![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Willing to Dish Out for Best Cooling
- Thread starter Jacez44
- Start date
crownonfire
n00b
- Joined
- Nov 20, 2009
- Messages
- 46
Talk to ultrasonic at xtremesystems for a TEC chiller
You mean like the Freezone Elite.. which is about as good as my TRUE and costs 350$? I'm a bit lost in regards to >Air coolers.
SicKlown42012
2[H]4U
- Joined
- Jul 6, 2008
- Messages
- 3,319
A custom made TEC cooler is miles ahead of any kind of air cooling. The pre-built ones you can buy don't represent what TEC coolers are capable of.
TEC water-chilling is wretchedly inefficient unless you run several TECs in parallel at low voltage (5-8v or so). At that point, the Delta Temperature achievable is reduced to about 10-20C below ambient, so that'll be the temperature of the water.
I've considered doing this... because you can control the voltage and/or simply turn off most of the TECs when you're not under load, and the power consumption would then be reasonable, but the cost of several TEC's, a good TEC PSU, and materials and manufacturing of the necessary water-blocks would be high.
At that price point, I could actually go with a DIY phase change setup, which will achieve lower temperatures, potentially cost less, and run more efficiently. The reason I don't get back into phase-change cooling: Noise (I'd have to place the setup outside and run it through the wall). If I went this route, I'd chill the water so I could inexpensively chill all PC components instead of just the CPU. I might also run a small evaporator directly in the CPU/Video case compartment (hopefully sealed off from the hard drives) and air-chill the PCBs. It wouldn't create condensation if done right, and it could actually reduce condensation on the water-chilled components (because the case's air would then be at a similar temperature). That's rather involved, however.
In the end, we have these best-case scenarios (and phase change, which can be subzero, but I wouldn't recommend it for the uninitiated):
- High-end Aircooling (noctua NH-D14 or Cogage Arrow), - Let's assume 75C under load
- High-end watercooling (Multi-Radiator, top tier waterblock, possibly dual pumps) - About 15C better under load at the same noise level
OR... if you're willing to do some machining on your own:
- A carefully designed and regulated TEC-waterchiller. - An additional 10-15C better, and depending on how much power you're willing to consume, you could potentially go lower... fortunately if you employ voltage control and/or switch on/off some of TECs, you can control the power consumption quite well.
- All of which would do best with a high-end thermal interface (coollaboratory liquid metal or Indigo Extreme, sufficient mounting pressure, and perhaps a good lapping job).
- There is also the option of running an open-loop evaporative water cooler instead of a radiator. This will only work if you're in a low humidity environment, and it'd be better if you're able to vent the created humidity, but these coolers can achieve water temps about 5-7C below ambient (In some cases I've seen 10-11C below ambient).
The downsides are: They create excess humidity in the room, they're open loop (refill more often, can't have harmful additives in the water, but silver might still handle most of the job, more maintenance), and the noise of dripping water (which can be minimized based on design). I personally think they're a PITA, and this is why so few people still run them.
I've considered doing this... because you can control the voltage and/or simply turn off most of the TECs when you're not under load, and the power consumption would then be reasonable, but the cost of several TEC's, a good TEC PSU, and materials and manufacturing of the necessary water-blocks would be high.
At that price point, I could actually go with a DIY phase change setup, which will achieve lower temperatures, potentially cost less, and run more efficiently. The reason I don't get back into phase-change cooling: Noise (I'd have to place the setup outside and run it through the wall). If I went this route, I'd chill the water so I could inexpensively chill all PC components instead of just the CPU. I might also run a small evaporator directly in the CPU/Video case compartment (hopefully sealed off from the hard drives) and air-chill the PCBs. It wouldn't create condensation if done right, and it could actually reduce condensation on the water-chilled components (because the case's air would then be at a similar temperature). That's rather involved, however.
In the end, we have these best-case scenarios (and phase change, which can be subzero, but I wouldn't recommend it for the uninitiated):
- High-end Aircooling (noctua NH-D14 or Cogage Arrow), - Let's assume 75C under load
- High-end watercooling (Multi-Radiator, top tier waterblock, possibly dual pumps) - About 15C better under load at the same noise level
OR... if you're willing to do some machining on your own:
- A carefully designed and regulated TEC-waterchiller. - An additional 10-15C better, and depending on how much power you're willing to consume, you could potentially go lower... fortunately if you employ voltage control and/or switch on/off some of TECs, you can control the power consumption quite well.
- All of which would do best with a high-end thermal interface (coollaboratory liquid metal or Indigo Extreme, sufficient mounting pressure, and perhaps a good lapping job).
- There is also the option of running an open-loop evaporative water cooler instead of a radiator. This will only work if you're in a low humidity environment, and it'd be better if you're able to vent the created humidity, but these coolers can achieve water temps about 5-7C below ambient (In some cases I've seen 10-11C below ambient).
The downsides are: They create excess humidity in the room, they're open loop (refill more often, can't have harmful additives in the water, but silver might still handle most of the job, more maintenance), and the noise of dripping water (which can be minimized based on design). I personally think they're a PITA, and this is why so few people still run them.
BillParrish
Supreme [H]ardness
- Joined
- Aug 25, 2006
- Messages
- 7,519
Put in a swimming pool and plumb it to the computers water cooling loop.
CPU water block and hold down
GPU water block(s) and hold down
Y and sufficient pipe/hose and adapters to tap into swimming pool circulation pump outlet
Y and sufficient pipe/hose and adapters to tap into swimming pool circulation pump inlet
pump (size dependent of lenght of water line runs and diameter of pipe used. 1/2 horse would be more than sufficient for most needs.
hose clamps
Swimming Pool.
Open invite for all local sororities for free usage of pool facilities.
Poolside cabana bar and barbecue optional.
CPU water block and hold down
GPU water block(s) and hold down
Y and sufficient pipe/hose and adapters to tap into swimming pool circulation pump outlet
Y and sufficient pipe/hose and adapters to tap into swimming pool circulation pump inlet
pump (size dependent of lenght of water line runs and diameter of pipe used. 1/2 horse would be more than sufficient for most needs.
hose clamps
Swimming Pool.
Open invite for all local sororities for free usage of pool facilities.
Poolside cabana bar and barbecue optional.
Last edited:
Where the heck am I going to find a horse??
Well, the FreeZone Elite has 12 TECS (allegedly), and it's about as good as air cooling.
Also, someone is saying it's a "great deal" for 350$, which is driving me nuts!
Overall, I have no idea what voltage has to do with cooling (no idea about TECs).
All for 10-20C less than air? That's beyond insane.
Okay, there we go. That's what I wanted to hear. Phase change is the only thing that really seems to be worth anything. I mean, it gets your temps BELOW 0! In which case, why do people even need LN2? The TDP is low enough to reach the CPU's maximum frequency.
Unfortunately, the prices seem to start at 900$.. which I might just have to pass on.
Also, this is going to be inside a quiet room, so the noise aspect seems crazy. I COULD put it outside the room (on the balcony), which would be easy enough, but I'm not sure if the hose would stretch about 5 feet.
This is my issue precisely!
High End Air - 50$ - 75C
High End Water - 300$ - 55-60C
TEC/Chiller - >300$ - 40-45C
Phase Change - 900$ - -52C
I am SUPER willing to work on the system and tweak. That's my passion.. but for that much money.. that low of a result.. it's pathetic.
Basically, I don't need -52C. That's ridiculous. The chip's highest frequency was reached WAY back.
SO! What I want is something that can reach it half way from 52C to -52C. Something that costs 450$ and gets it down to 0C.
Does that exist? Do you get my drift? I really hope so!
Your only option is DIY phase change.
- A lot of people buy an air-conditioner (5k-8K BTU will cost $100-$250) and convert it into a chiller so that half of the work is done already. They need to make a a heat exchanger to chill the water, and there are several methods to do this (pre-made compact designs, DIY piping into a coolant reservoir, in-between, or simply re-using the stock evaporator (which will have aluminum fins in direct contact with the water)).
I think this guy converted an A/C: http://hardforum.com/showthread.php?t=1424727
- Other people scope out a used compressor (maybe they're looking for a certain model. Some are quieter and more efficient than others). Then they size an appropriate condenser... just a complete DIY job, and I've now forgotten all the tricks... Unless you find some ebay or used deals, this will be more expensive and troublesome. You need to know what you're looking for already, but if you figure it out, you'll likely have a better unit overall.
- Last option is for pre-built units... which are prohibitively expensive.
- There are some A/C units that are designed with the evaporator separate from the compressor and condenser assembly for the purpose of running the "hot-end" (also the source of noise) outside. The line running to the evaporator would need to be heavily insulated (so it'd be a few inches in diameter) - but it can run through a wall onto a balcony without issue. Most A/C units are not designed this way, so it'd have to be a modification. You could, alternatively, run the chilled coolant line through the wall.
- At some point you'd need to invest in some equipment to be able to evacuate/collect (often using an air pump) and charge (needs certain valves) the system with refrigerant. If you could find a buddy with a license, you might be able to score a better-performing refrigerant or mixture of refrigerants... otherwise most people will stick with whatever the AC came with, or R134a or propane (R90), which work just fine.
* The TEC voltage issue: When running at full voltage, TECs will consume VAST amounts of power. When running at lower voltages, they can be surprisingly efficient (coefficient of performance can start at less than 0.5 and rise above 2.0), but the max temperature difference between sides is reduced to 10-20C (maybe 20-30C at a COP of 1? I'd have to go check the numbers again), somewhat defeating the purpose. Fortunately they can be run at higher voltages at load, and lower voltages at idle - at the whim of the user. There have been successful chillers taken to about freezing point, but it's difficult... and expensive.
- A lot of people buy an air-conditioner (5k-8K BTU will cost $100-$250) and convert it into a chiller so that half of the work is done already. They need to make a a heat exchanger to chill the water, and there are several methods to do this (pre-made compact designs, DIY piping into a coolant reservoir, in-between, or simply re-using the stock evaporator (which will have aluminum fins in direct contact with the water)).
I think this guy converted an A/C: http://hardforum.com/showthread.php?t=1424727
- Other people scope out a used compressor (maybe they're looking for a certain model. Some are quieter and more efficient than others). Then they size an appropriate condenser... just a complete DIY job, and I've now forgotten all the tricks... Unless you find some ebay or used deals, this will be more expensive and troublesome. You need to know what you're looking for already, but if you figure it out, you'll likely have a better unit overall.
- Last option is for pre-built units... which are prohibitively expensive.
- There are some A/C units that are designed with the evaporator separate from the compressor and condenser assembly for the purpose of running the "hot-end" (also the source of noise) outside. The line running to the evaporator would need to be heavily insulated (so it'd be a few inches in diameter) - but it can run through a wall onto a balcony without issue. Most A/C units are not designed this way, so it'd have to be a modification. You could, alternatively, run the chilled coolant line through the wall.
- At some point you'd need to invest in some equipment to be able to evacuate/collect (often using an air pump) and charge (needs certain valves) the system with refrigerant. If you could find a buddy with a license, you might be able to score a better-performing refrigerant or mixture of refrigerants... otherwise most people will stick with whatever the AC came with, or R134a or propane (R90), which work just fine.
* The TEC voltage issue: When running at full voltage, TECs will consume VAST amounts of power. When running at lower voltages, they can be surprisingly efficient (coefficient of performance can start at less than 0.5 and rise above 2.0), but the max temperature difference between sides is reduced to 10-20C (maybe 20-30C at a COP of 1? I'd have to go check the numbers again), somewhat defeating the purpose. Fortunately they can be run at higher voltages at load, and lower voltages at idle - at the whim of the user. There have been successful chillers taken to about freezing point, but it's difficult... and expensive.
Last edited:
You're suggesting I buy an AIR CONDITIONER to cool my CPU?
That sounds ridiculous to me, on first coherence.
I understand, from my love of physics, that an air conditioner works on the same level as a phase change system, i.e. gas (air) is compressed into liquid, which travels through tubes at high pressure (150-200 PSI), and is vaporized into cold air. But a phase-change system sort of "sucks" the heat from the CPU, through quick changes of pressure, and then repeats the process of turning the heated air into cold air.
I am not sure how an AC could cool a CPU, nor do I understand what liquid has to do with this procedure. Unless referring to the minute amount of liquid that does not become vaporized through the procedure.
I sure don't know what I'm looking for, and I don't need better results. Anything between -52 and 0 will do me just fine.
But can I get the same level of performance with a DIY method, without it being terribly inconvenient (I still have neighbors to think of)?
Yeah, obviously I'm not going the TEC way. It costs nearly as much as pre-build phase change and it's not nearly as efficient.
- An A/C unit IS a phase-change unit, as is a refrigerator, which is really just an AC unit wherein the evaporator is chilling the air of sealed chambers.
- The evaporator is the cold side of the system, and it can look like many things:
- A "radiator" (through which we can blow air as in an A/C), or
- A block of copper on the CPU and/or GPU (direct-phase-change cooling, without chilling water), or
- A bunch of coils of copper tubing/piping in a reservoir of water (to be chilled), or
- A different heat exchanger design, still chilling water.
So there are two basic options:
1. Couple the evaporators directly to the CPU and GPU, and forgo the use of a separate chilled coolant. This allows for lower temperatures, but more difficult assembly, tuning, etc... requires more experience (esp. if you want to cool more than just the CPU).
2. Chill the water, and then you're able to cool everything as you would in a regular watercooled setup (except with added insulation to prevent condensation).
In that last scenario, the evaporator is acting as a water-chiller.
The only way around insulating for condensation is to run dual evaporators, one of which will chill the air inside a sealed PC enclosure (not the hard drives though). This air must as cold or colder than the coolant. This adds complexity to the design, however, but it has been done.
Obviously the most practical solution is to adapt an A/C unit into a water-chiller, and insulate heavily.
Edit: and it won't be horrendously loud. I'm just picky. It won't be any louder than... an A/C or refrigerator. If you think the fans on your A/C = the loudest part, then you can replace the A/C fan. I happen to think the compressor is a bit too loud for a PC. In my mid 20's... I'm getting older and lamer, and much more practical
Edit 2: There are guides detailed how to convert an AC into a water chiller. I might look those up and post them here later... but if I don't, then I think a google search could return a few useful results anyway. The Xtremesystems forums might be of more use if you're considering researching this.
Edit 3: And LN2 is used primarily for temporary extreme benchmarking and world record attempts. They do get better results than on Dry Ice or Triple-Stage-Cascade phase-change (which can reach below -100C).
Edit 4: And I think you were originally thinking of phase-change in relation to Heat-Pipes... which is valid. "Phase-Change" is so vague... but we're really talking about active units with compressors.
- The evaporator is the cold side of the system, and it can look like many things:
- A "radiator" (through which we can blow air as in an A/C), or
- A block of copper on the CPU and/or GPU (direct-phase-change cooling, without chilling water), or
- A bunch of coils of copper tubing/piping in a reservoir of water (to be chilled), or
- A different heat exchanger design, still chilling water.
So there are two basic options:
1. Couple the evaporators directly to the CPU and GPU, and forgo the use of a separate chilled coolant. This allows for lower temperatures, but more difficult assembly, tuning, etc... requires more experience (esp. if you want to cool more than just the CPU).
2. Chill the water, and then you're able to cool everything as you would in a regular watercooled setup (except with added insulation to prevent condensation).
In that last scenario, the evaporator is acting as a water-chiller.
The only way around insulating for condensation is to run dual evaporators, one of which will chill the air inside a sealed PC enclosure (not the hard drives though). This air must as cold or colder than the coolant. This adds complexity to the design, however, but it has been done.
Obviously the most practical solution is to adapt an A/C unit into a water-chiller, and insulate heavily.
Edit: and it won't be horrendously loud. I'm just picky. It won't be any louder than... an A/C or refrigerator. If you think the fans on your A/C = the loudest part, then you can replace the A/C fan. I happen to think the compressor is a bit too loud for a PC. In my mid 20's... I'm getting older and lamer, and much more practical
Edit 2: There are guides detailed how to convert an AC into a water chiller. I might look those up and post them here later... but if I don't, then I think a google search could return a few useful results anyway. The Xtremesystems forums might be of more use if you're considering researching this.
Edit 3: And LN2 is used primarily for temporary extreme benchmarking and world record attempts. They do get better results than on Dry Ice or Triple-Stage-Cascade phase-change (which can reach below -100C).
Edit 4: And I think you were originally thinking of phase-change in relation to Heat-Pipes... which is valid. "Phase-Change" is so vague... but we're really talking about active units with compressors.
Last edited:
But a refrigerator is only designed to keep food cold, not continually cool new food. The motor will die out if you try doing that/leave it open. And it can't be because it's not powerful enough, since this would happen even with a freezer unit which is obviously much more powerful it terms of cooling. Although, even it can't cool much lower than the point of freezing.
I am still quite confused on the prospect of how this system works.
The evaporator allegedly turns warm air into cold air, which is then to be fanned across an area. But as the same time, instead of blowing the air, it could be used to cool the CPU, which would be a direct-phase-change cooling method. The cold air cools the CPU and then becomes warm, going back to the pressurized unit to become cold again.
In which case, it's almost the same as a water-cooling system.. except that instead of water, it uses air, and for some reason this is more efficient than an HSF solution.
That is, unless I have gravely misunderstood the system, as this seems bizarre to me.
I mean, the whole point is to cool AIR and then either use it to cool something directly (like an AC), or use the air to cool water (like a chiller - even if I am confused as to the difference between this and the previously stated 'TEC chiller')
So basically, we're back to watercooling. Somehow, I always knew I would choose chilled water. But you're suggesting is a little above the end of chilled water.. rather using a phase-change system to chill the water, making it even colder. And you're saying that the condensation can then be cured in some form? If it's cheap than direct phase, and I do have two CPU's.. maybe even some GPU's into the mix.. it sounds great. I mean, as I said, I really don't need anything below 0C.
Actually, that is not a problem. I keep all my hard drives (sparing a single SSD) outside my case. Would cooling the air cool the components down all by itself, perhaps?
Also, isn't insulation really cheap? Seems like it would be much more cost-effective than getting another evaporator.
Yes! And I feel like I am finally grasping the matter!
Edit: and it won't be horrendously loud. I'm just picky. It won't be any louder than... an A/C or refrigerator. If you think the fans on your A/C = the loudest part, then you can replace the A/C fan. I happen to think the compressor is a bit too loud for a PC. In my mid 20's... I'm getting older and lamer, and much more practical
I have a small mini-bar type of refrigerator running right beside the computer. I don't mind its noise. The AC that cools the room is right above it and I actually enjoy its peaceful murmur.
I'm in my very early 20's.. maybe I can be just as lame and practical as you
I usually prefer just talking to someone who knows all that needs to be known.
That is usually me, in regards to almost all other areas of computer hardware, but sometimes I too like to fill in some blanks, mostly because of a practical interest.
Since I have direct, daily and free access to dry ice, I was thinking of trying it out myself, at least once. Now, I need to find good insulation for the socket and a good pot. Although, I do not understand why there would be condensation at a mere 8.5C negative change. I mean, if the CPU is running @ 60C, and the motherboard is running @ 50C, it should begin to cause condensation.. but I probably have it wrong.
And I also don't fully understand the physics behind a compressor. All in due time
Basic Phase-change cooling overview with focus on A/C chiller conversion:
http://www.wc101.com/guides/refridgeration/ - This is a popular guide. In it you will find some basics:
- Compressors are the “pump” or “motor.” They provide suction on the low pressure “low-side” inlet, thereby inducting refrigerant in gas form. The compressors are the basic same design whether used in a chiller, refrigerator, or A/C.
- The refrigerant exits the compressor’s high-pressure “high-side” outlet, and leads to the condenser (essentially a radiator with narrow tube diameter) and then to the capillary tubing (which is of an even smaller diameter). Condensers in fridges are inadequate to deal with 24/7 constant duty, but you’ll notice that A/C condensers are much larger, and require a fan to remove more heat.
- The refrigerant is now in a stable liquid state, and ready to absorb heat in the evaporator, which can vary in form.
- The evaporator is placed wherever you want to remove heat. As the name implies, the evaporator absorbs heat, and evaporates the liquid refrigerant into gaseous form. In a video of a transparent evaporator in operation, you’d see what appears to be a low-flow stream of liquid… obviously that liquid is subzero refrigerant:
- In a water-chilled setup, the evaporator is separate from the water-loop, but it must chill the water, thus it is placed either inside a water reservoir (and it can look like coils of copper tubing, or a radiator (which is really the same copper tubing, but with aluminum fins to maximize surface area)), or in close-contact with the water in some other form (water-to-refrigerant plate heat-exchanger – which looks a bit like a water-air intercooler for turbocharged vehicles).
- In a direct-CPU/GPU setup, the “block” IS the evaporator. A small capillary tube feeds the block, and a larger diameter tube acts as the return line. There will always be a large layer of insulation hiding these details.
- In an A/C, the evaporator is a radiator. A fan blows air through the evaporator, and thus the air is chilled.
Refrigerants: Common legal refrigerants are r134a, and r290 (propane). A license is required for higher-performing refrigerants (r404/507/509, etc… I can’t remember). Someone more experienced might be able to design a mixture, but I’ve been told that compressor-refrigerant compatibility can be an issue.
Temperature Control: The inexpensive route is to simply replace the thermostat on the A/C with a low-temp thermostat. There are some other options with controlled switches, and pre-made control units, but that requires more research.
Efficiency: I would look for an A/C with a high EER or COP (coefficient of performance). I figure this improves the odds of getting a unit with a good compressor and properly-sized condenser. http://www.newegg.com/Product/Product.aspx?Item=N82E16896767748 – This one has an EER of 11, which is pretty good, but you’d need to make sure you have enough cooling capacity – there will be an optimal efficiency range, and maintaining lower temperatures will require more power (I once went over these figures extensively, and determined that if I ever did phase-change again, I wouldn’t go much below freezing point in order to keep the COP high). To get specifics, you’d have to identify the compressor in use and find the manufacturer data on it (or else maybe find a similarly sized Danfoss compressor and review the available Efficiency data on their website to get a rough feel for the numbers). Edit: Alternatively, you could just convert the best A/C you can afford, and insulate/build to the best of your ability, and then simply measure the power draw required (from the wall) to maintain different coolant temperatures at your system's heatloads (I'd use a load to simulate 100% CPU/GPU, and then an idle-load). Plot it out in 5-10C increments (requires variable temperature control), and choose the best compromise.
Coolant: If the water reaches below freezing point… well then let’s hope it’s NOT just water. You’d need to mix a solution of glycol/water (there are a few other options), and the ratio of the mixture would depend on your desired lowest temperature.
Insulation: Di-electric grease in the sockets, conformal coating on the PCBs, RTV-silicone optional to seal further, and Neoprene gaskets and tape jammed the hell in there TIGHT. You’ll need neoprene on all cold components. There are some insulation guides out there…
Tools: You’ll need some basic tools to modify the A/C if you want to add length to certain lines, refill with a new refrigerant, etc. (valves, air-pump, brazing or some kind of torch setup)
MORE LINKS (I need to add to this section):
http://forums.extremeoverclocking.com/showthread.php?t=101902 – Here is a log of an A/C to chiller conversion.
http://forums.extremeoverclocking.com/showthread.php?t=337515 – A more recent log, but of use are the comments by Drewmeister, the OP of the original log.
http://forums.extremeoverclocking.com/showthread.php?t=233054 – compilation of guides… includes information on refrigerants, insulation, everything… that is… if the links are still good.
http://www.xtremesystems.org/forums/forumdisplay.php?f=80 – The stickies at the top include some more guides, and this is the home of some Xtreme builders (-100C and lower cascade units – too [X] for me now).
http://www.xtremesystems.org/forums/forumdisplay.php?f=155 – Here is the chiller section. More useful stickies there obviously…
http://www.xtremesystems.org/forums/showthread.php?t=160872 – the GOLIATH build. This guy chilled the air to -40C, eliminating condensation. I’d want to do the same thing, but maybe just to around freezing point to save power.
NOW… from all of that… and after 2 weeks of studying at least 1 hour per day, you should be able to learn most of what you need to know. You can essentially ignore cascade, autocascade, and more exotic, multi-compressor setups (seriously… IGNORE, seeing those setups will derail you). You should preferentially read the essential guides and A/C conversion threads first, and keep a document or otherwise organize of all the relevant information you might use.
What I would do (if ever I got back into phase-change):
- Decide whether or not I should convert an A/C or build a DIY chiller for that extra 5% efficiency, the satisfaction of knowing I did it myself, a deeper familiarity with my setup, and possibly a slightly less noisy compressor (moot, because there’s no way I’d have the unit in the room with me).
- Decide whether or not I’m willing to do the extra work to insulate my case and chill the air as well (It does save me the work of insulating my components).
- Calculate power-consumption at different temperatures and make a decision. Hopefully I won’t be tempted to go below freezing so I can use WATER, which is a superior coolant (I’d get better CPU/GPU temps for the same coolant temp = lower coolant to CPU/GPU delta). The only problem would be to deal with aluminum-copper corrosion from the evaporator (which becomes an issue in regular water)… I’d want to use a different evaporator.
http://www.wc101.com/guides/refridgeration/ - This is a popular guide. In it you will find some basics:
- Compressors are the “pump” or “motor.” They provide suction on the low pressure “low-side” inlet, thereby inducting refrigerant in gas form. The compressors are the basic same design whether used in a chiller, refrigerator, or A/C.
- The refrigerant exits the compressor’s high-pressure “high-side” outlet, and leads to the condenser (essentially a radiator with narrow tube diameter) and then to the capillary tubing (which is of an even smaller diameter). Condensers in fridges are inadequate to deal with 24/7 constant duty, but you’ll notice that A/C condensers are much larger, and require a fan to remove more heat.
- The refrigerant is now in a stable liquid state, and ready to absorb heat in the evaporator, which can vary in form.
- The evaporator is placed wherever you want to remove heat. As the name implies, the evaporator absorbs heat, and evaporates the liquid refrigerant into gaseous form. In a video of a transparent evaporator in operation, you’d see what appears to be a low-flow stream of liquid… obviously that liquid is subzero refrigerant:
- In a water-chilled setup, the evaporator is separate from the water-loop, but it must chill the water, thus it is placed either inside a water reservoir (and it can look like coils of copper tubing, or a radiator (which is really the same copper tubing, but with aluminum fins to maximize surface area)), or in close-contact with the water in some other form (water-to-refrigerant plate heat-exchanger – which looks a bit like a water-air intercooler for turbocharged vehicles).
- In a direct-CPU/GPU setup, the “block” IS the evaporator. A small capillary tube feeds the block, and a larger diameter tube acts as the return line. There will always be a large layer of insulation hiding these details.
- In an A/C, the evaporator is a radiator. A fan blows air through the evaporator, and thus the air is chilled.
Refrigerants: Common legal refrigerants are r134a, and r290 (propane). A license is required for higher-performing refrigerants (r404/507/509, etc… I can’t remember). Someone more experienced might be able to design a mixture, but I’ve been told that compressor-refrigerant compatibility can be an issue.
Temperature Control: The inexpensive route is to simply replace the thermostat on the A/C with a low-temp thermostat. There are some other options with controlled switches, and pre-made control units, but that requires more research.
Efficiency: I would look for an A/C with a high EER or COP (coefficient of performance). I figure this improves the odds of getting a unit with a good compressor and properly-sized condenser. http://www.newegg.com/Product/Product.aspx?Item=N82E16896767748 – This one has an EER of 11, which is pretty good, but you’d need to make sure you have enough cooling capacity – there will be an optimal efficiency range, and maintaining lower temperatures will require more power (I once went over these figures extensively, and determined that if I ever did phase-change again, I wouldn’t go much below freezing point in order to keep the COP high). To get specifics, you’d have to identify the compressor in use and find the manufacturer data on it (or else maybe find a similarly sized Danfoss compressor and review the available Efficiency data on their website to get a rough feel for the numbers). Edit: Alternatively, you could just convert the best A/C you can afford, and insulate/build to the best of your ability, and then simply measure the power draw required (from the wall) to maintain different coolant temperatures at your system's heatloads (I'd use a load to simulate 100% CPU/GPU, and then an idle-load). Plot it out in 5-10C increments (requires variable temperature control), and choose the best compromise.
Coolant: If the water reaches below freezing point… well then let’s hope it’s NOT just water. You’d need to mix a solution of glycol/water (there are a few other options), and the ratio of the mixture would depend on your desired lowest temperature.
Insulation: Di-electric grease in the sockets, conformal coating on the PCBs, RTV-silicone optional to seal further, and Neoprene gaskets and tape jammed the hell in there TIGHT. You’ll need neoprene on all cold components. There are some insulation guides out there…
Tools: You’ll need some basic tools to modify the A/C if you want to add length to certain lines, refill with a new refrigerant, etc. (valves, air-pump, brazing or some kind of torch setup)
MORE LINKS (I need to add to this section):
http://forums.extremeoverclocking.com/showthread.php?t=101902 – Here is a log of an A/C to chiller conversion.
http://forums.extremeoverclocking.com/showthread.php?t=337515 – A more recent log, but of use are the comments by Drewmeister, the OP of the original log.
http://forums.extremeoverclocking.com/showthread.php?t=233054 – compilation of guides… includes information on refrigerants, insulation, everything… that is… if the links are still good.
http://www.xtremesystems.org/forums/forumdisplay.php?f=80 – The stickies at the top include some more guides, and this is the home of some Xtreme builders (-100C and lower cascade units – too [X] for me now).
http://www.xtremesystems.org/forums/forumdisplay.php?f=155 – Here is the chiller section. More useful stickies there obviously…
http://www.xtremesystems.org/forums/showthread.php?t=160872 – the GOLIATH build. This guy chilled the air to -40C, eliminating condensation. I’d want to do the same thing, but maybe just to around freezing point to save power.
NOW… from all of that… and after 2 weeks of studying at least 1 hour per day, you should be able to learn most of what you need to know. You can essentially ignore cascade, autocascade, and more exotic, multi-compressor setups (seriously… IGNORE, seeing those setups will derail you). You should preferentially read the essential guides and A/C conversion threads first, and keep a document or otherwise organize of all the relevant information you might use.
What I would do (if ever I got back into phase-change):
- Decide whether or not I should convert an A/C or build a DIY chiller for that extra 5% efficiency, the satisfaction of knowing I did it myself, a deeper familiarity with my setup, and possibly a slightly less noisy compressor (moot, because there’s no way I’d have the unit in the room with me).
- Decide whether or not I’m willing to do the extra work to insulate my case and chill the air as well (It does save me the work of insulating my components).
- Calculate power-consumption at different temperatures and make a decision. Hopefully I won’t be tempted to go below freezing so I can use WATER, which is a superior coolant (I’d get better CPU/GPU temps for the same coolant temp = lower coolant to CPU/GPU delta). The only problem would be to deal with aluminum-copper corrosion from the evaporator (which becomes an issue in regular water)… I’d want to use a different evaporator.
Last edited:
That is a fantastic amount of good information (for the most part; your condenser/compressor description is not 100% accurate), but I think the OP would be better off if he could see the bigger picture of heat transfer. Without getting into specific details about which specific refrigerants or TECs or phase changes and everything- you need to understand the theory behind moving heat (energy) from one place to another.
thermal (heat) capacity is the capacity a substance has to store heat (energy; usually measured in joules per unit of temperature). It is dependant on several things, but one of the main properties that influence heat capacity in the terms that we are using it for is the density of the substance.
thermal conductivity is the ability of a substance to readily absorb heat energy. Generally, for example, metals have a much higher thermal conductivity then plastics do.
For example- water has a very large capacity to store/transport heat. As do heavier metals like iron and copper and gold and mercury. Aluminum, being less dense, isnt able to store heat as well as copper for instance. This is why most good heatsinks are made from copper. Gold and silver are even better than copper, but for reasons of cost and reactivity of the metals, are not used.
Gasses acutaly have similar heat capacities of metals by weight . But in terms of heat storage per unit of volume, the thermal capacity of gas (air) is very poor in comparison. You would need tons more air volume blowing over a heatsink to remove the same amount of heat than you would if you poured water over the heatsink instead.
Phase-change cooling works by utilizing the properties of a gas under different environments- by changing temperature and pressure. Gasses liquify under high pressure or 'cold', and then return to a gaseous state when they get hot enough or when under a low enough pressure. Its an inverse relationship.
Putting enough pressure on a gas causes it to liquify. But pressure produces/releases heat.
Pulling enough vacuum on a liquid causes it to gassify. But that process uses heat.
Liquifying a gas releases as much heat energy as gassifying a liquid uses .
if you force the refrigerant gas to liquify under pressure in the condenser (hot side of the air conditioner, the part you put outside the window) you can then pump that now liquid refrigerant into the evaporator (cold part of the air conditioner) where there is a low pressure environment and it will turn back into a gas, taking the heat available in the evaporater with it back up the loop to the condenser again.
Warm room air is then blown over the chilled evaporater and circulated thru the room, cooling it.
Though from a scientific standpoint it is more accurately described as this - warm room air passes over the evaporater and provides the energy the refrigerant needs to turn into a gas. Theoretically if the room were very cold to begin with, the refrigerant would not have the energy it needs to transform into a gas.
Remember that cold is really the absence of heat. Heat is energy, absence of heat = absence of energy. The word 'cold' as we normally say it is only a relative view from the standpoint of a human.
A passive heatpipe works on the same gas-liquid principals except exactly opposite execution from an active air conditioner. The liquid heats up near the CPU, then turns to gas which travels up the center of the pipe where it releases energy into the fins of the heatsink and condenses on the walls of the tube. It is then transfered thru convection and capillary action back down the inside walls of the pipe to the CPU where it is reheated and repeats.
Instead of controlling the pressure , as an air conditioner does, a heatpipe controlls the temperature of the refrigerant but uses the exact same phase-changing effect.
thermal (heat) capacity is the capacity a substance has to store heat (energy; usually measured in joules per unit of temperature). It is dependant on several things, but one of the main properties that influence heat capacity in the terms that we are using it for is the density of the substance.
thermal conductivity is the ability of a substance to readily absorb heat energy. Generally, for example, metals have a much higher thermal conductivity then plastics do.
For example- water has a very large capacity to store/transport heat. As do heavier metals like iron and copper and gold and mercury. Aluminum, being less dense, isnt able to store heat as well as copper for instance. This is why most good heatsinks are made from copper. Gold and silver are even better than copper, but for reasons of cost and reactivity of the metals, are not used.
Gasses acutaly have similar heat capacities of metals by weight . But in terms of heat storage per unit of volume, the thermal capacity of gas (air) is very poor in comparison. You would need tons more air volume blowing over a heatsink to remove the same amount of heat than you would if you poured water over the heatsink instead.
Phase-change cooling works by utilizing the properties of a gas under different environments- by changing temperature and pressure. Gasses liquify under high pressure or 'cold', and then return to a gaseous state when they get hot enough or when under a low enough pressure. Its an inverse relationship.
Putting enough pressure on a gas causes it to liquify. But pressure produces/releases heat.
Pulling enough vacuum on a liquid causes it to gassify. But that process uses heat.
Liquifying a gas releases as much heat energy as gassifying a liquid uses .
if you force the refrigerant gas to liquify under pressure in the condenser (hot side of the air conditioner, the part you put outside the window) you can then pump that now liquid refrigerant into the evaporator (cold part of the air conditioner) where there is a low pressure environment and it will turn back into a gas, taking the heat available in the evaporater with it back up the loop to the condenser again.
Warm room air is then blown over the chilled evaporater and circulated thru the room, cooling it.
Though from a scientific standpoint it is more accurately described as this - warm room air passes over the evaporater and provides the energy the refrigerant needs to turn into a gas. Theoretically if the room were very cold to begin with, the refrigerant would not have the energy it needs to transform into a gas.
Remember that cold is really the absence of heat. Heat is energy, absence of heat = absence of energy. The word 'cold' as we normally say it is only a relative view from the standpoint of a human.
A passive heatpipe works on the same gas-liquid principals except exactly opposite execution from an active air conditioner. The liquid heats up near the CPU, then turns to gas which travels up the center of the pipe where it releases energy into the fins of the heatsink and condenses on the walls of the tube. It is then transfered thru convection and capillary action back down the inside walls of the pipe to the CPU where it is reheated and repeats.
Instead of controlling the pressure , as an air conditioner does, a heatpipe controlls the temperature of the refrigerant but uses the exact same phase-changing effect.
Agreed. I was thinking about posting a detailed follow-up with some basic thermodynamic principles, comparing water, other coolants, gases, and some metals with specific figures and formulas (which I'd have to reference)... but it looks like you've done some of that work.
I see that you've explained the processes more accurately, but if you could point out the specific errors in my descriptions, then that'd be helpful. My descriptions were for illustrative purposes to get an idea for the workflow, but obviously I am very rusty.