Loop Heat Pipe (no condenser) cooling of 1U system

[psdillon]

** That should be no "NO COMPRESSOR" in the thread title... doh.

I'm looking for advice on implementing LHP/Loop Heat Pipe cooling in a 1U enclosure.

So far I've only been able to find prototypes and press releases. Has anyone come across anything commercially available along these lines?

Alternatively, is this the kind of thing I could commission an XS builder to do?


FYI here are some photos of a prototype by Passive Thermal Technology:






PS - apologies if this is the wrong forum - while LHP is a vapor change system I realise this forum is more about systems that use a compressor.

[n00b 0f l337]

This isn't phase change in the sense that your thinking.
This is simply a heat pipe. Kinda alcohol in a tube. Moves along on it's own via a wick or pressure difference.

That makes it not a vapor phase change kinda thing.

[Buckeye]

Not sure why you would want cooling like this for 1U setups when normally they are air cooled and work just fine.

There are companies that use vapor change/water cooling for custom racks, but they cool large racks with them.

The pictures you posted look rather old and that's for a AMD system. Plus there seems to be a whole bunch missing from the setup. Heat sinks like those need powerful high speed fans to blow across them to cool. Plus those are at the front of the case so any hot air being blown across them would go right over the motherboard and heat that area up.

Well maybe they are at the back of the case now that I look at it more.

[psdillon]

Not sure why you would want cooling like this for 1U setups when normally they are air cooled and work just fine.

I'd like to overclock an i7-3970X, and I need to run it in a data centre where space is a premium.

The pictures you posted look rather old and that's for a AMD system. Plus there seems to be a whole bunch missing from the setup. Heat sinks like those need powerful high speed fans to blow across them to cool. Plus those are at the front of the case so any hot air being blown across them would go right over the motherboard and heat that area up.

Well maybe they are at the back of the case now that I look at it more.

Yeah it's a weird chassis configuration. There are definitely some poor design decisions in there, even for a prototype, but it does show the LHPs clearly.

Here's another old prototype photo, but this one using a more typical chassis. The long aluminium bars below the fan bar are the condensers. In a typical 1U case this is the only space available to install a condenser as the motherboard is always hard up against the back of the case. As you say not ideal as the hot air flows over the motherboard.

[Buckeye]

Well good luck with that, there is nothing out there that will work in a 1U case.
What motherboard do you plan on using for this ?

You found some prototypes and press releases ?

[mytekcontrols]

This isn't phase change in the sense that your thinking.
This is simply a heat pipe. Kinda alcohol in a tube. Moves along on it's own via a wick or pressure difference.

That makes it not a vapor phase change kinda thing.

Actually it could be phase change, by using the right refrigerant. Basically you just need something that will boil off at the maximum heat you wish to run the CPU at. So for instance R123 boils at +28 C (+82 F), and would make a phase change from liquid to vapor at this point. So to insure the most effective heat transfer, the CPU side should be slightly lower then the condensing side. This way the heated vapor travels up, condenses, and then by gravity the condensed refrigerant flows back down to the CPU side. Basically a phase change thermal siphon.

I built something like this many years ago for solar heating, using R113 as the refrigerant. It worked quite well as I remember.

This can also be implemented using a single pipe of a larger volume, where the liquid and gas travel together in a concurrent fashion, passing by each other.

[Buckeye]

I am not sure if this is a phase change system as there is no comp to power it

[psdillon]

Actually it could be phase change, by using the right refrigerant. Basically you just need something that will boil off at the maximum heat you wish to run the CPU at. So for instance R123 boils at +28 C (+82 F), and would make a phase change from liquid to vapor at this point. So to insure the most effective heat transfer, the CPU side should be slightly lower then the condensing side. This way the heated vapor travels up, condenses, and then by gravity the condensed refrigerant flows back down to the CPU side. Basically a phase change thermal siphon.

You're right about all that. As I understand it, a loop heat pipe is more efficient than normal heat pipe because it also uses convection in a loop, with a wick structure in the condenser->evaporator tube if it needs to operate against gravity.

In the second prototype above, they used water, with a cold plate temperature of 70 C. I'm guessing they lowered the boiling point of water by lowering the pressure.

But yeah, liquid->vapour means phase change to me too, but I agree with others that "vapour phase change cooling" commonly means a refrigeration system with compressor.

[psdillon]

Well good luck with that, there is nothing out there that will work in a 1U case.
What motherboard do you plan on using for this ??

You found some prototypes and press releases

Probably one of the Asus X79 overclocking boards. I've searched high and low for a board that supports overclocking as well as IPMI/KVM-over-LAN but no luck. Thankfully memory orientation isn't an issue with LHP, otherwise I'd be trying to jam an oversized eVGA SR-X into a 1U case.

Buckeye, can I ask, are you active in the server space? The LNO website claims to have done this with their "flash servers", but it's all very unconvincing. Was this after your time?

[Buckeye]

Loop heat pipe technology has been around for sometime really. From I guess 2008 on or so, maybe earlier. Some companies make these for various systems but are generally pretty costly setups. Most of the companies that make these do so for data centers and have the system installed in full sized racks with the cooling units being built into the racks itself.

The 1U area is a pretty hard nut to crack as far as cooling goes as space is so limited, but with an external cooling unit you can do a lot and perhaps cool many 1U servers.

As far as self contained cooling units I build several different types for 1U, 2U & 4U and even some with external cooling units that cool stacks of servers.
I have several I am starting work on Monday and later doing on site installs for a larger amount, plus other projects I am not ready to disclose atm.

So to answer your question, yes I am very active in the server space.

As far as LNO goes the server you mentioned is pure fabrication and not possible at all.

As you brought that up I feel I must direct you to this link as it will explain LNO & Mr. Ed Trice
http://www.xtremesystems.org/forums/...c&daysprune=-1

[psdillon]

Thanks Buckeye. Hope you can post some details in the future of your server builds, would love to know more about them.

Thanks, I've read those posts before - even though his website is full of exaggerations and inconsistencies, it never clicked to me that he was a fraudster until seeing those posts on XS. It's great that you posted your experience, it must surely have saved others from similar experiences.

[Buckeye]

Actually it could be phase change, by using the right refrigerant. Basically you just need something that will boil off at the maximum heat you wish to run the CPU at. So for instance R123 boils at +28 C (+82 F), and would make a phase change from liquid to vapor at this point. So to insure the most effective heat transfer, the CPU side should be slightly lower then the condensing side. This way the heated vapor travels up, condenses, and then by gravity the condensed refrigerant flows back down to the CPU side. Basically a phase change thermal siphon.

I built something like this many years ago for solar heating, using R113 as the refrigerant. It worked quite well as I remember.

This can also be implemented using a single pipe of a larger volume, where the liquid and gas travel together in a concurrent fashion, passing by each other.

I think this might be something like what you are saying Michael.

http://www.frostytech.com/articlevie...id=2576&page=2

It did not do very well in cooling tho.

[mytekcontrols]

I think this might be something like what you are saying Michael.

http://www.frostytech.com/articlevie...id=2576&page=2

It did not do very well in cooling tho.

Yep by it's image and description of having to keep the fan cooled condenser higher, it most likely is very similar to what I was mentioning. If it's cooling was poor, I would probably attribute this to poor heat exchange design and/or inappropriate refrigerant charge. If done correctly, it should yield much better results than a passive heat exchanger/fan combo. But not as good as true a complete phase change system utilizing a compressor.