The "summary and conclusion" says the key thing: it doesn't matter much which paste you're using.
Proper application is far more important: No gaps/bubbles, seated with proper pressure, thinnest possible layer.
I've had success in a production environment with the arctic silver 5. This was for a switch fabric IC that can hit 200W, forced-air-cooling through a vapour-chamber heatsink. FWIW other pastes worked fine too, but the factory preferred working with viscous compounds.
My beef with arctic silver stuff, is that it's electrically conductive. Not that I've ever run into an issue with it, not something I would risk on big production pieces.
However, these are in a colloidal suspension. The stuff does not conduct. I know because I had the same concern and I tried to measure it. Moreover, the voltages that something like this is likely to come into contact with is on the order of 1V for BGA packages upon which it gets mounted.
Proper application, of course, mitigates the rest of the concerns about this material.
Arctic silver is not electrically conductive. It is slightly capacitive, but even getting gobs of it randomly across a motherboard is most likely not going to do anything.
It may be my work proxy and/or ad-blocking extensions, but I frequently have issues with Chrome choosing the wrong character encoding with pages on tomshardware.com, resulting in a page full of gibberish, like this:
http://imgur.com/a/0YrWb
Reloading the page always fixes the problem. Anybody else have this issue?
That looks more like you're getting raw gzip content - perhaps your proxy is messing with the Accept-Encoding headers or trying to ungzip content for inspection and somehow messing it up.
No, but my phone ad blocker (?) kills all the figures, and with it disabled, I get page hijacking ads about 10% of the time, figures 10% of the time, and ads where there should be figures 80% of the time. I can read about half the page before the ads stop loading.
Basically, this site is completely unusable for me.
It is a shame, since the article's text does not stand up on its own, and I used to like Tom's Hardware.
Thermal Grizzly Kryonaut seems to be the standout paste after cross-referencing; it's an "8" for usability, and is the only 8 or above usability rating I could find in the top ten of both the low and high pressure air-cooling tests.
Always amuses me how there's heat sink compounds sold to the PC enthusiast market and they're completely different brands than you'd see stocked by a traditional electronics supplier.
I recently had to resupply myself with heatsink grease for non-computer uses so I got a boron nitride grease (Chemtronics CW7250), having been told that it's the best truely non-conductive material, short of the toxic beryllium oxide. I recently also used some reassembling the heat sink on a modern CPU and it seems to work just fine.
And liquid metal alloys seem like a stupidly risky thing to use.
Maybe conductivity is an issue, but I don't see why things should be set up that way.
I would ensure that both the heat sink and the top case of the CPU were grounded. I wounld't rely on that ground path, but any conduction along it would be welcome in my world.
Yeah, I have noticed it appears in some of the PC enthusiast materials, I wonder why it has not received more widespread industrial acceptance. Obviously it is very abrasive, but I think tiny synthetic diamonds aren't that expensive. Maybe it's so far down the scale of diminishing returns to not be worth it?
Diamond is used in some of the higher end thermal pastes and they do perform marginally better than alternatives, but the problem is that they tend to cause micro-abrasions caused by the high-speed vibrations ICs do. This isn't too big of a deal when you're mounting a cooling block to an IHS surface, but if you're doing direct-to-die cooling (which many hardcore enthusiasts do) it can spell disaster.
Well, boron nitride should be almost as hard and abrasive as diamond. I'm guessing it's just the law of diminishing returns that works against diamond here.
As has been the case for a number of years now perf/$ the best choice is Noctua NT-H1. While there are better TIM on the market, they tend to be more exotic which requires special care during application or have specific lifecycles where they must be re-applied. Alternatives which perform better are also more expensive.
I've been building extreme-performance enthusiast systems for more than 20 years which has included using about every sort of exotic TIM you can imagine, including all the various types of liquid metals. Noctua NT-H1 is my continued go-to and the numerous benchmarks of TIMs continues to provide strong evidence that for the money you can't beat it.
Call me old-style, but I still use the traditional white silicone based thermal paste. It might not be the best product where top efficiency is desired, but it won't deteriorate over time: I've removed TO3 transistors from 30+ years old equipment and their white thermal paste was still soft to the touch. Some of the silver/gold ones I tried in the past with CPUS (before going back to the old one) would become solid and deteriorate in a couple years max.
I wish they would have included Desitin (baby rash cream). I've seen it recommended to use in a pinch and wonder how it stacks up. (I've used it on little homebrew radios, and it seems to work, but I've never gotten around to comparing how it works against 'real' compounds.)
Proper application is far more important: No gaps/bubbles, seated with proper pressure, thinnest possible layer.
I've had success in a production environment with the arctic silver 5. This was for a switch fabric IC that can hit 200W, forced-air-cooling through a vapour-chamber heatsink. FWIW other pastes worked fine too, but the factory preferred working with viscous compounds.