Hey Canary and yes I have abused your nice little miners a little bit since I got them but they are still alive
You are technically correct that the grey thermal pastes usually contain silver particles which are themselves probably conductive, however a large percentage of the material is actually oils and binder materials (the exact composition is usually proprietary though you can work it out with some acetone, filtering equipment, and precision balances). These binders are generally non conductive so you end up with something that is a less conductive than a single piece of solid metal.
Of course, alumina and other metal oxide based (white) pastes are generally assumed to be entirely non-conductive because most metal oxides are excellent electrical insulators.
Just how conductive is silver based thermal paste? Well I did a quick test, I took some Arctic Silver 5 and smeared it over some perfectly smooth plastic in about that kind of thickness you would use for any CPU/GPU. Then I took some aluminium metal and made two conductors on either side. The conductive path of paste was 50 mm long and 10 mm wide and then I used a multimeter to measure the resistance across the paste:
For the measurement I used my digital multimeter, which caps out after 40 megaohm. Just how much is 40 megaohm? Well its not allowing much conduction at all. I managed to get 7 megaohm measuring between my fingertips on each hand, through my body, pressing very hard on the probe pins. Pushing lightly with my fingertips, I could raise the resistance until it reached 40 megaohms and then finally went over its limit.
The 50x10 mm strip:
However, it did not give any measurement, indicating the resistance was >40 megaohms! I even pressed down firmly on the aluminium to ensure it was in good contact with the paste on both sides.
Not one to give up, I reduced the area to 10x10 mm. Again, no reading was possible as the resistance was far too high.
So I reduced the measurement area to about 10x 0.3 mm, perhaps the spacing of pins you would find on a chip. In the photo, you can barely see the gap between the aluminium conductors its so small:
Again, it was suprisingly still >40 megaohms.
I raged, and simply pushed out a glob of AS5 and dug the multimeter pin probes straight into the paste:
Still, there was no conductivity.
I found that even if the metal probes are lightly touching each other inside the paste, there is still no measurable conductivity. In the picture above the red probe is actually resting ontop of the black probe, but still no measurement can be had. Only by pushing the tips together fairly hard within the glob can you get a reading (and as you'd expect, when the metal tips actually touch, it goes to <0 Ohm ie. the other limit of the DMM's measurement range.)
This is kind of suprising to some extent but then I realise that I have seen plenty of CPU and GPU where the onchip SMD capacitors have been drowned in thick silver based pastes but still functioned perfectly. In fact in my experience this is an amusing 'problem' that used to be seen a lot with Pentium 3 and 4 equivalent Xeon chips in severs (especially from Dell!). This is kind of like how you can probe pins on a running PCB circuit with digital components using an oscilloscope: because of the high probe inductance/resistance it usually won't effect the circuit/devices operation. I realize that at kHz and above frequences other electrical effects in seemingly non-conductive materials become considerable though I think in this case you can get away with it thanks in part to the design tolerances of the components and signal:noise will probably still be pretty good.
tl;dr - Getting silver thermal paste (at least Arctic Silver 5) on pins / exposed parts of CPU/GPU/ASIC has no effect on the performance.Edit: In the
overclocking thread I have discovered that capacitive effects from AS5 actually DOES cause problems with the USB miners. However I found that Ceramique 2 compound worked fine, although it too would have some capacitance so YMMV. The optimum solution seems to be using 2 sided tape.
PS: I scraped some 'dried' thermal paste off some scrap sever CPU heatsinks and measured the resistance of that, and again it was the same. The conductivity doesn't really increase in any measurable way when the paste has set/dried.
(Disclaimer: A few manufacturers, including Arctic Silver inc, discourage you from getting it on exposed pins/components because of the other electrical effects I mentioned above. Don't assume from my shitty 10 minute tests that your $$$ device won't implode if you're messy with paste.):
AS5 instructions:
4) Arctic Silver 5 ONLY: While much safer than silver greases engineered for high electrical conductivity, Arctic Silver 5 thermal compound should be kept away from electrical traces, pins, and leads. Arctic Silver 5 is slightly capacitive and could cause problems if it bridged two close-proximity electrical paths.
Although this argument is all kind of irrelevant... because its easy to not make a mess with the thermal compound and avoid getting it on the pins, but also because the ASIC die is apparently positioned upside down relative to how die are usually set in epoxy on a CPU/GPU. As a result, cooling is actually superior
through the PCB and into the heatspreader rather than putting a tiny heatsink on the topside of the ASIC. See my link above in the overclocking thread for more info. That's why the manufacturers went with a heatspreader under the PCB instead of shipping them with a tiny heatsink ontop.
