- Joined
- Jan 8, 2015
- Messages
- 87
- Motherboard
- Gigabyte Z390 Aorus Pro-F12d
- CPU
- i9-9900K
- Graphics
- Radeon VII
- Mac
- Mobile Phone
I was looking at videos about methods of cpu paste, when I came across one of those videos that use glass or plastic plates, and they managed to do this:
The video this came form is here, around the 2:00 minute mark:
If you watch the video, they're using the "spread it like peanut butter" method, and things start out fine (around the 2:00 minute mark), but then as he presses harder and harder it all goes sideways and makes this pretty fractal pattern. All the dark parts are air pockets (or perhaps even a complete vacuum). I'd estimate that IF this ever happened in the real world, your thermal results would be massively horrible. It looks like only about 30% coverage over the surface, and horrible coverage over the hot center.
CPUs and CPU coolers are not made of glass or plastic or whatever he used here, but at the same time, all things flex under pressure mechanically. So if your cooler screws down, as some do, with a flat plate bolted at four corners, which is larger than the CPU heat spreader, then the tighter you go the more you're going to flex the cooler into a concave shape above the CPU heat spreader, and potentially open up a gap. Does this ever happen in the real world? No way to be sure, but there are certainly people who report horribly bad thermal performance that's cured by a simple reinstallation.
The traditional center dot method, especially in a thicker paste, is going to be much safer, as even if you warp the interface a bit, the dot of paste will remain as the fulcrum and you'll maintain full contact in the middle of the two surfaces (at least up to a point). But you will leave all the edges our your heat spreader out of contact, and while some say that doesn't matter, I think of it as a giant air bubble all the way around the edges. Yes most of the heat goes through the center, so this is clearly the safer choice, but not perfect. After all the heat spreader... spreads heat.
I certainly believe that pressure matters (in theory there should be a linear relationship between pressure, and number of molecules in direct contact), and I certainly also believe that ideally you'd cover the entire surface of your heat spreader, and not just the middle. So I think we have a situation where the best possible paste application — the thinnest possible layer between two perfectly flat mated surfaces and mounted at very high pressure — is also the riskiest, as the moment you warp the cooler bottom into a concave shape, you risk popping the two surfaces apart and losing the vast majority of your conductivity, especially in the center.
It's also interesting to note that some coolers do NOT mount at the four corners, for example, the Noctua coolers I've looked at use a bracket at pairs of corners that lets the cooler itself screw down on two center posts. In one dimension it can still warp upwards, but not in both dimensions.
This rambling does however hint at a method one could use to optimize your performance:
The video this came form is here, around the 2:00 minute mark:
If you watch the video, they're using the "spread it like peanut butter" method, and things start out fine (around the 2:00 minute mark), but then as he presses harder and harder it all goes sideways and makes this pretty fractal pattern. All the dark parts are air pockets (or perhaps even a complete vacuum). I'd estimate that IF this ever happened in the real world, your thermal results would be massively horrible. It looks like only about 30% coverage over the surface, and horrible coverage over the hot center.
CPUs and CPU coolers are not made of glass or plastic or whatever he used here, but at the same time, all things flex under pressure mechanically. So if your cooler screws down, as some do, with a flat plate bolted at four corners, which is larger than the CPU heat spreader, then the tighter you go the more you're going to flex the cooler into a concave shape above the CPU heat spreader, and potentially open up a gap. Does this ever happen in the real world? No way to be sure, but there are certainly people who report horribly bad thermal performance that's cured by a simple reinstallation.
The traditional center dot method, especially in a thicker paste, is going to be much safer, as even if you warp the interface a bit, the dot of paste will remain as the fulcrum and you'll maintain full contact in the middle of the two surfaces (at least up to a point). But you will leave all the edges our your heat spreader out of contact, and while some say that doesn't matter, I think of it as a giant air bubble all the way around the edges. Yes most of the heat goes through the center, so this is clearly the safer choice, but not perfect. After all the heat spreader... spreads heat.
I certainly believe that pressure matters (in theory there should be a linear relationship between pressure, and number of molecules in direct contact), and I certainly also believe that ideally you'd cover the entire surface of your heat spreader, and not just the middle. So I think we have a situation where the best possible paste application — the thinnest possible layer between two perfectly flat mated surfaces and mounted at very high pressure — is also the riskiest, as the moment you warp the cooler bottom into a concave shape, you risk popping the two surfaces apart and losing the vast majority of your conductivity, especially in the center.
It's also interesting to note that some coolers do NOT mount at the four corners, for example, the Noctua coolers I've looked at use a bracket at pairs of corners that lets the cooler itself screw down on two center posts. In one dimension it can still warp upwards, but not in both dimensions.
This rambling does however hint at a method one could use to optimize your performance:
- Start by spreading a thin layer across the entire surface
- Add a very small center dot on top of that.
- Install with low pressure.
- Monitoring your temperature, give all the mounting screws a quarter or eight turn tighter ever hour or so until the temperature skyrockets (or some very expensive part breaks - disclaimer: I didn't make you do this).
- if you didn't crack your CPU, back the screws off until the temperature comes back down.
