Since I have played around with my Hailea chiller with water down to 3 Celsius and not being able to get high enough scores for HWbot I though of trying sub zero cooling. Having also dealt with the condensation issues and learning the issues with such I got hold of some new toys.

I received the package from PCS computing in USA who was the closest that had the equipment i needed. On the left is the EK SF3D GPU pot and in the middle the KingPin F1 Dark CPU pot . On the right the Kingpin Heater for the back of the CPU. Real heavy pieces indeed 🙂

Reading up on various pages I found some more or less conclusive posts about how to set the equipment up.  Meanwhile finding the material needed in Manila is an adventure by itself.

I decided to try out the toys on my older Maximus  iX Apex and the 6700K process as a learning platform before i move on to my 2066  more expensive CPUs.

Preparing motherboard for sub zero cooling with dry ice

The first thing to do is the preparation of the Motherboard  and for that i decided to go with the vaseline approach.  Thus out with a toothbrush and smear up the areas around the CPU socket components to ensure they are waterproofed in case there is any spillage of splattering. Then I decided to go for the Chamois extra absorbant PVA material used for polishing cars rather than paper as it is equally absorbing compared to the papertowels but rather more sturdy and thicker.

I put 1 layer of about 1mm thickness close cut to the CPU socket . (note the picture show the watercooling pins ratehr than the pot mounting pinsthe vaseline covered areas are the more shiny parts of the PCB.

Then I close the CPU socket above the same so it is firmly in place with the CPU.

Then  I prepared a second gasket that will be snug to the socket mounting bracket that can ft for both the watercooling block or in use with the F1  pot.

I vaselined the areas around the CPU on the backside of the board as well. Since I got the heater for the backside of the CPU . Rather than using the thinner gasket that comes with the pot as a full cover I cut out the fit for the CPU backplate from the thicker gasket and then used the thinner gasket just on the CPU center components. This means I can get a snug fit when of the critical areas of the CPU components covered. Its otherwise very hard to tighten the heater and gasket enough even though it is extremely soft as material.

Since neither the heating element nor the Fl1 Dark Pot comes with any instruction manual its a bit of a trick first time to match up the mounting holes. As you can see on the pictures the only way to get it to work is to use different holes thus not symetrical. Maybe Kingpin could color code the holes that match with the different sockets and or supply a manual 🙂

Here the two home made gaskets are attached.

Then i used another chamois piece where i cut out a square hole that is slightly wider than the CPU but not as wide as the two bottom ones. This one  I place just when mounting the pot so it will give a snug fit with the pot.  I did not find any nice polypropylene tubes in the shops so alternative solutions is used to give more isolation for the pot. Learning and experimenting  🙂

Here the pot is on and some dry ice is in the pot. Since it was evening when i came back with 20 kg dry ice in 1 kg blocks the hardware store was already closed.  Reading up on internet I could see that i needed some type of alcohol or acetone to get a good contact with the dry ice as that will help the cooling.

What type of alcohol to use for dry ice cooling of CPU

I could not find any pure alcohol in the pharmacy but they did have some nail remover with acetone marked 30. Not knowing if it was 30% or 30 Proof (15%)  reading up on internet i concluded it would freeze around -5 to – 10c.  Since i wanted to test things out this was perfectly of for me as an initial trial.  So i went ahead and poured it in and dropped in some dry ice. Not having any understanding of the coldbug effect that i see mentioned here and there i turned on the computer with the pot at room temp after having poured in the acetone. then i started realtemp and Hwmon.

As i dropped in a dry ice peace it started bubbling in the acetone and the combination of a 25 C pot and dry ice with the computer on means there was a lot of vapor from the dry ice. If you try this the first time Be Careful! due to the dry ice working relatively fast the acetone will look like it boils  until the pot has reached a negative temperature. thus only use small amounts of liquid and small amounts of dry ice so it does not splash all over the MB. It will splash far enough to land outside the areas covered with vaseline. To ensure no damage was there for my first test i put papertowels all over the MB and shielded the GPU and other parts.

then i put in bigger pieces of dry ice as the pot was cooler. Realtemp stopped at 0C but the Hwmon continued to show temperatures thus the cpu sensor internally worked. while overclocking and testing the 30% acetone nail remover started to freeze around -5 to – 7 C while i tested some benchmarks. No matter how big chunks of dry ice i put in and even if i put in more acetone the temps would not go lower as the acetone solvent would freeze and melt to insulate the pot from the dry ice.

For me this was not a failure but rather a good way to  start testing in a controlled way. Meanwhile since it was in the middle of the night and no shops open i went to my fridge and got hold of the next best thing in alchohol I had 🙂  Bacardi 40% alchohol 🙂 Replacing the nailpolish remover and pouring in some Bacardi and lubricating my throat with the same I restarted the  process. This time the temperatures rapidly fell to – 17C on the hwmon. Effectively it should be freezing at -21C.  Again i did some benchmarks and could see that I could increase core speeds  100- 300 Mhz and still keep the  Hwmon temps  on the CPU within a good range even if i increased the vcore.  I would max out the temperatures during a benchmark session at around 60C when even if i increase the voltages I could not go higher on the core clocks.

I was playing around for about 8 hours until the sun came up and as you can see on the top part of the pot there is ice forming in a thick but powdery form. the moment you touch the pot the powdery snow will fall off.  Meanwhile my MB and CPU is safe below. Important to note is that with the benching session over it is important to remove the pot from the MB as once the ice starts melting the MB will otherwise get wet and that also goes for the CPU.  I have a heatgun but no blowtorch thus rather remove the pot completely to be safe.

Doing so I also noticed that the Pot had not been mounted 100% correct. I had noticed that one core was 10C – 30C higher than the rest in any benching and I believe this is the reason.

As you can see on the picture above the TIM is not showing the same pattern at the bottom part of the CPU as the other areas. I had pressed down and twisted the pot to make sure to even the tim before I fixed the same with the standard crosswise tightening.

the next morning I headed over to the closest hardware store and I found Denatured Alcohol which is used in some wildlife stoves or as a cleaning solvent. For day 2 and from now on I will use this instead. It is pure alcohol and  has a freezing point of around -114C. Considering that dry ice  melts or sublimates at – 78.5C. Thus no matter how much dry ice i put in the pot it would not freeze the denatured alchohol. Do however note that the denatured alcohol at room temperature in a CPU pot even with the computer off will look as if it starts to boil. thus be very careful putting in small amounts of dry ice in the beginning. You will hear noise and notice splatter from the dry ice initially due to the temperture difference but this will stabilise once you have the pot down to -20 – 30 C and then its safe to put bigger amounts of dry ice in the pot. Again since I dont know the coldbug issues and how to start the machine i first tried with the computer on in the process. Meanwhile i did continue to bench the whole weekend and after about 20 or so hours I had more or less consumed the 20 kg of dry ice.  testing out benching on the 6700, 8350 and the 8700K only focusing on the CPU I managed to go from around 59 -60th place in the apprentice league to 40th place in the same league.

So not a bad weekend playing with the new toys. Next weekend I will continue on the CPU benchers but on the 2066 platform 7740x and 7800x and maybe test it out on my 8086K.

Meanwhile the machine is back on the chiller while i play around with memory timings.