Thermosyphon Cooling Solution

[Dr_Emmett_Brown]

As a mad scientist, I have experimented with many different heat removal systems. Apparently, with Intel and AMD engaged in a little "heat war," where Intel can only outperform AMD with higher overclocks, even a triple radiator can't quite keep up if you are running all 8 cores of an i9-9900KS at something like 5.2 GHz.

A new type of heat removal was needed, since boiling water in a radiator won't be cooled sufficiently, even with 6 foot tall fans from the Boeing Wind Tunnel.

So here is my mock up solution, perhaps a little extreme:



The cold plate is solid copper.

The fins in the condenser (no radiator, vapor flows through it, not water) are also copper.

Copper has the wonderful property of having nearly twice the thermal conductivity as aluminum. That basically means you get the cooling performance as if you had twice as much surface area in the fins when compared to aluminum. Even still, I went with a huge condenser. It's the length of my elbow to my wrist.

I ordered two high speed fans that were BIG, as you can see. They can hit 5000 RPM with no problem. Of course, the pressure drop across the fins creates about 65 db of noise, far too loud to be practical. I will look to replace them with smaller fans, possible 4 of them in push-only configuration. The lower fan speeds x more fans should be able to deliver the same cubic feet per minute I require.

I use a very volatile cocktail of refrigerants so that they turn to vapor upon striking the cold plate. This vapor phase change is very endothermic. It produces fantastic temperatures even running Prime95 on all 8 cores 24x7 at 5.2 GHz.

I will show some more crazy builds and some build logs in the future. I also design my own custom cases, which you probably could tell.

 

[thenrz]

So it's a phase change system where rather than using a compressor, you're using some kind of other pump? What refrigerants are you using to keep liquid at lower pressures? Do you have any issues with pooling at the bottom or is that essentially what you want here? What kind of pressure we talking? I'm assuming middle of the road considering the materials that comprise the loop. Copper is really the only way to go... Even machinable alloys like tellurium 145 only lose a couple percent. Are you machining this?

That setup reminds me of the first phase change system I built. Things were everywhere, but it worked so who cares!

Post some more pictures. It's difficult to tell specifics from the one. Also, where do you get your extrusion? My usual supplier has raised prices when I went to buy some the other day. I can get it through work, but they aren't ordering from that vendor for a couple more weeks, so my free ride on shipping has to wait.

 

[Dr_Emmett_Brown]

So it's a phase change system where rather than using a compressor, you're using some kind of other pump?

The thermosyphon operates pumpless but needs a decent gravity drop to be effective. When the liquid touches the cold plate, it vaporizes, increasing in volume, and also increasing the pressure in the loop enough to push it back up to the condenser. That's why the build is so tall.

What refrigerants are you using to keep liquid at lower pressures? Do you have any issues with pooling at the bottom or is that essentially what you want here? What kind of pressure we talking? I'm assuming middle of the road considering the materials that comprise the loop. .... Are you machining this?

It's a cocktail of refrigerants plus some glycol for good measure. As long as the cold plate is hot and above the low vaporization temperature of the fluid, there's no chance of pooling. The solid copper cold plate is custom machined, yes.

Post some more pictures.

The first condenser. I went with a larger one later.


A good image of the aluminum from another case build


The finished version of the frame build shown above. Internal monitor and wireless keyboard and mouse

 

[thenrz]

Ah yes, gravity, the free pump. That condenser looks a lot like the cubic ones I used to use. I always enjoyed the unfinished ends so I could braze on my own U bends to complement whatever geometry and fan system was being used for that build.

I really like the extrusion you're using. I have a design with what may be the same stuff, but it's a much different target use case. I would imagine refrigerant wise you're not using anything like R510 or anything like that... off the top of my head, i can't think of any that would behave like you'd want without a lot of pressure. I'm way out of the loop with such things anyway.

Do you have any hard numbers? Running specific wattage tests? I've had to learn a lot lately about consistent testing with my waterblock designs, and it's not as easy as it appears. Luckily with Zen 2, I can dial things in to hit specific targets relatively easily after a lot of trial and error.

 

[Dr_Emmett_Brown]

Ah yes, gravity, the free pump.

That's half of the free pump The other half comes from the PRECISE calculation of the volume of each refrigerant added to the carefully calculated volume of the circulation loop. I can't tell you how many times I literally "blew a gasket" by not getting this correct to a single decimal digit. The pressure of vaporization pushes it up, like the pump would. The gravity provides the fluid with the correct Bernoulli Brothers' viscosity velocity needed to impart the correct momentum prior to the vaporization transfer. A pain in the butt to calculate once. Thankfully, and Excel spreadsheet with the correct formulae will allow you to handle different heat loads easily.

I would imagine refrigerant wise you're not using anything like R510 or anything like that... off the top of my head, i can't think of any that would behave like you'd want without a lot of pressure. I'm way out of the loop with such things anyway.

It's the change in pressure you have to worry about. Once the loops are sealed, the pressure itself does not matter (too much). But if it changes too drastically, poof!

Do you have any hard numbers? Running specific wattage tests?

I can tell this this: don't go by the published Intel TDP numbers, at least for the i9-9900KS chip. It's giving off way more heat than 127 watts. At 5.2 GHz I'm closer to 260 watts when running all 8 cores/16 threads. It's very very hot. The CORSAIR H150i water cooled triple-radiator with all 3 fans set at 1500 RMP through the iCue software app (all 3 functionally dongeled together as 1) could not keep up with the heat load. It was as 94 C initially, then 96 max at 10 minutes, 99C after 20 mins, 102C a short time later, then thermal shutdown.

 

[thenrz]

I run numbers for my own project, and tdp values are marketing nonsense for base clocks and nothing more. It's about power levels and basic monitoring. Things like hwinfo are key for monitoring things.

Sounds like you're doing things correctly. When you feel like posting some test data, do so. Curious to see how this hybrid approach handles some wattage.

 

[Dr_Emmett_Brown]

I run numbers for my own project, and tdp values are marketing nonsense for base clocks and nothing more. It's about power levels and basic monitoring. Things like hwinfo are key for monitoring things.

Sounds like you're doing things correctly. When you feel like posting some test data, do so. Curious to see how this hybrid approach handles some wattage.

Each success comes at the price of many failures. But I like seeing how thermo theory maps to hardware specs, and how well they match the anticipated results. The devil is in the details, as they say, and especially with fluids, one must be ever-vigilant with regard to vortices and eddies. Pumpless is one layer of complexity removed from a system with a pump, but the system must be in a bedrock-stable environment and not jostled too often.

Maybe you can show me some sample data so I can see what you are referring to. Most of my setups and teardowns were just that. Not much history recorded when stuff didn't work! I also find HWMonitor lacking in one important regard: It does not produce time-lapse data, as the way the Task Manager does with its 60-second window.

I started rating the performance of the system as a whole by the WORKING FLUID TEMPERATURE. If you think about it, this is a great way to rate your system as a whole. If the working fluid temperature = ambient temperature, you have an idealistic (and impossible) heat removal system. You measure the temperature of the fluid as it leaves contact with the cold plate. Track this as a function of time. Initially, t0 = ambient room temperature. It should spike up sharply, level off, and, hopefully, maintain a steady state.

You can then experiment with constant-load tests to see if your system is radiator-performance bound, airflow (fan) bound, or other-stage-bound (if you have one).

For example, I discovered my smaller condenser was overloaded and too small. It was hypersensitive to the fan speed. Drop the RPMs a little, it got way hot. Increase them to full speed, not much cooling gained, if any. That means the condenser (in your case, the radiator) would be an insufficient size. Next, I went too big. With 4 fans at max speed, it cooled as much as a single fan at 500 RPMs. That means the condenser was doing all of the work, with more surface area that was needed. Somewhere in the middle, I got the perfect match. Cooling varied linearly with fan speed and at 80% max speed, it reached the asymptote line which was good enough for me. If I ever swap up to a slightly hotter chipset, I should be good for a while.

 

[Zerileous]

What were your temps under prime95? Also which test, blend or small FFT? Any AVX offset / was AVX disabled in prime?

I wonder how much of the difference compared to the CLC has to do with using a thermosyphon vs using 5000rpm fans. Did you create your own block / cold plate as well? Looks like a pretty cool project and if the results are good that's awesome! Is this a personal project or something you're hoping to sell? If the latter I believe there are some applicable patents to keep in mind.

 

[Dr_Emmett_Brown]

What were your temps under prime95?

To be absolutely sure, I should rerun the tests in this final configuration. I had this system running at 5.0, 5.1, 5.2, and 5.3 GHz,each with slight tweaks to the hardware here and there.

Any AVX offset / was AVX disabled in prime?

It was run at full bang for the buck with 0 down-throttling.

I wonder how much of the difference compared to the CLC has to do with using a thermosyphon vs using 5000rpm fans.

As I alluded to in prior posts, the condenser does most of the heat removal. The fan speeds were not critical. In fact, I think they are running at a nice low 800 RPM x 4 fans in push-only mode.

Did you create your own block / cold plate as well?

Yes. The cold plate is solid copper. All the cool stuff I experiment with is solid copper. Well, there's a parallel plate heat exchanger in the photo, that has copper and aluminum plates alternating, but mostly copper.

Looks like a pretty cool project and if the results are good that's awesome! Is this a personal project or something you're hoping to sell? If the latter I believe there are some applicable patents to keep in mind.

Thank you.

I'm putting it together into a custom aluminum frame case with plexiglass walls.



Most of the DIY box builders don't really want to pay for builds. Or, they only want to pay for the price of parts and $0 for the assembly time. So I'm in no hurry to sell anything to folks on here.

Patents are beyond the scope of what I'd like to discuss on here. They seem to only invite fights.

I just want to build cool, fast systems, run them as fast and as cool as I can, and put them into attractive cases when I am finished.

 

[Zerileous]

Very cool, I'll look forward to seeing the numbers.

No worries re: the patents, I just wanted to give you a head's up before you made substantial investment since the technology is fairly obscure.

 

[Dr_Emmett_Brown]

Very cool, I'll look forward to seeing the numbers.

No worries re: the patents, I just wanted to give you a head's up before you made substantial investment since the technology is fairly obscure.

Thanks. By the way, I have been trying to figure out where the AVATAR settings are. I have no idea how to edit that in my profile. Reminds me of the book "Don't Make Me Think."

Any chance you can tell the new guy how to do it?

 

[Zerileous]

Go to settings in the top right of the page, then you can edit avatar and signature on the bar to the left.

 

[Dr_Emmett_Brown]

Go to settings in the top right of the page, then you can edit avatar and signature on the bar to the left.

I don't have any such option on my page. Did they remove the avatar option for new members?

 

[trents]

At the top right corner of the page, click on Settings and then you will see the choices on the left side that Z alluded to.

One question, will this system fit into an ITX case?

 

[Dr_Emmett_Brown]

At the top right corner of the page, click on Settings and then you will see the choices on the left side that Z alluded to.

One question, will this system fit into an ITX case?

Trust me, this is as annoying for me as it is for you guys trying to point me in the right direction!

Here is what I see on my settings page

The text at the top in red is a complete mystery. Obviously I completed my registration, but some software bit in the forum is not set.

And, as you can see below the message, I don't have the options you guys have.

If it takes more than one click to get there, nobody has told me what to do after that first click.

- - - Auto-Merged Double Post - - -

Dr, please post photo here rather than posting an unknown (to me) link.

OK, sorry about that. If you google for "image hosting" you will see imgbb.com show up in the list. They're not harmful. I uploaded my image there so you guys could see the screen shot but I must not have used the correct BB-style to get the image to show. I am still learning

- - - Auto-Merged Double Post - - -

And no, the system won't fit into any case, which is why I build my own.

 

[Dr_Emmett_Brown]

Thanks to everyone who helped with my avatar. It turned out to be a tech support issue. A flag in the database needed to be changed, then, viola, everything you guys said was in the correct place! Thanks again!

 

[trents]

Now you just need a Signature.

 

[RollingThunder]

Thanks to everyone who helped with my avatar. It turned out to be a tech support issue. A flag in the database needed to be changed, then, viola, everything you guys said was in the correct place! Thanks again!

Doc,

We're good now, thank you.

 

[Dr_Emmett_Brown]

Why isn't this new video working?

 

[JLK03F150]

Are you trying to insert a youtube video? If so change the https to http in the embedded link.