BarneyP

I would be fairly amazed if using liquid metal TIM didn't bring down CPU temps quite a lot. I can't understand why one would go to one extreme (putting water inside a PC!) and not go to the other (metal on the heat spreader).

You might try using pure indium as the "solder". It's expensive, but its low melting point (152C) means you can do at a much lower temperature and be sure of not damaging the heat pipes. It also has great thermal conductivity compared to any solder. Those two reasons are exactly why it's used to bond CPU dies to their integrated heat spreaders. It's more difficult to get indium to wet to surfaces than it is solder (most solder is self-fluxing) though you can apply a flux to the surfaces to be bonded first off. You can actually easily melt indium on your stove It's non-poisonous - so just use an aluminium saucepan (though probably not your favourite one!). Not sure of the best way to actually apply the indium to the pipes and heat sinks though.

In the desktop market, Intel should be worried, but their situation is not as bad as AMD's was a few years back. In the server market ... Intel look to be in massive, massive trouble. Epyc Rome puts them to shame to a quite ridiculous extent. In some areas of the product stack AMD are offering 3 and even 4 times instructions per second per dollar. Not only 3-4 times bang for buck. By practically every criterion, Epyc Rome soundly thrashes Intel - much more so than Intel was ever thrashing AMD in the desktop or server market, performance-wise. And the server market is much more lucrative.

The thermal conductivity claims of metal TIM manufacturers are absolute and utter lies. When you alloy two metals, the thermal conductivity is always much lower than the the thermal conductivity of either constituent. It is physically impossible for an alloy of indium and gallium to have a thermal conductivity even close to that of pure gallium, unless the proportion of indium is huge - in which case the alloy would not be a liquid anywhere near room temperature. Addition of silver (or any other high thermal conductivity metal) can only reduce thermal conductivity further. For the addition of silver to increase thermal conductivity, it would need to make up about 50% of the alloy. Such an alloy would have a melting point of around 500 degrees Celsius (search for "Ag Ga phase diagram"). Manufacturers can get away with claiming whatever the hell they like about the physical properties of their products, because nobody has the time, energy, money or motivation to take them to court over it. Eutectic gallium/indium (known as eGaIn) performs identically to Coollaboratory's and Thermal Grizzly's products (I've made it, and I've tested it extensively - it actually marginally outperforms both - presumably because it doesn't contain any tin). Its thermal conductivity is 24.9 W⋅m. K−1. That's as good as it gets.

Gallium is still corrosive when it is part of an alloy. Pure gallium isn't significantly more corrosive than the gallium alloys used in metal TIMs. It won't destroy your IHS though. Gallium alloys with the copper, but the Ga/Cu alloy layer never gets more than a few hundred atoms thick. And it doesn't react with silicon at all.

Yes, I've tried it! It's not hard to do. Obviously the gallium needs to have been melted, but you will also need to heat the CPU first. Not hard, just boot to BIOS with no cooler fitted and it will get hot enough quite quickly. Remove the CPU before applying, of course - you don't want to be fishing around inside the case with a conductive molten metal. The results were dissapointing ... exactly the same temps as a galinstan-based TIM. Not sure why. But as it's about 50 times cheaper, go for it. You also get to play with gallium, which is fun. I wouldn't worry about the freeze-thaw cycle hurting the chip. Gallium will actually remain liquid *way* below its melting/freezing point if it is under pressure, which of course it is when you fit a cooler. When I removed the cooler (at room temperature) mine was liquid and it stayed liquid for quite a while.

Metal polish removes the residue very easily and very thoroughly. I've used "Mother's Mag & Aluminium Polish" (which works on all metals), rub in for about 30s, rub off until cloth shows no black residue, finish with alcohol (again till no staining on the alcohol wipe/cloth). Works brilliantly on the CPU, A bit harder to get the heatsink completely clean, as it's not as smooth as the nickel plating on the CPU. I've found if you apply the polish with fine (600 grit) sandpaper to the heatsink it gets it very clean (and ends up smoother than when you started, which is a bonus).

Ah, that makes a lot of sense, must have used averages. Thanks

Hi Recently looked at some thermal paste comparison articles, and the results gave CPU temperatures to 1 decimal place. I haven't been able to find any monitoring software which shows anything finer than whole degrees. Anyone know of such a program? Thanks

Oh, by the way, what software/site was used to create those cool steampunk logos (I mean the ones used on this page, saying Kingston etc.)?

Probably the greatest mod in the history of modding. The only thing about the finished article which isn't perfect is that the most spectacular stuff is inside, but there are only two small windows in which to view it. I think I would have opted to have no left panel at all (not even a glass one), so that the spectacular "machine" inside is the main feature. Perhaps some (themed, of course) lighting might be cool - radio valves (probably mock - they're available) give a lovely amber glow and are very steampunky. Just amazing, I'm in awe.

Trying to do a bit more research on it myself. Every query I try just gives me loads of tips about using them, but no no "scientific" analysis (fair enough, why would I expect otherwise?). Did find the full product sheet for one manufacturer's inks which gave a breakdown: alcohol (fine, widely known to be safe on electronics); 2-propoxyethanol (a form of anti-freeze, conductivity < 1e-6 mhos, which I think makes it an insulator to all intents and purposes); "polymers" (sic), which I assume are like any other plastics very good insulators, and (1-2%) "pigments" (sic). Obviously the alcohol will evaporate almost instantly. My gut feeling is that there wouldn't be enough of the other ingredients there to worry about blistering or flaking, but only testing could verify that. I imagine the biggest concern is whether the pigments might be conductive. I'm going to be buying some alcohol inks in any case (to see whether I can stain an aluminium heatsink/cooler to look reasonably like weathered copper) and I have an old and useless graphics card on which I can test the direct application of alcohol inks on a PCB. If I get round to trying this, I promise to report back here.

Wow! Much, much more extreme than anything I'm going to be attempting (this is my first "non-vanilla" build). But the patination looks absolutely stunning - I guess because it's real patination. I might well try something similar with my fans. TBH, I'm not sure I'm really going to attempt any painting/inking on the actual mobo - if I did I'd certainly have to test it on some old worthless mobo first - but still curious as to whether it can be done safely. I suppose I could actually test the ink for conductivity with electrodes/ammeter/battery. By the way, what got me interested in the alcohol inks is a few things I saw while browsing for metal finishing techniques e.g. here, here, pretty stunning fake patination/aging I thought. Thanks

Cool, do you have any pictures of coppery builds? Also finding it very hard to find large copper heatsinks/coolers for either CPU or GPU (they're just not popular any more, nearly all discontinued). So many great copper coolers on Google images, but all the links are to end-of-life stuff. Know of any?

Alcohol inks are not paints, they're very strong dyes dissolved in alcohol, and work even on bare metal (in fact they're primarily designed for bare metal, but work on pretty much anything). Most of them are also non-acidic. So, if I wanted to airbrush them onto a motherboard, including electronics, what could go wrong, right? Looking to get an aged copper sort of look for a steampunk mod.

And if cards needed good cooling, then stock coolers be good. And if cards needed high quality thermal paste, then stock thermal paste would be high quality, right? Just about every corner is cut on just about every card.

Afraid not, the backplate holes are in different positions, AM3 and AM4 backplates are entirely non-interchangeable.

I think the correct answer is that having the iGPU in addition to a more powerful discrete GPU probably won't help on most current games (which will ignore the iGPU and just use the discrete card) ... but that is likely to change with the upcoming generation of games using Vulkan and DirectX 12 APIs. Both of these platforms have extensive features to support multiple graphics devices, and these are under the tight control of the developer. This is different from AMD's Crossfire technology, which attempted to spread the load over two cards automatically. With Vulkan and DX12, the division of labour is up to the developer, and it is possible to divide that labour in such ways that the two (or more) devices have no need to to communicate with one another (poor inter-gpu communication was the bottleneck which prevented AMD Crossfire from reaching the full potential of the two cards). With the new APIs, it should even be possible to take advantage of one NVidia discrete card and one discrete AMD card (plus, say, integrated Intel graphics). For example, the weaker card (say, the AMD integrated Vega in your setup) may well be used for things such as physics processing and post-process shading (screen-space shaders such as advanced lighting effects), which frees up more capacity in the main card to do its job of doing all computations for the graphics pipeline.