The ultimate watercooling FAQ for everyone

[Xilikon]

The ultimate watercooling FAQ

You got curious about watercooling or wanted to jump on it but don't know where to start? You are in the right place :)

Basic FAQ:

A: It's a method to extract heat from the computer components and dissipate it in the air with water as a heat carrier. Unlike air cooling, which is basically a base with lots of fins (tied by either heatpipe or directly from the base) and dissipated with the help of a fan or passively, water is a more elaborate closed loop system comprised of those elements: a pump (to circulate the fluid in the loop), a reservoir (to hold the fluid and feed the pump), one or more blocks (to extract heat from computer components and transfer in the fluid) and one or more radiators (to dissipate the heat carried by the fluid in the air either passively or with fans).A: Since water has a much higher specific heat capacity (http://en.wikipedia.org/wiki/Specific_heat_capacity) and thermal conductivity than air, it is more efficient at extracting heat from components. This will also allow you to place the radiator anywhere you want for the most efficient heat dissipation unlike air cooling, where you are limited to the CPU socket area enclosed in a case with potentially hot ambient air inside. With a very good setup, you can have a much more silent setup than air cooling while getting more thermal dissipation. A: When done correctly, watercooling is as safe as air cooling. When done poorly, watercooling is as safe as an underwater LAN party. The main thing to remember about a watercooling system is you have to be CAUTIOUS, you have to follow COMMON SENSE, and you have to TAKE YOUR TIME. When it's all said and done, there will be no real risk to your system. If you rush through setting it up, then you put your system in danger. Remember, most failures are due to human error and not a faulty component. A fan can die just as easily as a pump can die.A: Honestly, it's very hard to build a good setup under 100$ and be as efficient as the best air cooler but when you get past initial setup expenses, it can be very cost effective if you pick the right parts. With the right parts, all you need to change is the water blocks ONLY IF the block manufacturer doesn't provide a new mounting adapter or the block isn't efficient to cool a component. The expense of a block is the same as the price of a good air cooler at this stade.

This is why it's very important to take your time to do a research based on your needs. Don't be afraid to spend a bit more if this expense will warrant you a long term use of the setup (This is true on pumps, reservoirs and radiators, which rarely change even if the blocks get updated) and try to overshoot if you think you will be adding more blocks like video cards and chipsets/mosfet.

After all, if all you want is to cool the CPU and don't want to mess, don't bother with watercooling and stick with air cooling. It's basically a kind of hobby for computer enthusiasts since when you get the WC bug, you will want to tinker with it more ;)

WC loop FAQ:

A: Before talking about the order, the main rule is to ALWAYS put the reservoir before the pump because the current pumps will be damaged if it gets starved for water. Having the reservoir before the pump will ensure there is a constant supply of water and to trap air bubbles (more on it later).

Beside this rule, the order doesn't really matter and it's mostly user preferences. To determine a good order, there is 2 good advices : Try to put the block you wanted to cool the most (like a CPU block) immediately after the radiator so it get the coldest water and make sure your tubing is as short as possible so water doesn't travel too much. A recommended loop order you will often see in lots of watercooling setups is like that: RESERVOIR -> PUMP -> RADIATOR -> CPU BLOCK -> (CHIPSET BLOCK) -> (GPU BLOCK) -> BACK TO RESERVOIR (The items under parenthesis are facultative items).
A: The best fluid we can use in a loop is pure distilled water with nothing else added. Pure distilled/deionized water have the best thermal properties of any fluid while not having anything which can cause issues (precipitates, dye staining, chemical reaction or acidity/alkalinity). For those who wanted to keep distilled water for more than 1 month should consider adding 1 or 2 drops of a biocide (copper sulfate like PT Nuke) to stop the growth of algae. If you need to add color, use a dye sparingly or use colored tubing to avoid using a dye.

There is pre-mixed fluids available in the market so it's tempting to buy them for the "cool" or no hassle factor. They often have the "non-conductive" marking on the bottle but don't get duped by this because no fluid is non-conductive, even plain distilled water (even if a fluid or distilled water is deemed non-conductive, it will become conductive over time due to the migration of metal ions in the water). It's also expensive and sometime poisonous.

About the anticorrosion additives like Pentosin G11, if you setup the loop correctly with the right parts, there is no corrosion risk (as outlined in the next Q/A about corrosion) so you don't need to add them. The only reason is mostly for color and less risk of biological growth if you keep it over a few months without flushing. In case you insist to use some aluminium parts, it's required to add them to mitigate the corrosion effects or you can risk voiding the warranty.
A: Without elaborating too deeply into the process, galvanic corrosion is a electrochemical process where a metal will corrode when put in presence of a dissimilar metal in a electrolyte. Each metal alloy has a electrode potential and a electrolyte is a medium where ions will migrate from anode to cathode, causing the corrosion process. One of the ways to reduce galvanic corrosion the most is to try to mix metals with the closest electrode potential so there is virtually no ion migration. brass and copper are one of the closest because brass is a alloy of copper with zinc but aluminum is very far from copper so it will act as the anode, losing ions. When aluminum acts as the anode, it will dissolve and stick to copper, causing holes in aluminum and those holes will eventually get big enough to cause a leak.

So to answer the question about mixing aluminum and copper, the best one is to avoid aluminum like a plague even if it's anodized or plated. With the current selection of watercooling components in the market, there is no reason to buy components with aluminum in contact with water. People will often say it's safe if you take precautions like using an anticorrosion fluid (a glycol, as in antifreeze fluid) and drain the loop pretty often but why try to use bandaid remedies if you can remove aluminum from your loop?
A: It's indeed a tricky part but let me try to make it simple and use a few rules of thumb for basic selection. We know that the bigger the radiator surface is, the more it will dissipate heat in the air but you are limited by the ambient temp. There is no point trying to buy 10 radiators so if you wanted to cool your system below ambient, you will need to get to exotic cooling systems (chilled water, phase change, immersed in mineral oil or liquid nitrogen). The radiator surface is usually calculated by the amount of 120mm fan area so a 1x120 is the same size as a single fan, 2x120mm is the size of 2 fans together and 3x120 is 3 fans.

The actual heat dissipation capacity will depend on what fan to use so a silent/low-CFM fan will dissipate less than a high-CFM fan. The normal average is usually 50 CFM so with this in mind, a 1x120mm area will dissipate about 150w so a triple can dissipate a total of 450w. With this info in mind, the next thing to do is to calculate the total amount of heat dissipated by all the water blocks and the pump in watts then divide by 150w to get the amount of radiator surface.

If you want to be more precise and is not afraid to calculate the exact amount, there is a good page about that written by Marci (from Thermochill):
http://www.over-clock.co.uk/ivb/index.php?showtopic=20277
A: Short answer : both are equally good since the total radiator surface is the same. Long answer : It depends on a lot of factors including the flow restriction, actual heat dissipation efficacity of a 120.1 vs 120.2 (doubling the surface doesn't always mean doubling the dissipation) and loop size. Ideally, it's best to get a single radiator with the biggest surface you can use depending on your case modding skills, ease of installation and price. If you aren't good at modding and/or your case cannot hold big radiators but have a few 120mm fans holes, you can get away with some smaller radiators but make sure you get those who are not very restrictive, like Thermochill or Swiftech.

Also, if you want to avoid a component heating the loop too much to cause issues to the rest, you can split in 2 loops. In that case, you have no chice to get 2 radiators so that question will become useless.
A: When you receive the components new or used, it's a good idea to give a good cleaning. For the blocks, just disassemble them and clean them with a brush and some ketchup to remove oxidization and stains if you got it used (you can also use vineger if you think ketchup is gross). After the initial cleaning or with new parts, make sure you rinse them thoroughly with distilled water (don't use tap water as this will leave mineral deposits if you area have hard water) until it come clean then it's ready to be used. This is the same with a pump but skip the ketchup and/or vineger as it's not needed and only rinse it with distilled water.

The hardest part to clean in the loop is the radiator because you cannot open it. If you bought it new, it will have a lot of deposits in it, especially flux from the manufacturing process so you need to remove them but even used can have some residu from the cooling fluid or algae. The best way to clean them would be to use boiling distilled water so pour it in the reservoir outlet till it's filled up, leave it for 5-10 mins then remove half of it and shake it (make sure to wear oven mitts since it will be very hot). Do this for 6-10 times till the water come out very clean but it's possible you will need more. A easier way would be to buy a cheap submersible pump, put it in a kettle, connect the radiator to the pump then let it run for 1-2 hours till it come clean (just replace the water once or twice to help).

Some people will tell you to use vineger to clean it but don't do this since it's pointless and this will tarnish copper if you let it sit for a bit.
There are 2 components in the fitting dimensions : The threading size and the tubing inner diameter. The G1/4" is the threading size, which is the same as 1/4" BSPP (British parallel plumbing threading) and it's currently the standard used in almost every watercooling component beside a few oddities (Thermochill radiators use G3/8" as example). The tubing side diameter is the tubing inner diameter size it can accept so logically, for a 1/2" ID tubing, you should be looking at 1/2" barbs. With compression fittings, things get more complicated because it will list the inner diameter and the outer diameter too so with a compression fitting, make sure you get the right one for the correct tubing (like 8mm/10mm compression fitting requiring 8mm/10mm tubing and nothing else).

Don't confuse BSPP or G1/4" with NPT which is often used in North America. If you want more information about threading standards, Google is your friend.
A: It's always a good idea to have clamps on the barbs because there is always a risk of leaking if you don't use them. The only case when you don't need them is to use a undersized tubing on oversized barbs or if the tubing looks to be very snug on the barb but you must leak test to make sure it's ok before running the computer. There are a few options available to clamp and here is a good list : Worm Drive (also called Jubilee Clip), Nylon hose clamps and Zipties in the order from best to average. Zipties is good as long as you install it correctly and tighten them strong enough so the tubing won't pull off the barb but don't rely on them to seal them since there is always a gap around the mechanism. With worm drives, you should try to tighten just enough to make a snug fit but not too much since it can pierce or cut the tubing and/or crush the barb. Nylon hose clamps is a good alternative but make sure you get the right size or it won't work and if you try to tighten too hard, it will break.
A: Compression fittings is a great way to fix the tubing to the blocks without using any form of clamping. It's also very nice to look for those who care about the looks. However, selecting the wrong compression fitting for your tubing can lead to disastrous results. To avoid this, first determine what is the tubing size you plan to use (both the inner and outer diameters) then make sure the compression fitting you select is made for the exact tubing size (if you select a 3/8" ID 1/2" OD, you must use 3/8" ID 1/2" OD compression fittings only). If the outer diameter of the compression fitting is larger than the outer diameter of the tubing, no sealing is possible and it will leak. If it's the other way and it's smaller than the tubing OD, you will have a hard time screwing the ring and risk cutting the tubing.

The most common sizes in the market currently is 3/8" ID 1/2" OD, 3/8" ID 5/8" OD, 1/2" ID 5/8" OD and 1/2" ID 3/4" OD so you are limited to those sized when selecting tubing.
A: Each one has its advantages: A reservoir will help a lot with bleeding (remove air bubbles from the loop, which can prevent heat transfer) by trapping air bubbles better but it's a bit more cumbersome to install and you need a place for it. A T-line (a vertical line with a T coupler placed in a loop) is inexpensive and less cumbersome to install but bleeding will take much longer to do as it doesn't trap air that well. A T-line will also serve as a drain line if it is placed in the lowest part of the loop and long enough to be able to be moved out of case. In my opinion, if you can afford the expense and the room to have a reservoir, it's strongly suggested to get one and with the current choices, you have no real reason to avoid them.
A: Again, it's a matter of personal preferences and budget. A full cover block has the advantage to cover everything on a card including the core, memory, mosfets and the IO chip and get it cooled more than with air but it's a bit expensive to own and very often, it works only with a specific model of card. You cannot get a full cover block designated for card X on card Y as the mounting holes is different and the components might not be positioned the same for good contact. On the other hand, a GPU core only block will only cool the core so you are required to find other ways to cool the rest of the card with ram sinks, mosfet sinks and IO sink (screwed or glued with thermal pad/epoxy) and have some airflow over the surface to dissipate heat. However, since it's only for the core, you can reuse it in the future as long as the block manufacturer release the mounting plate or the holes fit. It's also less expensive.
Finally, a full cover block will let you avoid dumping the hot air in the case so it will help lower the case temp.
A: The main goal of a backplate is to avoid bending the board and ensure the mounting is as straight as possible. With a CPU block, having a backplate is a good idea since the mounting pressure is often strong enough to bend the motherboard. Bending the motherboard will possibly damage it or causing the mosfet (the chips around the CPU socket) to lose good contact with the heatsink and overheat. For the GPU, it's not necessary and usually, there is already one installed on the card or the full cover act itself to straighten the card. As for the chipset and mosfets, it depends on how it's mounted, how strong the mounting is and how easy it is to bend so you can fabricate some yourself or order from a place who have them.
A: It's a bad idea since a refrigerator is built to cool food and keep it cool and not to continously dissipate the heat generated by a radiator. If you do this, you may see nice results at first but after a short while, the compressor of the refrigerator will fail. Beside this, if the water in the radiator is cooled below the ambient temperatures, you will have the risk of creating condensation and risk damaging the computer components without proper precautions (If you still insist on doing this, hop in the chilled water forum and ask since it's out of our scope). You can also risk freezing water and cause damage to the radiator and other components.

Specific parts FAQ :

A: Good question, the Thermochill radiators use a G3/8" (BSPP 3/8") threading and the current barbs threading is G1/4". There is currently a few ways to get around this :

1- If you are in the US, you can order a pair of G3/8" to G1/4" adapters from http://www.mcmaster.com. Look for part 4860K657. With the adapter, you can use any G1/4" barbs or compression fittings you want. this is in my opinion the best choice, giving you the most freedom.
2- If you are in Australia, GAM sell G3/8" 1/2" barbs : http://www.gammods.com.au/store/index.php?main_page=product_info&cPath=1_91&products_id=4
3- Lately, a few manufacturers like EK and BitsPower started manufacturing G3/8" barbs and adapters so you have more choices.
A: First of all, if you have the nozzles kit, install the washer since this will stop the leakage of water flow in the middle chamber and reduce the flow on the base plate. This alone will often give you some improvement in temperature. As for which nozzle to get, it depends on each setup but there is a good guideline agreed by the experts : For dual-cores, if you have a very strong pump like a Iwaki and CPU-only loop, the 3.5mm or 4.5mm nozzle is the best way to go. For a normal loop with a DDC or D5 and possibly more than the CPU in a loop, a 5.5mm nozzle is a better choice but if it is very restrictive due to the rest, get the largest (6.5mm). However, if you own a quad core, it's almost a given to get a quad nozzle (the one with a split in the middle) for the best performance but make sure the orientation it correct (the cores is across the IHS with the markings on the bottom when you look toward the cpu).

However, if you have the Fuzion CPU v2 (released to market recently), they don't use nozzles but they have a quad-core mid-plate available for those running a Intel quad-core. This new version is slighty more restrictive but that's more due to a design fix of the v1 mid-chamber leakage issue. The performance is slighty better with the design fix as well.
A: Before answering this question, D-Tek made 2 blocks called the same, a Fuzion CPU block and a Fuzion GPU block. When we look at the CPU version, it's the most free-flowing block we can get and it get restrictive only if you get a nozzle in it (it get more restrictive with smaller nozzles as outlined above). With a quad nozzle, it's still very free flowing. The GPU version (the one you put on video cards) is the most restrictive block in the market and that's why lots of people get confused about the restriction level. That's also why lots of people will recommend getting a dedicated loop just for the Fuzion GPU or get very strong pumps like Iwakis.

Now, it's becoming a moot point because D-Tek released a new v2 of the GPU block, a lot less restrictive than v1 and working great.
A: If we compare to the curren crowd of waterblocks, it's indeed much more restrictive so it's better to use it alone in his own loop or to use either 2 pumps in series or a Iwaki pump. However, with the restriction also come good performance on large dies. In the end, it's a good pick with the Fuzion but the final decision is determined by the amount of restriction you can live with and the number of blocks in a loop.

[Xilikon]

Kayin's brain busting scientific facts about watercooling

If you are scrolling down to this post, we think you got more and more curious about the science behind watercooling and learn why experts give certain answers to many popular questions (and why certain will try hard to steer people against certain uses).

Many people equate the word "acrylic" to the words "forced anal rape." That is an incorrect statement. However, it is a material that must be treated with respect. Acrylic is not brass, delrin, or copper. It is less chemically inert, less malleable, less tough. However, the trade that you get is the ability to check on any component that's installed under, around or inside it. In some cases, that's a Godsend. I cite the case of the EK Supreme. I (and many others) do not recommend the acetal version for the simple fact that the design, with the thin accelerator plate and microchannels, can easily become clogged if you have anything free in your system (anything from teflon tape to algae to broken loop pieces-I've even seen the inner lining of a tube come loose and travel through a system) and that clogging not only destroys flow but can cause catastrophic failure, especially in a high-pressure loop. The acrylic version allows you to watch for restriction, heading off at the pass any problems that might come from anything growing or leaching into your loop.

Acrylic is chemically less inert, being acted upon by chloroform as a welding agent, cyanoacrylate as a molecular destabilization agent, and acetone as well. Alcohol, however, causes spiderwebbing (cracking) on contact. My only theory on that comes from the rapid evaporation of alcohol, I do not at the moment confess to know the real reason, only that it does occur. There are alcohol-resistant strains (often those are polycarbonate, though) but they are almost never encountered in the grade we see in watercooling. Another benefit is it may be any color, as well as color-change and UV reactive.

The other leading all-around plastic is polyoxymethalone, sold under the trade names Delrin or Acetal. It is normally black (there are white versions though), hard, tough, easy to machine and somewhat indestructible. It is classified as a "synthetic stone" on many plastics websites. It is many people's plastic of choice, especially as it is more durable than acrylic. The trade off is, of course, that it is opaque. Six of one, a half-dozen of the other. It is far more durable, being mostly chemically inert, though there are reports of shattering Delrin (Enzotech Sapphire, first revision) and one instance I have seen of melted Delrin (MCW60, pump failed.) It is generally regarded as the safest plastic in watercooling.

Continuing on, there are more exotic choices available, and some find their way into loops.

Next most common is actually practically universal. We all know what it is, most of us use it-polyvinyl chloride. We know it in loops as Tygon, Durelene, Masterkleer-most mainstream tubing is made of it. It is (or can be) mostly chemically inert, food/medical safe, soft, flexible, and clear. Certain grades of Tygon are not-Tygon Antimicrobial is impregnated with silver, Tygon R-3400 is black, and another Tygon, Tygon Norprene, isn't even PVC but another plastic altogether. By varying the amount of plasticizer, you can get other grades-the standard for WC R-3403, the extremely soft R-1000, and the plasticizer free (B-44-3, 2275 *the latter being medical grade.*)

Another overlooked use for polyvinyl is reservoirs. A few have used the material, which without plasticizers and using a different catalyst hardens to very slightly flexible, water-clear and ver, very tough-sometimes able to survive trips off my roof, or being hit with a hammer. It machines very easily, leaving long curls when drilling and a dangerous white dust when Dremeling or machining. The fumes associated with doing so are also not healthy, and if you make/modify a part from this, it is best done outside with a mask.

Also on the docket is polycarbonate (trade name Lexan), a hard clear, and very tough plastic, but one which does not machine as easily and taps with difficulty. This is the same material as bulletproof glass, with a high tensile rating and impact durability. It is also immune to acrylic's alcohol vulnerability. Polycarbonate is most likely best used as a cast item, where any machining was done into the mold itself.

Last but not least are the syntetic rubber substitutes (and real rubber) used in O-rings. These of course stop leaks, but each has its application too. Buna-N is rated for water only apps (straight distilled) but may degrade over time in antifreeze/coolant type loops. EPDM is a synthetic rubber approved for most anything. And silicone O-rings can be used in cases of poor mating fit, as it is the most malleable/flexible of the three offered. Silicone is also very resistant, and does not degrade over time, even in straight isopropyl.
When it comes to additives-if you can't pronounce its scientific name correctly, it probably doesn't belong in your loop.

The most common (and most harmful) is ethylene glycol. We know it as green antifreeze. Sweet, acrid-smelling, and incredibly deadly to pets and children, it shuts down the kidneys. Avoid if at all possible. I'm personally pushing to see it out of the industry in 2009.

Propylene glycol is its replacement. It's not sweet, and it's judged food additive safe while still being an anticorrosive. It's often blue. However, just cause it's blue doesn't mean it's propylene-read the label!!!

We then come to the next line of additives, the Pentosin/Glysantin family. There are the G1x series of Pentosins, an antifreeze marketed for higher-end vehicles containing a lot of aluminum, and G48 Glysantin, a super-duty antifreeze that sees use in long-life colants, tractor engines, and snowmobiles. The G1x series are denoted by their color (pink, purple, red, yellow, and blue) and G48 Glysantin is a deep blue. More info than I can even begin to work with is available from their manufacturer, BASF.

Last, and certainly least common, is GM Dexcool. Another fully synthetic coolant/anticorrosive, it is orange in color. It is designed for long times between service, and is very stable, though not that great a PC coolant.

Water Wetter-don't. It coats parts, reducing thermal transfer greatly.

Hydrx is simply Valvoline Zerex for marine use with green dye. Usage in the SWFT recommended level of 5% is way too much. Thanks Ricey.
Surfactants seek to reduce the surface tension of water, making its boundary layer thinner and thereby making it more efficient at transferring heat. The only general use surfactant is glyerine (not glycol, which is an alcohol-based substance), which is basically soap. It can cause foaming (not common, but seen it happen) but does not normally do so. It is purported to lubricate pump bearings, but only on older mag-style pumps is this the case. On newer pumps, they are designed for the loads that we impress upon them, and they need no extra lubrication. However, the only claim that is confirmed is that glycerine is a true surfactant. It use in loops, however, is debatable.

Biocides come in multiple types. Some simply destroy all life they touch, while others are targeted at algaes. The broad-spectrum biocides are almost all dangerous to humans. They begin with the copper-sulfate based PT-Nuke. It employs a higher concentration of copper ions normally found in copper-bearing watercooling loops to kill most anything living in the loop. It does alter the pH of the water, though, making it more acidic. It is not safe for mixed-metal loops, and will degrade the stability of straight distilled (to what point it is not known yet, only the fact that it lowers pH measurably is known.)

Next is PT-Nuke PHN. It is composed of benzalkonium chloride, which is pH stable, as well as being designed for mixed-metal loops to work in concert with anticorrosives. However, its buffered pH means that it is perfectly suitable for distilled only loops.

Both of these are not safe for internal consumption. Please do not lick, snort, or doctor your drink with them.

Also in this category are targeted biocides, otherwise known as antialgal compounds. These are basically aquarium anti-algae medications, and as such are generally safe for human consumption, though I do not recommend it anyway. God gave us Kool-Aid, don't drink the biocide.
The clear ones are a chemical preparation that is a targeted algae-destroyer, and the dark colored ones are potassium permanganate, which will stain tubing and any porous plastics brown. Both are fish safe, though. Your mileage may vary, as they say, but potassium permanganate WILL wipe out any algae you have. Not sure about the cooling capacity of the water afterwards.

The other category is dyes, which pretty much polarize the community. Yes, they are a thermal compromise, but many people have watercooling as much for looks as for performance, so they proliferate.

Of dyes, you have the UV reactive and the non-UV reacive. If you have no UV lighting, there's no difference. However, most dyes are made up of a suspension of particles in water or another carrier liquid (one reason why dyes separate over time is the changing of distilled water's ionic proerties, as they are more properly classified as an emulsion and not a true suspension. The change breaks the bonds of the emulsion, and the forced mixture separates.) There are some dyes that are truly liquid, though-the aniline series, a poisonous oil-based (and oil-dyeing) set not too often seen any more, but VERY potent, a drop can color a loop oftentimes. Also, many people have drained highliters for the ink inside. As it is a true liquid, it does not separate easily, but it is very staining as well. It's what we used to do before good dyes originated, if we wanted glowy water.

Any pre-prepared off-the-shelf liquid will be any number of these things. Readthe label carefully, and if necessary, request an MSDS (Material Safety Data Sheet.) We at RRR.com do NOT endorse the use of ethylene glycol in any coolant, and it is my personal quest to convince vendors to drop it altogether in 2009. Please join us in saying that ethylene glycol is simply no longer acceptable-for our pets, our children and the good of the industry.
Silver has been making quite a showing as of late, in coils, blocks, and tubing, among other things that we don't supply ourselves (coins, ingot, etc.) There are two major benefits to silver, each of which I'll cover, as well as pros and cons.

The first benefit of silver is its thermal conductivity. At 429 W/mK, it is slightly more conductive than copper, at 401 W/mK (both are assumed at 300 K.) Therefore, it is to be assumed (and is borne out by testing) that silver is the better material for waterblocks. It is, however, softer and far more expensive-repeated high-pressure mounts will leave the impression of the core or IHS in the block, and oly lapping will remove it. This can be easily seen as a destructive cycle on silver based-blocks, so it is something to watch out for.

The second, and of far more use to the watercooling world, is as a wide-spectrum biocide. It is simply disruptive to life on a microscale. It works on the same principle as heavy metals toxicity in humans, without harming us in the process (however, too much silver exposure can cause argyria, which leaves you looking like a Smurf. You were warned.) It works on bacteria as well as algae, so in theory nothing should grow in the loop, however, there ARE silver-resistant strains out there, be they incredibly rare (stuff grows in Chernobyl too, so never discount the tenacity of life.) be that as it may, they are rare enough that they should never pop up in your loop. In this instance, a small amount of silver will do, as deionized water seeks to gain ions as fast as possible, and if silver is present, it will readily react, distributing silver ions throughout the loop, maximizing the efficiency of a small item.

It goes without saying that a silver block will perform the same function. Also to not is that silver is close enough in nobility to copper to corrode aluminum as well, so if you mix them, you do so at your own risk. Yes, silver killcoils or coins are enough to cause a reaction.
Don't. If that's not good enough, I will enumerate the answers.

Vinegar is acetic acid. The solution is an imprecise one, which is problem number one. You don't know what's what with it, unless you wanna litmus test all your vinegar. Also, it's an acid. Acids eat stuff, unscientifically put. Inside your rad, there may be leftover solder, as well as if it's older, lime scaling as well. If you have a HWLabs or Swiftech rad, all their solder is and has been done with water-soluble flux. If it is a ThermoChill, there is a chance that it is a non-water soluble flux. This is where the concept of flushing with vinegar originated. It would dissolve the flux, as well as any scaling in the rad as well. However, acetic acid will attack both the metal and the solder, doing damage on both counts. If not flushed immediately, the copper starts to ionize at a very rapid rate, filling the tubes with a solution of copper carbonate (there is carbon in the vinegar) so that when it dries will coat the tubes in azurite/malachite microcrystals, destroying much of the effectiveness of the radiator. It has an affinity for the walls of the rad, so it does not easily come out when flushed. It also weakens the solder bonds, and can cause a rad to burst if left in overlong.

Since the problem has to do with a non-water soluble material, a solution specifically designed for flushing radiators is recommended. It will remove scaling as well as any solder flux, without damaging the metals. Just so you know, CLR is a proprietary phosphate-free acid and water mixture with surfactants, so I'm not too sure about recommending it either. Radiator flush is designed with the copper and brass tubes in mind, CLR is not always.
Before I start, I want to say this. If you don't read the entire article all the way through, and pay attention, you WILL NOT understand what the point of this article is. You are forewarned.

In the PC cooling world, there are many compromises, and the one that most people think of is coolant versus distilled, with distilled being the normal winner. However, there are things about distilled that people should know, and how to exploit those weaknesses for our own profit.

Distilled water is not neutral pH. Distilled tends toward acidic, and pH can go as far down as 2-3. This is bad for many parts, though often not over the timescale that we see. What it does cause, however, is mass ionization of any and all parts in contact with it.

What we often call distilled is in fact distilled, but also run through reverse osmosis and other filtration to produce a water without ions-which are the sole reason that water conducts electricity. Deionized water is dielectric, which in layman's terms means that it doesn't conduct electricity. However, water, as with any other substance, it aggresively forms ions with the free electrons inherent in the water. The process is self-stabilizing, it will not continue to happen as rapidly as the electrons form pair bonds and the shells are filled.

However, consider this. Copper forms ions incredibly rapidly, and if you have aluminum in the loop, the copper will react with the aluminum, giving perfect conditions for galvanic corrosion. Brass will do the same thing. Nickel and other metals that are stable in their outer shells (less reactive or nonreactive, as they have relatively full outer electron shells) do not do this as readily. As it picks up ions, its pH rises, but even so, the initial pH and the subsequent seeding of the water with copper is enough to do major damage with mixed metals.

Since there is this weaknes inherent n deionized or distilled water, what to do? The simple answer is to exploit this by using something that we want its ions in the loop-silver. Silver is not as free as copper in ionization, being a more noble metal, but it will give them up, and when the water is laden with silver ions, it becomes biocidal, spelling death to all manner of things that find their way into loops.

Water is truly not the "ideal" coolant, as the pH will not be constant, it still picks up ions for long periods after filling, and once it has done so it is not dielectric. Even with the ability to use it ionic properties in our favor, it is still an imperfect solution. The biggest detractor is of course the fact that water does not retain its dielectric properties. However, adding things to water to make it retain that property reduce its most stellar one-its thermal conductivity. So, it is unfortunately, the perfect compromise-it's still conductive, but at least it conducts heat well too. Hopefully science will give us a new toy to play with in dealing with this issue, but for now we're stuck.
This section is dependent upon the principles set down in the distilled water discussion. If you have not read it yet, please do so.

First off, in dealing with corrosion, most of you think I'm going to immediately jump on aluminum. Not so. I'm going to set out the principles of the galvanic table, metal nobility, and corrosion in general. By then, it should be evident.

First off, I talked about ions last time. Ion are a big part of corrosion, as is a process called oxidization, where oxygen (which is always ready to combine) chemically bonds with the subject at hand. Some objects ionize, some oxidize, some do both, and a few do neither. That information is important given the nature of what we're working with.

First up is metal nobility. A metal is considered more "noble" if it has fewer electron spaces in its outer shell empty-if its shells are full or close to. The less empty spaces, the more noble, all the way to full and almost completely nonreactive. A metal is classified as "active" if it has more open spaces, or fewer completed shells, and will readily seek to either steal electrons (ionize) or chemically bond (often oxidization, alloying is another process) with the materials around it. In a loop, copper is a "noble" metal, aluminum is an "active" metal. The resultant giving up of ions creates an electrical current, and that hastens the ionization process of the aluminum, destroying the structural integrity. The reason that glycols and other anticorrosives work is that they inhibit the formation of the ions (thereby also inhibiting the formation of silver ions for biocidal purposes) and disallow the development of the charge. A copper/aluminum loop is a simple type of battery.

However, in a copper-only loop where there is a problem with varying pH or concentration of a solution (as in some commercially prepared fluids, and a few dyes, mostly no longer sold) you can still create the same conditions and corrode the metal in a lower concentration. This is a "concentration cell" battery, effectively, and the copper can become permanently stained, or even pitted this way. Another hazard of crappy fluids.

Corrosion does not only happen in metals though. Plastics we would consider to corrode by plasticizer leech, that nasty white scum in tubings. It is brought on by many things, such as UV light, free radicals and the presence of oxygen (water) or chlorine (the old tendency to drop a few drops of bleach in for UV and bug killing) and results in brittle tubing or barbs. The Tygon that we use is often subject to it, and new tubing formulations have been tried in the PC watercooling industry to attempt to combat the issue. One formulation, R-3400, simply has carbon black added to combat the UV issue, and has performed satisfactorily. If you have ever had nylon/polypropylene barbs break, you've experienced plastics corrosion.

While it IS possible to stop galvanic corrosion entirely, by the use of sacrificial anodes or the introduction of a current designed to stop the reaction, these approachs are much more complicated, and so normally stopgap solutions are used. These include using various fluids designed to stop corrosion, or simply draining and refilling often. The fluids employed are often not nearly as good thermally as water, and the concentrations necessary to halt corrosion (10% and up) are detrimental to temps. Simple changes of water are not enough to halt it, merely slow it down. Also, most of the fluids are designed with a higher working temp in mind, which can destroy their effectiveness as well.

Another method that will stop galvanic corrosion is to plate the more noble metal in something that is nonreactive. Nickel, chromium, gold, silver and platinum are all viable choices, and as they will not readily ionize, they will not allow the reaction to start. Gold, for example, simply does not corrode by any normal means, and is a perfect choice for plating of components. In a gold-plate and aluminum loop, it would be safe to leave out any additives and run straight distilled, assuming the plating job was good enough. However, any item that is not as rigorously plated tips the balance and starts the process.

A final note on plating. The Swiftech Apogee GTX originally shipped with a plated aluminum top. The top was first plated with zinc, then with nickel. a few words on this. Aluminum and zinc are galvanically dissimilar, but in most cases the aluminum is protected from water, so it is a nonissue. However,the plating was not as good as advertised, and the zinc, instead of being the passivator it was designed to be, was quickly consumed as a sacrificial anode and then the aluminum began to give way. The zinc, being the most active metal, was also simply plated too thin. After the nickel was gone (or where it ddn't plate properly, as some GTXes were shown as having) the zinc wore itself out rather quickly. This was the reason for the rash of GTX top deaths, the consumer outrage and the final solution of the all-copper top. This could have been avoided by plating the copper instead (of course, it would have happened down the line) or by never using the aluminum in such close contact in the first place.
The rage right now it toward "bowed blocks", blocks for which the center of the base is proud of the rim. The issues with bowed bases are numerous, but let's start with the original issue at hand.

The Intel IHS itself is a truly craptastic piece of engineering. Many are concave, some are convex, and on my own Q6600 it was both and then some. I had multiple dips and one high spot, necessitating many hours of lapping to flatten and then polish the IHS. Bowed blocks were developed in response to the concave (usually) Intel IHS, as a way to quickly increase areal contact. The bow often fits into the concavity, helping to reduce the air gap and therefore the amount of the TIM joint. While this sounds good, there are some issues to look at.

1. The bow is a one-way street. If you afterwards need that block on a flat processor, you are SOL. If you need a flat block in the future, hit it up. This limits blocks to one-hit wonders, or at the most to a product line.

2. Increased mounting pressure. Yes, we should all use backplates to help evenly distribute force, but the issues with bowing mean pressures may go above 150 psi, which can cause long-term damage to boards, and if it's uneven can even destroy a board. Even with backplates some boards have been observed flexing. That flex means that board, chip and block are the only thing that will work together in the future, and some boards have popped after being changed out due to stresses from deformation and the attempt to correct it.

3. Ill fit even at best. Unless you can scan your IHS, 3D print a reverse copy, and then mill/lap your block base to match, you will have a less than optimal joint. There will be gaps that must be filled by TIM, which has a much lower thermal coefficient and therefore will lower the efficiency of the entire system. Also, additional force is required to mount a bowed block properly in order to try and minimize the gaps (see answer 2 above.)

4. Stress deformation of other substances. It is entirely possible to damage a block that has been intentionally bowed because not only are you mounting it harder, you're also exerting stresses on other parts of the block, ones not designed with that in mind at all. Delrin/acrylic are not exactly elastic, and elastic materials aren't very common in waterblocks.

5. Propagation of disinformation. By allowing bowed blocks to proliferate, we cloud the truth and reward design flaws. If we insist on flat blocks, then we also insist on research, proper machining and most importantly reusability.

As far as the anatomy of the TIM joint, as well as the block/IHS interface, that will be the subject of my next article.
In the modern watercooling loop, there are two main places for radical gains to be made: the IHS/block interface and the radiator. Radiator gains can simply be expressed as increases in areal density: simple but very effective. However, the other area in which large gains can be made is much more complicated: the IHS/block interface, and the TIM joint.

These both bear at least a cursory explanation before I go on. Most people have never heard of the first, or think it is indistinguishable form the second.

The IHS/block interface is the place where the rubber hits the road, so to speak. It is also called the contact patch, though that is rightly a term given to where the water impacts the block as well (one term for two concepts, which is why we choose this terminology.) If the IHS is deformed, there will be gaps in the interface, severely impeding cooling performance. If the block is bowed, it may help make up for that. If the block is bowed, but the IHS is flat, it will rock, or in a worst case scenario damage the IHS and the block base from mounting pressure. The best joint is created when both block and IHS are perfectly flat. This ensures even contact with each other and maximizes the area in which the block has to work.

A bowed block necessarily has less contact than a completely flat one, and while point cooling has its gains, the general issue is that a bowed block is less efficient than a flat one on the basis of simple areal density. Also, even if using a bowed block on a bad IHS (which WILL produce gains over a flat block on a deformed IHS, no doubt-for the same reason as aforementioned) the area of contact is still less than flat on flat. Once again, much of the gains there are from areal density.

The TIM joint is the amount of TIM (thermal interface material) that is used to join the two surfaces together. Under normal circumstances, there are microfine grooves in both block and IHS from machining, and it serves to fill those air gaps (air is a dead insulator-such a low specific heat it's scary in this circumstance) which serve to insulate the block from the IHS and severely impact cooling capacity.

However, the problem is that all TIMs are significantly less thermally conductive than copper, or even aluminum. Therefore, minimizing the TIM while maximizing contact is the name of the game. In a bowed block/bad IHS scenario, TIM serves to complete the joint in some places, making thick layers of poor transfer and thin layers with a higher transfer coefficient. A bowed block on a flat IHS and a flat block on a cupped IHS do the same thing-huge areas of thick TIM, and poor contact. On a flat block and flat IHS, though, the TIM is there only to push out any air that may be in any machining imperfections, and its overall effect is much more negligible. Of course, increases in the thermal conductivity of the TIM will result in gains either way, but with the more inefficient joints it is more readily seen.

The proposed solution to all this is lapping block and processor. The IHS, once properly lapped, is perfectly flat-as will be the block. This allows for a much thinner TIM joint, less wasted TIM, better areal contact, and less mounting pressure. I'm going to introduce a machining concept here-the "wringing fit." Answers.com defines it as "(design engineering) A fit of zero-to-negative allowance." Imagine the concept of a zero-tolerance fit. Both surfaces are completely flat. This theory is espoused by the concept of gage blocks, blocks used to zero dial gauges, calipers and micrometers. They come in multiple sizes, and they fit together simply by hand pressure-the fit is so tight it pushes out the air and they stick together by force of attraction. With a fit that tight, the TIM joint is practically nil, and there is no better interface short of water directly upon the IHS. While difficult to create with lapping, I'm convinced it can be possible, and my initial experiments have yielded a fit so tight that the block can pick up the processor if it is pressed down upon it so as to force the air out. Not wringing, granted, but close-possibly close enough to shave 2-3 degrees off working temps, idle and load.
We told you so :ahha: