AMD Ryzen 5 2400G delidded and temperature tested

[Corky34]

AMD stuff seems to be decent:

Not sure why the need to copy the benchmark I've already seen, seems decent isn't exactly very scientific.

Also I am not sure what Z height has to do with it,since they are re-using the IHS,ie,its not a direct die cooling solution they are using.

I was referring to the Z height between the die and the IHS, they say they replaced the AMD epoxy fixing the IHS to the substrate but they said nothing about replicating the Z height so for all we know, and going by their description of how they re-fixed the IHS, the distances between the core and IHS has been reduced.

From my own experience of using several types of non-conductive thermal compounds in a SFF system,most of the decentish stuff is more or less the same,and if anything the spreadibility of the compound seems to make much more difference. Its most likely all of these formulations are made by a few largish OEMs anyway,and just put into fancy tubes for the parts companies.

True, the application probably makes more difference than the type.

Like i said from what i can tell AMD are using TC-5022, I'll try to find some more details if i get the time, but TBH I'd be surprised if AMD are using anything drastically different than what Intel uses.

[CAT-THE-FIFTH]

Not sure why the need to copy the benchmark I've already seen, seems decent isn't exactly very scientific.



I was referring to the Z height between the die and the IHS, they say they replaced the AMD epoxy fixing the IHS to the substrate but they said nothing about replicating the Z height so for all we know, and going by their description of how they re-fixed the IHS, the distances between the core and IHS has been reduced.

Which is going to be closer to how people doing it in the real world,and in the end I doubt its really going to make much difference. Most of the decent non-conductive stuff is within a few degrees of each other.

It shows the aftermarket non-conductive thermal compound to be slightly better,but not massively.

True, the application probably makes more difference than the type.

Like i said from what i can tell AMD are using TC-5022, I'll try to find some more details if i get the time, but TBH I'd be surprised if AMD are using anything drastically different than what Intel uses.

My main systems have been SFF since 2005,and if you are going to notice big differences in temperature,its the kind of system where any differences would be magnified due to thermal constrictions.

Going back to this:

http://www.tomshardware.com/reviews/...n,5108-11.html

The AMD stuff is equal to some of the better compounds.

There is only a few degrees difference and I have used IC Diamond which is near the top of the list too,and it was a pain to apply to get a thinner coating. I have used perhaps probably close to 8 of the non-conductive compounds in the top half of that list. I would just get the cheapest one and the one which is easiest to spread. I would not be spending £13.50 for 5g of Thermal Grizzly,when 8g of MX2 costs £5!







I remember using Arctic Silver Ceramique 2 ages ago!! Its better to just use cheaper and reasonable quality stuff and re-apply it more often TBH!!

In the end why its an issue for Intel is because the K series CPUs can generally overclock very well,and they use quite small chips,so it means things get very toasty.

For the APUs its not really a big deal,as even with conductive thermal compound you cannot overclock any better,and in the end even with an Intel CPU I wouldn't bother delidding as I want to have a warranty still!

[peterb]

Of course if you want the best heat transfer, dont use thermal paste at all, just lapp and polish the surfaces to remove the small imperfections so you get a metal to metal surface (which is what some enthusiasts do) - although probably not practical for mass production. Thermal paste is just a filler for the air gaps in normal metal to metal surfaces - it has a lower thermal resistance than air, but still higher than metal/metal.

[Spreadie]

For those with access to a 3D printer, there are a bunch of design on Thingiverse for de-lidding CPUs.

Oh, good shout. I might run a couple of those off if I can spare the time tomorrow.