Cooling [SFF Network] Thermal Paste - A Roundup

[iFreilicht]

I wish you had unlimited funds so you could test this

You could always become a Gold supporter to push their funds a little closer to unlimited

 

[Josh | NFC]

You could always become a Gold supporter to push their funds a little closer to unlimited

Naw, I wouldn't want to show you up.

 

[confusis]

The issue with testing thermal paste is that some have a "cure time" (hours to days) and that it's often tested with one platform/TDP. To be accurate, one should include two TDP platforms, for a more mainstream 50-60W and a more power-hungry 90-120W. Along with a curing period of about 3 days with 50% load. But I'm not confident curing will yield much different results for most paste.

Also, a Noctua cooler should be used as these have in my experience the best distribution of pressure and you can wind these down completely without torque or warp/bend being a problem. Didn't you have an NH-L12 ? Those have the SecuFirm 3 mounting or what's it called.

Firstly, I used a NH-D15 using secufirm2.

Secondly, I don't have the budget or time for all that! (maybe when SFFn pays my salary.. ) In all honesty though, unless someone comes out with some magic bullet, the high wattage test platform should cover all bases.

 

[Phuncz]

Oh the Noctua NH-D15 should suffice ! But if you have a beefy AMD CPU at your disposal, you can probably downclock it or disable cores to adjust the TDP lower. Some motherboards even have profiles you can save so you could switch between them. Not that I'm saying you must do all this, just saying it is possible. But indeed it costs a lot of time when you have to manage a job, kids and a bunch of other stuff.

 

[jØrd]

you can probably downclock it or disable cores to adjust the TDP lower.

From a strictly theoretical perspective (because i doubt it would make any difference in practice) wouldn't this change the heat loading of the die and as such the potential performance of the heat spreader and by extension the cooler? Thinking about it though I guess that's an argument that can be made regardless for any two CPU's even those of the same model, etc, etc

Edit @Josh | NFC Whilst I dont consider Linus to be an authoritative source of actual knowledge on anything he did try it, YMMV I guess ->

Edit 2: People seem to be doing this enough to kickstart the first tool to success & lead people to design 3d printable tools like the second link
https://rockitcool.myshopify.com/
https://www.youmagine.com/designs/skylake-delid-tool

 

[EdZ]

The tool is less due to any increase popularity in delidding, but more because Skylake is harder to delid safely: the PCB the die is mounted to is MUCH thinner than previous generations (fewer layers required to route pads) and the IHS itself is thicker and sturdier, so it is easier to damage attempting to delid using the normal clamp-it-and-razor-blade method. Conversely the thinner PCB is more flexible, making it easier to remove the IHS by mechanical separation than it is with previous generations. Thus, there is a small market for tools designed to induce a controlled flex into the PCB to 'pop off' the IHS while minimising the risk of damage to the CPU.

 

[jØrd]

The tool is less due to any increase popularity in delidding, but more because Skylake is harder to delid safely: the PCB the die is mounted to is MUCH thinner than previous generations (fewer layers required to route pads) and the IHS itself is thicker and sturdier, so it is easier to damage attempting to delid using the normal clamp-it-and-razor-blade method. Conversely the thinner PCB is more flexible, making it easier to remove the IHS by mechanical separation than it is with previous generations. Thus, there is a small market for tools designed to induce a controlled flex into the PCB to 'pop off' the IHS while minimising the risk of damage to the CPU.

This makes alot of sense. I was more alluding to the bit where enough people feel there is enough advantage to delidding that there is viable support for a small after market of tools. Not empirical evidence of a benefit by any stretch though

 

[Josh | NFC]

@j0rd

I did see that one (thanks for linking though) and I am not knocking on Linus as I greatly appreciate him and his team, but I am really curious about more HD cooling setups than the stock cooler. That might be all that is needed to prove the bottleneck, but I am removing heat much faster than him and I would be quite happy with 2 degrees change.

 

[iFreilicht]

I actually like the idea this tool quite a bit more:


Not that I did any calculations, but this one seems a lot safer because the force is equally distributed on all four corners instead of just two.

 

[Qrash]

Thanks for the article, @confusis!

I've pretty much been only using NT-H1 because it never sets--which is nice because I have to replace coolers for corp customers yearly as part of service plans.

This is why I chose Arctic Alumina to replace the Intel TIM after delidding my 4790K. The Intel TIM was covering the entire rectangular die, but it was hard an brittle. I cleaned away all of the old TIM and adhesive from the core and the underside of the heatspreader and then applied the Arctic Alumina, using instructions from Arctic, including "tinting" the both sides of the heatspreader and the bottom of my Noctua NH-C14.

My initial results were very good, but Prime95's Small FFT Torture Test still gets the core temperatures up to 90 C. It just takes longer to get there now. I can't for the life of me figure out why it gets so hot.

 

[Qrash]

I actually like the idea this tool quite a bit more:


Not that I did any calculations, but this one seems a lot safer because the force is equally distributed on all four corners instead of just two.

Slick. Better than my raxor blade method, though that was fairly easy (patience is key!).

 

[Josh | NFC]

Stop linking me cool stuff that I didn't know I needed!!!!!

Man, it is sold out. I would so buy it...

 

[marce4]

Yeah almost expected, I saw one of Linustechs latest videos where they tested thermalpaste and different ways of applying it, almost nothing made any difference.

Thermal paste is really good but there isnt such a big difference between them unless you really go for high-end material. For example something like that. Then you gonna have notably better results but also a pretty steep price.