Heatspreader, thermal paste and material analysis
Of course, this is my favorite topic at the moment, but why not go into more detail about the heatspreader? First of all, there is a microscopic examination of the surface, because this is fundamentally different from that of a current Intel CPU. You can make nice comparisons without benchmarks. AMD on the left and Intel on the right, clicking on them then results in the popular full screen. And yes, AMD’s surface looks like the snout of a pug that has run into a wall more than three times while hyperventilating. Wrinkled mountains in copper, but with valleys of different depths, as we’ll see in a moment:
 |
 |
In AMD, the depressions are only up to 2 to 3 µm deep, but there is a greater gradient in the overall course, so that one can assume 20 to 30 µm both in the horizontal extension and in the vertical level. Conversely, leaving aside the additional bending due to clamping in the base and production-related curvature, we can assume a minimum layer thickness of 30 to 50 µm at the thinnest, i.e. the most concave point for the thermal paste.
The next picture shows the tear-off of the paste (in the picture a DOWSIL TC-5550) and we can also see the small beads in the form of the larger aluminum oxide particles lying in the depressions. You should have seen this to understand my constant comments on interface resistance. Here we can also see the boundary layer.
What happens if you use a lousy paste (EC360 Diamond) with particles that are too large (more than 15 µm and significantly larger) can be seen in the next picture. The cooler will no longer rest optimally at this point, but will instead stand on stilts and the paste will become a runaway object due to its lack of viscoelastic properties. With Intel, the effect is even more pronounced, as the surface is less rough overall.
This brings us to the recommendation: Any normal to viscous paste is suitable for the Ryzen 7 9800X3D, the largest particles of which should normally be less than 15 µm in size. If it’s not too cheap, it will last longer. It doesn’t have to be a premium paste costing 20 euros per gram, a good paste in the 4 to 5 euro range will do. That’s what my database is for, which I’ll link to here:
Material test
I have now drilled a laser hole with 15 individual shots in an unsuspicious corner and evaluated the plasma. The galvanic nickel layer is around 5 µm thick (plenty) and we can also see the slow transition to pure electrolytic copper inside the IHS. And don’t forget: a thermal transfer resistance is also created here. In addition to the loss of warranty as a disadvantage, the surface grinding of a CPU also has certain advantages.
With that out of the way, let’s move on to the test system and the methods.
- 1 - Introduction and technical data
- 2 - Heatspreader details and thermal paste
- 3 - Test setup and methods
- 4 - Fast overclocking and undervolting
- 5 - Gaming Performance HD Ready (1280 x 720 Pixels)
- 6 - Gaming Performance Full HD (1920 x 1080 Pixels)
- 7 - Gaming Performance WQHD (2560 x 1440 Pixels)
- 8 - Power consumption and efficiency
- 9 - Temperatures, summary and conclusion
78 Antworten
Kommentar
Lade neue Kommentare
Urgestein
1
Veteran
Urgestein
Urgestein
Veteran
Veteran
Mitglied
Urgestein
Mitglied
Veteran
Mitglied
1
Urgestein
Neuling
Mitglied
1
Mitglied
Mitglied
Alle Kommentare lesen unter igor´sLAB Community →