I am comparing different thermal conductivities in the market and found that the thermal conductivity goes up to 1600W/m·K for a thermal sheet, but a thermal paste is still 0.65 W/m·K.

Does it make a difference to use either? How are they calculated and tested?

Just out of curiosity, I purchased both products and used them between a CPU and heatsink. I compared the temperature difference between the CPU and heatsink using a thermal probe.

As you would probably guess, the graphite sheet performed terribly and caused the CPU to overheat.

enter image description here

enter image description here

  • 4
    \$\begingroup\$ You have a typo. That's not "solder paste". It's "thermal paste". \$\endgroup\$
    – Transistor
    Commented Jan 19 at 15:16
  • \$\begingroup\$ Mhairi - Hi, To comply with the site rule on referencing, please edit your question & add the document or webpage name & link to the source of each image. TY \$\endgroup\$
    – SamGibson
    Commented Jan 19 at 15:21
  • \$\begingroup\$ Graphite is conductive, thermal paste is not. \$\endgroup\$
    – bobflux
    Commented Jan 19 at 15:36
  • \$\begingroup\$ What if it? What will the interface between your chip and the graphite sheet behave like? Will you improve it by applying a paste in between? Will the graphite retain its properties when almost powdered between the electronics and heat sink? How will you limit particulate? \$\endgroup\$
    – Abel
    Commented Jan 19 at 16:03

4 Answers 4


Well, check the datasheets. The latter is pyrolytic graphite, an electrically conductive material that is highly conductive in-plane, but rather modest thru-plane. The quoted number is the former, not the latter. It's good at spreading out heat laterally, but doesn't help with air gaps, for which grease or pad is still required.

The first item is a paste, ceramic particles in an oil (usually silicone) base, which conforms between components in a joint, squeezing out and excluding air (an excellent insulator). Thermal conductivity is essentially irrelevant because the material mostly squeezes out; thermal resistance of the joint is dominated by interface effects, the insulator if any, and the parts themselves.

Note that copper itself is ~400 W m-1 K-1, so a copper sheet/foil 0.1 mm thick has as much heat-spreading capability as this specialized, pricey material. In other words, 3oz copper -- one can achieve a similar effect by fabbing the PCB with inner layer planes with as much total thickness. The cost of pyrolytic graphite is therefore justified only in very specialized circumstances, where there is truly no room to spread out heat otherwise, and most likely where a PCB isn't present, like across the tops of components, or components in places where a PCB won't fit anyway (Idunno, cell phone cameras let's say?). It's a fantastic material when you need it, but rare that you do.

The type of sheet needed for general component insulating use (TO-220s, etc.) is either a thin insulating "foil" (cleaved mica, or polyimide film) plus grease, or a thicker rubber compound loaded with conductive particles -- "Sil-Pads" or the like (which is squishy enough that grease is not required). None of which are exceptional thermal conductors; ceramics are the best among them (Al2O3, and even moreso, AlN and mixtures/alloys) but obviously are quite stiff and don't conform between surfaces, and are very brittle so poorly-fitted surfaces can even crack them.

Some quite soft rubber materials are available, with quite high thermal conductivity; I've seen over 6 W m-1 K-1. Simultaneously an advantage and disadvantage of these materials, they are so soft they can't be clamped down with screws or stiff springs; some are tacky enough they can be placed between components without clamping force, but preferably a modest-tension clamping spring is used.

  • \$\begingroup\$ Thanks for the answer. Funny enough, I bought both and tested them and found the thermal paste to performs better. \$\endgroup\$
    – Mhairi
    Commented Jan 24 at 16:11
  • \$\begingroup\$ If you used the foil alone, yeah, it won't do very well; it's meant to be used with paste or gap-pad. But if you don't need to spread heat around, it'll just add two more interface resistances, another layer of paste, and provide no insulation, so still come out worse \$\endgroup\$ Commented Jan 25 at 0:38

The unqualified 1600 W/(m⋅K) number is a bit of advertising sleight of hand since that number is the in-plane (XY) value. Pyrolytic graphite is, on the microscopic level, sheets of carbon atoms that are weakly bonded to each other (sheet-to-sheet). The carbon bonds are very effective at transferring heat (see the thermal conductivity of diamond) but PGS is highly anisotropic and the Z-plane conductivity is probably an order of magnitude less, which makes it useful for heat spreader applications but not especially good for thermal interfacing. For example, in this datasheet, Panasonic does not even tell you the Z-plane (or C-plane, as they call it) conductivity because it's that's not why you pick PGS. This other datasheet for a PGS interface material gives Z-plane conductivities of 28 W/(m⋅K). At 50% compression, it's thicker than a good or even adequate thermal paste application so you're still choosing it mainly for heat spreading. We tried using PGS sheets for an application that had a small heat-generating area on a relatively enormous heatsink to try to spread the heat out a bit but any benefit was negated by the extra thermal interface it created.

Edit: the 1600 number is a lot of sleight of hand; note that the RS datasheet makes no mention of anisotropy or Z-plane conductivity being less than that of the XY-plane and just gives a single number for thermal conductivity which is, in my opinion, extraordinarily negligent if not intentionally misleading.

Thermal paste, on the other hand, is made of a thermally conductive material - boron nitride is a common choice - in a carrier vehicle. The heat needs to go through both the particles and the relatively unconductive carrier so even though the particles themselves can be highly conductive the overall value is fairly low. The advantage being that the paste can squeeze into and fill the microscopic pits and scratches in the materials being joined and provide the greatest surface for heat transfer, something a PGS sheet cannot do well. Properly applied, the bond line should be a few dozen microns so even your 0.65 W/(m⋅K) paste (and, as others have mentioned, there are better products out there) is going to be a negligible contribution compared to your heatsink resistance to ambient.

  • \$\begingroup\$ I don't know what they're like today, but historically the company whose datasheets are being discussed was very poor and ensuring that they all used consistent units: some would be SI while others (for a similar part from a different manufacturer) weren't and so on. \$\endgroup\$ Commented Jan 20 at 13:03

You want a conductive path through solid materials with good thermal conductivity. If your two surfaces are perfectly smooth, then you don't need an interface. But they often aren't perfect. You need something that will deform to make up for the unevenness, or something to fill in the gaps. Gaps are bad, the thermal conductivity of air is about 0.02 W/m·K. Either of your proposed materials is much better than air.

You need to determine which is best for you. Thermal paste isn't nearly as good but may be good enough and will cost less. Thermal paste is a hack, something that deforms is preferred IMO. If cost is not a factor and a graphite pad will deform enough to fill the voids, then use graphite.

  • \$\begingroup\$ I think you've missed the point! Why is one apparently 2000 times better than the other? \$\endgroup\$
    – Transistor
    Commented Jan 19 at 16:51

You can get heatsink paste that is a lot more conductive than that. Beyond conventional paste there are liquid metal "pastes" that are more than 100x higher conductivity. Sony for example uses these on the PlayStation 5.

To an extent yes this will make a difference. But the thickness of the interface material matters too. Halving thickness will double heat transfer, same as doubling conductivity. For very thin layers the heat transfer can be very high even if the conductivity (per meter) is low. Usually exotic high conductivity materials are reserved for very high thermal density interfaces (tiny high power computer chips for example).

As for graphite/diamond, carbon crystals/sheets are very conductive. But they're also solid materials, so the actual interface won't make perfect contact unless atomically flat. The gaps will limit heat transfer.

  • 2
    \$\begingroup\$ Not just layer thickness, but also interface quality--paste will do a much better job of getting in crevices in the part and heatsink than the graphite will, for instance. \$\endgroup\$
    – Hearth
    Commented Jan 19 at 17:43
  • \$\begingroup\$ This doesn't really answer the question either. Why is one 2000 times better than the other. The contact interface is unlikely to be a factor in the calculations by the manufacturers. Thickness changes the conductance but not the conductivity. \$\endgroup\$
    – Transistor
    Commented Jan 19 at 17:55
  • \$\begingroup\$ @Transistor As I said carbon crystals are very conductive, at least in isolation. The actual interface would be nowhere near that value though. If you want to know the quantum dynamics of those bonds that make them able to transfer heat efficiently, thats probably a question for the Physics stack exchange. \$\endgroup\$ Commented Jan 19 at 18:13
  • \$\begingroup\$ It should be noted that liquid metal thermal compounds contain gallium, so are not suitable for direct contact with aluminium. (See "gallium vs aluminium" videos for graphic illustrations of why.) \$\endgroup\$ Commented Jan 20 at 20:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.