-test-of-27-thermal-compounds-part-2-en/2/

Fewer pastes: But finishes first:
https://www.kitguru.net/components/...to-head-does-it-matter-which-brand-you-use/3/

47 pastes: First place after liquid metals:
https://overclocking.guide/thermal-...d-with-air-cooling-and-liquid-nitrogen-ln2/6/
 
First link in google search. And Toms is a widely recognizable site in English, Where did you get your graph?

It's one of the top if not the #1 most effective paste after liquid metal, in nearly every test at every reputable site, so here are some more after Toms:

27 pastes: Tied for first:
https://www.hwcooling.net/en/the-test-of-27-thermal-compounds-part-2-en/2/

Fewer pastes: But finishes first:
https://www.kitguru.net/components/...to-head-does-it-matter-which-brand-you-use/3/

47 pastes: First place after liquid metals:
https://overclocking.guide/thermal-...d-with-air-cooling-and-liquid-nitrogen-ln2/6/

some russian site they often release benchmarks early, i don't remember the name as its in russian | i don't speak russian. This is something i compiled a year ago.
 
Prolimatech PK-2 is my go-to. PK-3 is better, but much harder to get an good application. PK-2 works very well with the pea method on AM4 and Intel sockets.

I use PK-2 on my personal systems.
 
Also, I gave rgMekanic all the samples I had purchased to do TIM testing. I think ~25 different TIMs... Hopefully we will see something out of him soon. A lot of work to do a fully new test scenario however.
 
Also, I gave rgMekanic all the samples I had purchased to do TIM testing. I think ~25 different TIMs... Hopefully we will see something out of him soon. A lot of work to do a fully new test scenario however.

It is still in the works for sure, in fact it's the test I'm looking forward to doing the most, It's just an incredibly time consuming test to perform. Since I test in a case, + am planning to do a 30 minute test per TIM, with the mounting and unmounting of the cooler between, it's gonna have a lot of hours in it.

Couple that with the fact it's hotter than Satan's taint in FL right now, I would have to dedicate specific hours, probably at night, to do the TIM testing to make sure the ambient is as stable as possible throughout all the tests. Even if I put everything on the back burner except the TIM roundup, it would likely take me about 2 months to get through it :(
 
inI've used a few different pastes over the years. Rarely have I used a paste longer than 2 years, without re-applying (not because I needed to. Just because I often end up refreshing paste, along with other maintenance).

I've never had any paste fail after a year or two. Especially not via "pump out". However, I understand how it could happen with a softer paste. But in my experience, most pastes don't last many years, anyway.

I used to use Arctic Silver Ceramiqe, because its known as "semi-permanent" and when I was younger, the idea of that was good to me. Also, older hardware didn't blast the heat nearly as badly. It was a lot easier to keep Athlon 64's, X-2's, etc, cool. Than today's quad cores and higher.

I did have the stock paste on a fat PS3 fail after 5 years. I replaced it with AS Ceramique.

More recently I've been using the expensive stuff. Gelid GC-Extreme was my choice for the past 3 years or so. Easy to apply, better temps than Arctic Silver Ceramique. 2+ years on an overclocked 7600k with no issues. I also put it into a laptop about 4 months ago.

I recently tried Scythe's own paste, which they supplied in a tube with my Mugen 5 Rev. B, because I couldn't find my GC-Extreme. The Scythe stuff was about 1c worse.

I just got some Thermalright TF8. I didn't apply it in a situation in which I could directly compare. But, I have seen it tested and based on a few tests, seems to be 1-3c better than GC-Extreme and also compared well with Kingpin. Application was very easy. The package claims an application will last 8 years.

*part of the reason I strayed from GC-Extreme, is due to an Indonesian guy who livestreams paste tests on youtube and he has a newer tube of GC-Extreme which consistently tests way worse than an couple years old tube he has.

For me personally, its worth it to spend a few extra dollars, to shave off a few degrees.
 
Last edited:
*part of the reason I strayed from GC-Extreme, is due to an Indonesian guy who livestreams paste tests on youtube and he has a newer tube of GC-Extreme which consistently tests way worse than an couple years old tube he has.

I know what you mean, I've seen that in many tooth pastes. (I have pleasure of applying plenty of paste to servers)
though I never had a -really- bad batch of paste on servers.

we had ordered some kryonaut since one of the admins was telling us how great it is - well it turned out to be 40% great, 40% mediocre, 20% below expectations. While Artic and Noctua produced more linear results almost every time we bought them. I personally noticed that each generation after ivy bridge benefited more from denser pastes - that put more pressure.

(Now that I think about it, it may be a global problem with contact plates, as maybe the quality is going down year after a year - to point where amd gpu's now also suffer from poor contact or chip makers cheap out.)
 
I know what you mean, I've seen that in many tooth pastes. (I have pleasure of applying plenty of paste to servers)
though I never had a -really- bad batch of paste on servers.

we had ordered some kryonaut since one of the admins was telling us how great it is - well it turned out to be 40% great, 40% mediocre, 20% below expectations. While Artic and Noctua produced more linear results almost every time we bought them. I personally noticed that each generation after ivy bridge benefited more from denser pastes - that put more pressure.

(Now that I think about it, it may be a global problem with contact plates, as maybe the quality is going down year after a year - to point where amd gpu's now also suffer from poor contact or chip makers cheap out.)
That's an interesting note about dense pastes.

According to Der8auer, denser pastes may have more longevity. Maybe I'll pick up some IC-Diamond, next time I want to switch out paste. See how it compares to this Thermalright TF8.

Der8auer also said that they have a new Kryonaut ready to release and a new Conductonaut nearly ready. So maybe they noticed consistency problems and fixed it with a more sustainable recipe?

*and yeah, the contact plate on the new AMD cards is pretty disappointing. AMD really missed with the reference coolers on NAVI.
 
Come to think of it, I had used GC-Extreme for a re-lid and my results seemed to get slightly worse over time. So, maybe that was pump out?! Maybe GC-Extreme isn't suite for being directly on the core.
 
Come to think of it, I had used GC-Extreme for a re-lid and my results seemed to get slightly worse over time. So, maybe that was pump out?! Maybe GC-Extreme isn't suite for being directly on the core.
GC Extreme is known to dry out quickly. You would have to replace that stuff every year. I would never touch the stuff for any application.
 
Prolimatech PK-2 is my go-to. PK-3 is better, but much harder to get an good application. PK-2 works very well with the pea method on AM4 and Intel sockets.

I use PK-2 on my personal systems.

Just saw this post. Thanks for the recommendation. Ordered a tube of PK-2 from Amazon today.
 
That's an interesting note about dense pastes.

According to Der8auer, denser pastes may have more longevity. Maybe I'll pick up some IC-Diamond, next time I want to switch out paste. See how it compares to this Thermalright TF8.

Der8auer also said that they have a new Kryonaut ready to release and a new Conductonaut nearly ready. So maybe they noticed consistency problems and fixed it with a more sustainable recipe?

*and yeah, the contact plate on the new AMD cards is pretty disappointing. AMD really missed with the reference coolers on NAVI.

I got some IC Diamond. I've been using that on the test bench lately. I ran out last night and I won't be getting more. I'd say skip it. The performance is decent, but its expensive and it doesn't spread worth a crap.
 
GC Extreme is known to dry out quickly. You would have to replace that stuff every year. I would never touch the stuff for any application.
Not in my experience. I've used GC-Extreme for 2 years before changing and it was fine. The only reason I changed it for fresh paste, was because I got a new heatsink.
 
Been using LM for years and years for GPU and CPU. Never had any issues. It's the only stuff I've never had to touch.
 
I got some IC Diamond. I've been using that on the test bench lately. I ran out last night and I won't be getting more. I'd say skip it. The performance is decent, but its expensive and it doesn't spread worth a crap.
I've been using IC Diamond 7 for awhile. Performance is good, but when I removed the heatsink after 2.5 years, it looked like there may have been some drying/air in between. Hard to say.

It is abrasive, and will polish your heatspreader, but after 2.5 years, I can still clearly see the batch info. It is near impossible to remove all of the compound (diamond dust?), because it seems to embed itself in the spreader. I think the key is to use 91% isopropyl to first remove all the grease with a q-tip and then use some pressure with a dry q-tip to remove the dust. The heatspreader feels polished.

The only way I could see it actually removing the etching, would be if someone were using a high pressure heatsink and moving their PC around alot or reapplying the paste many times. Otherwise, the only time there would be any grinding action would be during cleanup.

Going to try Coolermaster MasterGel Maker Nano. Uses diamonds, but they seem to have figured out the polishing problem and results are similar or better than most of the others. Have heard Grizzly Konductonaut dries out. Also concerned the aluminum would react with the copper heatsink, over time or be conductive. Having used the liquid metal stuff, it's conductive and likes to fuse to heatsinks - that's out.

Pics of my 1400 after 3-4 reapplications and 2.5 years of use.
 

Attachments

  • 20190818_124244[1].jpg
    20190818_124244[1].jpg
    329.7 KB · Views: 0
  • 20190818_124322[1].jpg
    20190818_124322[1].jpg
    392.5 KB · Views: 0
Last edited:
Back
Top