ykot wrote on 2020-08-16, 13:42:

Did you measure the time it takes to heat the heatsink from room temperature and operating "stable" temperature? Because there are multiple parameters to watch for (that is, the whole temperature curve), not just some temperature at some random point of time.

I had this aluminium foil (I don't think it's graphite) on my Deschutes P2 as well and several socket 370 heatsinks too, but in practice I've found it not so good for heat transfer. Before removing the foil, you could do a very simple test yourself - disconnect the fan and power up the PC and measure how quickly the external heatsink becomes uncomfortably hot for your hand to touch. Then, remove the foil and instead apply thermal paste, but not just at the place where the foil was, but either at the whole surface or at least in multiple large spots, then repeat the test. This is subjective but very effective test - the difference in heat transfer is so big that you'll notice difference without using a thermometer.

With the fan working this was difficult to notice as the heatsink only gets mildly warm, so I agree it would have been a good idea to do some tests without the fan, unfortunately I only thought about that after I had removed the pad...

ykot wrote on 2020-08-16, 13:42:

I think this aluminium foil was used not much as a replacement for thermal pad, but as a softer surface to avoid damaging CPU's die, before manufacturers started to use integrated heat spreader (IHS) somewhere around P3/Tualatin time.

If you look at my picture you'll notice the heatsink surface is really rough, so maybe they thought the pad would provide a smoother surface for contact.

swaaye wrote on 2020-08-16, 15:34:

I found the Intel Thermal Design Guidelines for Pentium II SECC1. It's clear their preferred TIM is paste, but suggest a thermal pad might be adequate too.
http://datasheets.chipdb.org/Intel/x86/Pentiu … II/24333102.PDF

I imagine a pad is the cheapest, easiest cooling option for the assembly line.

That document has a lot of interesting info! Thanks!

auron wrote on 2020-08-16, 17:28:

all klamath and most SECC1 deschutes didn't have exposed dies, it's just the pii 450 that had been moved to flip-chip packaging afaik.

Right, as I mentioned above even some 450 P2 like mine (SL2U7) are of the "older" type, without the exposed die, where the heatsink actually contacts what Intel calls the "thermal plate" in the document posted by swaaye.

auron wrote on 2020-08-16, 17:28:

ykot wrote on 2020-08-16, 13:42:

I think this aluminium foil was used not much as a replacement for thermal pad, but as a softer surface to avoid damaging CPU's die, before manufacturers started to use integrated heat spreader (IHS) somewhere around P3/Tualatin time.

all klamath and most SECC1 deschutes didn't have exposed dies, it's just the pii 450 that had been moved to flip-chip packaging afaik.

anyway as there's thermal pads now that have been tested on modern 100-200+w CPUs and perform acceptably with those, i really don't see much reason to use messy thermal paste on 30-40w parts.

Not all thermal pads are created equal. The only one I would bother using now days is the one that Linus did a review on. Kind of pricey but if you never want to worry about your paste drying out or separating it is the way to go. Just as good thermal transfer as the best pastes as well.

cyclone3d wrote on 2020-08-16, 19:11:

auron wrote on 2020-08-16, 17:28:

ykot wrote on 2020-08-16, 13:42:

I think this aluminium foil was used not much as a replacement for thermal pad, but as a softer surface to avoid damaging CPU's die, before manufacturers started to use integrated heat spreader (IHS) somewhere around P3/Tualatin time.

all klamath and most SECC1 deschutes didn't have exposed dies, it's just the pii 450 that had been moved to flip-chip packaging afaik.

anyway as there's thermal pads now that have been tested on modern 100-200+w CPUs and perform acceptably with those, i really don't see much reason to use messy thermal paste on 30-40w parts.

Not all thermal pads are created equal. The only one I would bother using now days is the one that Linus did a review on. Kind of pricey but if you never want to worry about your paste drying out or separating it is the way to go. Just as good thermal transfer as the best pastes as well.

The same stipulation of having one that covers the whole IHS is a given though as having a smaller area to dissipate the heat from the CPU to the cooler is a bad idea.

auron wrote on 2020-08-16, 17:28:

anyway as there's thermal pads now that have been tested on modern 100-200+w CPUs and perform acceptably with those, i really don't see much reason to use messy thermal paste on 30-40w parts.

Can you please recommend a good 125W thermal pad? I just happen to be building a new rig with Core i9 10900K with NH-D15 cooler right now, so could definitely try something like this.

cyclone3d wrote on 2020-08-16, 19:11:

The only one I would bother using now days is the one that Linus did a review on.

Which one? Are you referring to this?

cyclone3d wrote on 2020-08-16, 19:13:

cyclone3d wrote on 2020-08-16, 19:11:

auron wrote on 2020-08-16, 17:28:

all klamath and most SECC1 deschutes didn't have exposed dies, it's just the pii 450 that had been moved to flip-chip packaging afaik.

anyway as there's thermal pads now that have been tested on modern 100-200+w CPUs and perform acceptably with those, i really don't see much reason to use messy thermal paste on 30-40w parts.

Not all thermal pads are created equal. The only one I would bother using now days is the one that Linus did a review on. Kind of pricey but if you never want to worry about your paste drying out or separating it is the way to go. Just as good thermal transfer as the best pastes as well.

The same stipulation of having one that covers the whole IHS is a given though as having a smaller area to dissipate the heat from the CPU to the cooler is a bad idea.

there is no such stipulation. see here: https://www.gamersnexus.net/guides/3346-therm … h-thermal-paste

as seen, the "pea size" method in that test leaves one corner of the CPU uncovered, yet the results are pretty much identical across the board. and that is with an 8086k@5.1 ghz, needless to say how much difference in TDP that is to the late 90s stuff we talk about in this thread.

the random thermal pad that came with my k6 233 does its job fine enough, still there is a charred spot on the back from past use, but that's because the tiny pentium cooler that it came with is totally insufficient for that CPU and would get hot even its fan running. i'm rather happy with swapping that for an adequate s370 cooler that i can reuse on different setups without having to deal with silly thermal paste every time.

@ykot: that IC graphite pad seems to be quite decent, but i wouldn't mind using paste in this case, because i wouldn't be frequently swapping CPUs like with older gear.

Yeah, the IC graphite pad it the one I was referring to.
If you watch the review of that one, it has crazy good heatspread properties as well. So theoretically , even if it doesn't transfer heat quite as well as some other things... The heatspread will maybe make up for that because it will be spreading the heat to the cooler possibly even better than the IHS of the CPU.

That being said, I generally use Arctic Ceramique II for most applications because it is easy to spread, easy to cleanup and doesn't dry out or separate.

I was using IC Diamond for a while but that stuff is a huge pain to apply because it is so thick and doesn't spread near as thin as I would like, even after many thermal cycles. To top it off, I have had it leave craters in a lapped IHS and a lapped Heatsink.. probably due to diamond bits being clumped together.

I'm not going to get into the discussion about thermal pads since I only use them for mosfets and ram memories, that kind of things. for the rest I use thermal paste. A tip about the scratched heatsink.
I use "Autosol Metal Polish" to polish it to a mirror-like finish. polishing it with the dremel is simple and not very laborious. The product is also very good for cleaning glues and other things.

Hoping wrote on 2020-08-17, 11:23:

I'm not going to get into the discussion about thermal pads since I only use them for mosfets and ram memories, that kind of things. for the rest I use thermal paste. A tip about the scratched heatsink.
I use "Autosol Metal Polish" to polish it to a mirror-like finish. polishing it with the dremel is simple and not very laborious. The product is also very good for cleaning glues and other things.

How do you create an even surface with a small tool like a dremel on a large heatsink like that?

I am aware of that, that's why I only polish the contact area or the whole according to the size of the heatsink, and, I know that it is still not perfect but it is comfortable, fast, and using sandpaper and other methods it is not easy to make it perfect by doing it. by hand.
I only use that polish because it is easier and faster than sandpaper.
Let's face it, most of the heatsinks don't have a polished base and I do it that way because it's easy and fast. Sorry I repeat myself:)

If that question seemed inadequate, sorry dude! I didn't mean to be a dick to you or anything. I just don't think a little scratch is going to harm temperatures in any significant way that can't be overcome with thermal paste because just as you said they're far from perfect to begin with. So why bother with a Dremel?