Gigabyte’s so-called “Server-Grade Thermal Conductive Gel” and a degrading thermal paste on the Radeon RX 9070 Gaming OC

1 - Overview and an important foreword 2 - Thermal putty: Consistency, microscopy and issues 3 - Thermal Putty: Measurements and material analysis 4 - Thermal paste: problem case and microscopy 5 - Thermal paste: measurements and material analysis 6 - Summary and conlusion

What is thermal putty and how does it get onto the cooler?

Thermal putty is a pasty, viscoelastic material for thermal coupling between heat-generating electronic components and heat sinks. In terms of function, it is somewhere between classic thermal pastes and solid thermal pads, combining a certain flowability with a stable structure. In contrast to paste, it retains its basic form after application and does not require exact dosing for each application. It can embed into surface depressions with moderate mechanical pressure, filling gaps without completely running or segregating – at least ideally. The thermal performance is based on a fine-particle dispersion of inorganic, electrically non-conductive fillers in an organic carrier matrix. Typically, silicone-based polymers (usually polydimethylsiloxane, PDMS) are doped with oxide particles such as zinc oxide, aluminum oxide or silicon nitride.

Compared to conventional polymer pads, thermal putty allows significantly better adaptation to uneven or variably high components, as it yields specifically under pressure but does not melt like a paste. At the same time, it generally retains greater structural coherence and is more resistant to pump-out effects than low-viscosity thermal pastes. The contact resistance to the contact surface can be improved by targeted compaction, provided the material has not been formulated to be too brittle. This is particularly relevant for power electronics where large component tolerances occur or low contact pressures are available, as is common with GPU VRM zones or uncooled RAM banks.

In industry, thermal putty is preferably applied in large quantities using spraying or dispensing processes. Pneumatic or servo-controlled dosing systems are used here, which apply the material from cartridges or large containers to the component in finely defined quantities via flexible lines. The material is applied in dots or as a continuous bead and requires precise coordination of viscosity, thixotropic behavior and grain distribution. For reproducible processing, the putty must lie within a narrow flow window: it must not run in a static state, but must yield sufficiently under shear in order to be extruded through nozzle systems at a defined pressure. The consistency is typically between 100,000 and 500,000 centipoise, with temperature-dependent thixotropy being a key control criterion.

Microscopy and consistency

In terms of material technology, however, there are considerable conflicts of interest with this form of application. High thermal conductivity requires the highest possible proportion of thermally conductive fillers, which, however, greatly increase the viscosity. In order to still keep the material conveyable, the filler content is reduced or supplemented with larger particles, which worsens the homogeneity. Alternatively, softer, oil-rich carrier media can be used, which improve pumpability but tend to segregate and release oil under thermal load. This results in sedimentation, pore formation or an increased risk of bleeding. In addition, poorly dispersed or coarse-grained fillers can lead to nozzle erosion, micro-clogging or unstable flow behavior, which makes complex process control necessary.

The microscopic images shown here clearly demonstrate the effects of such a formulation optimized for sprayability. The overview at 1000 µm scale shows a blurred material edge with uneven edge adaptation. This indicates a weak matrix structure with low local compaction.

The observed characteristics certainly indicate that during continuous operation, especially when the graphics card is installed in a vertical or orthogonal position, progressive gravimetric and mechanical creep of the putty may occur. The microscopic images with a lateral resolution of around 250 µm show a material image that indicates several potentially critical factors.

Firstly, clearly distributed cavities in the form of pores and channels can be seen, which extend both in depth and horizontally within the matrix. This porosity is most likely due to air bubbles trapped during application, which were either incompletely displaced from the paste during dispensing via a dispenser or even introduced by the abrupt closing of the dispensing nozzle. Too high a viscosity in combination with too fast a feed rate also favors such inclusion formation.

On the other hand, the clear demarcation of the material close to the edge in the direction of unwetted component edges is striking. This indicates a comparatively low surface tension of the carrier matrix, which means that there is insufficient spreading on hydrophobic or less wettable substrates. Together with the recognizably weak adhesion to metallic surfaces, this not only creates the risk of partial delamination, but also promotes material migration under sustained mechanical or thermal stress.

In this specific case, this means that a vertically installed graphics card is subject to permanent shear stress over long periods of time due to the weight of the putties themselves, exacerbated by cyclical thermal expansion. The combination of internal porosity, low adhesion and a lack of cohesive strength can then lead to parts of the material slowly shifting laterally or being displaced from their original position – a process that accelerates further depending on the installation position and operating temperature.

Targeted modification of the rheological properties, for example with additives to stabilize the structural viscosity and adhesion promoters to improve adhesion, would therefore be urgently recommended to counteract this potential long-term problem. Without such measures, the use of the analyzed putty in this form remains a clear risk factor for the thermal stability over the entire service life of the card.

Particle analysis

High magnification under polarization contrast reveals the actual inhomogeneity of the particle distribution. Particles with diameters of less than 3 to more than 20 micrometers occur together, partly embedded in agglomerates or isolated. The dark areas between the particles are evidence of matrix retraction, possibly due to thermal relaxation or viscoelastic stress. The next image after the tear-off edge clearly shows the wave-like surface structure that occurs in materials that only deform unevenly under load. The high light refraction gradient and the spatial topography are a clear sign of plastic creep with simultaneous local matrix failure.

Finally, the last image shows a strongly decoupled grain structure with open pores and partially visible sedimentation edges, which indicate an irreversible material change. Particle binding is weak, the bulk material loses coherence and tends to crumble mechanically.

These characteristics indicate a thermal putty that was primarily designed for industrial processing with an automatic dispenser. The formulation was apparently adjusted to focus on pumping performance, stability and dispensability – at the expense of microstructural integrity and long-term thermal performance. The observed compromise between flowability and grain bonding inevitably leads to suboptimal behavior under thermal cycling. Although the applied layer is initially dimensionally stable and covers the entire surface, it begins to age prematurely, which manifests itself in the reduction of the thermally active surface and the increase in interface resistance.

Alternatives to the dispenser

A technically more sophisticated, albeit more cost-intensive alternative to this solution would be the use of pre-pressed putty material in pad form, as already used by a few competitors. These are prefabricated, pressure-compressed molded bodies that can be cut to size and applied like a normal thermal pad. During production, the material is calendered or laminated into a defined geometry that precisely specifies both the thickness and the grain distribution. The great advantage lies in the consistent thermal performance, the clean application and the ability to simplify complex assembly processes. However, as these pads are significantly more complex to manufacture and have a lower material yield, they are often too expensive for mass-produced products in the consumer class. Here, it is better to opt for the more cost-effective material that can be sprayed – even if it is called “Server Grade Thermal Conductive Gel” to promote it.

Does the putty really run out?

Yes, it can, unfortunately. The next picture clearly shows two different material classes of thermal interface materials in direct comparison. On the left is a mechanically manufactured, pre-pressed semi-putty pad from Ziitek with a stable, grid-structured surface. On the right is the classic dispenser putty in unhardened form, as used by Gigabyte (presumably from the same OEM). The darker, blurred edge area around the dispenser putty, which has formed after just two days at room temperature on a simple sheet of paper, is striking. This behavior is a clear indication of incipient material migration, i.e. the escape of low-molecular components from the polymer matrix.

This effect is referred to in the technical literature as bleeding. It occurs when the carrier matrix is not able to sufficiently bind or stabilize volatile, low-viscosity silicone oils. These oils are usually added to the formulation to improve the processability of the material. They reduce the viscosity, facilitate the dosing and injection process and support the short-term flowability during pressing. In the long term, however, they cause the material to lose its dimensional stability. Migration takes place through diffusion along surfaces or through capillary action in porous substrates – as here with paper. The fact that this effect is already visible after 48 hours in an undisturbed resting state indicates a comparatively weak physical retention of the volatile phases in the matrix. This instability is further exacerbated under thermal load or mechanical stress.

This becomes particularly problematic when graphics cards are mounted vertically or orthogonally. In addition to thermal stress, gravitational forces also act in one direction over longer periods of time in this installation position. If the putty is not sufficiently structurally cross-linked or at least rheologically stabilized, it begins to creep under its own weight. This creep process can cause the material to move out of the contact zone, spread unevenly or build up at the edges. The thermal coupling deteriorates locally and inhomogeneous thermal resistances with potential hotspots arise. This is particularly critical for components with point loads such as VRMs or memory modules, where complete surface coverage is required to ensure thermal integrity.

In contrast, Ziitek’s pre-pressed pad shows no visible migration. This is due to the compaction and shaping carried out during production, which stabilizes the geometry of the material and reduces the free oil content to a minimum. Such materials are usually based on polymer-modified silicone elastomers with low oil separation and higher internal cohesion. The thermomechanical stability is usually significantly better, which makes them particularly suitable for applications with vertical mounting or longer operating times.

Observation of the wet edge around the Gigabyte putty therefore suggests that it is a formulation optimized for injection processability, where long-term thermal stability and structural integrity appear to have been given lower priority. This decision may be understandable from a cost perspective, but inevitably leads to problems in demanding or long-term applications. Creep effects, oil separation and material migration are real risks that can lead to functional impairments depending on the installation orientation and thermal profile of the card. Let’s take a look at this.

But the real problem comes later in the form of the highly problematic heat-conducting paste, because degradation is pre-programmed. A purely accidental find, as always.