[karma0013]

Got the new fan installed. Stole the fans from the Kracken and added a couple of additional fans to circulate air downwards. Since it is all air over now, circulating the warm air downwards will be more important than it was with a liquid cooler.

Also made a custom sensor panel for Aida and added a 10”lcd touch panel to monitor the system without having to be at the keyboard / monitor. Sadly I made it the wrong resolution, or cannot correct the screens aspect ratio so I will have to remake the sensor panel. Amazon said it was 1024x600 so I made a panel that size and windows says it’s 1024x768 ugh back to the drawing board.

 

[karma0013]

Dont get you panties in a knot. Constructive criticism from somebody that has been around this industry for 25+ years. Nothing more. I have seen a lot of different ideas on cooling PCs. A PC inside a freezer is nothing new. One that is user friendly, doesnt exceed the cost of the system it is cooling, efficient, not a colossal, and cost effective, is new.

"I have been stress testing the system with benchmarks all week."
You never mentioned this testing before, or I missed it. That is why I suggested a static controlled heat source inside during testing. If the inside heat source isnt static and controlled, how could you possibley have accurate data if no control numbers? There has to be a constant somewhere.

"because you don’t understand the dynamics of refrigerant adequately"
I wont lie. No, I dont have the refrigerant dynamics understanding you have. I do have a basic understanding though. And a little knowledge of physics. Also, I have been around this industry long enough to see and know what works and what doesnt. You arent the first to try this nor will you be the last. I am sure some of those who have attempted this, have your level of knowledge and some greater.

"-5c under heavy load was achieved day 1 lol"
I was suggesting for integrity, the inside temp cant vary more than +/- 5C. Not maintain -5C. A variation of 5 degrees at sub zero is huge when overclocking. This I know for fact. Binderdundat.

"I can tell from your last two openers that you probably get a lot of kick back from your posts on forums."
History will tell. Yes, I have been somewhat of an *** in some forums in the past. But in this case, I am simply offering my observations. Is your ego so big that you beleive you are on a level above all and nobody can offer any insight?


"It’s much easier to ask questions or ask how something works as opposed to making assumptions with incomplete data."
If I had a question, I would ask. As I said, I simply offered observations from a different view. Take it as you wish. Dont get too high on yourself.

See, historically I have been an ***. People change. I managed to get through this reply civilly. I am kinda proud of myself. LOL

No more comments from here. Good luck.

I appreciate your civil reply, I will untangle my panties at my first opportunity. I do get irritable when people make assumptions with incomplete data. It’s an easy path to misunderstanding, miscommunication and typically, failure to execute.

The box holds 0°f to -15°f very well under load. Games, benchmarks, VR, mining, etc. I tested all of that before I brought pics and build info here.

Under -15f it struggles to maintain temp and I have been testing cold storage to push temps lower for longer with fewer run times. This has proven problematic under -20f so far. But, testing is not complete and I have confidence 👍

I apologize if I appear to have an ego issue, I do not intend to come off that way. It’s likely because the refrigeration dynamics we are talking about are things that I discuss everyday.

My background, before HVAC controls and refrigeration near 20 years ago now was in PC hardware engineering, so, I’m not exactly new to the PC world.

I will assure you,I’m not going to have any world record benchmarks or overclocks with this box, and won’t be trying. What I will have, is something cool to engage my newer techs and teach them about the dynamics of refrigerant and how to use it effectively. And then they can play games on it. Its sure to pique Some interest.

 

[karma0013]

Updated the sensor panel to fit the screen.

After watching the system with a thermal camera for a few min, I positioned the kracken fans to push air down into the refrigerant coil and the other 2 fans to pull air up from below to circulate over the top and across the cpu.

Anyways, I think I’m done playing with it for the night. I’ll run some tests on it tomorrow.

 

[Logic11]

Nice!


I'm wondering if you aren't freezing the cooling liquid in the CPU cooler's heatpipes, leading to high temps!??
Sounds crazy right! Bear with me:

Heatpipes work be boiling off liquid at the heat source (CPU)
The steam then flows to the cold side due to the suction caused by its condensing there, then the liquid flows back to the heat source by capillary action.
But
If the cold side is too cold; all the liquid will freeze there and capillary action will stop.
On the hot side there simply wont be any liquid to boil off.

That would explain your observations pretty well wouldn't it..?


Maybe a stock loop (Glycol excepted) is the better option...
The water volume will add to your 'Cold Battery'...
I'd add and extra pump or 2, in series, to keep the sludge moving/ impinging on the WB base.
Perhaps water in direct contact with end part of the evap is a better option than a stock rad..?
"end part":
You want to keep the air @ a lower temp than the water, just in case of condensation on the WB...

Oh ye;
I think moving the air is more important than you realize:
You want to reduce boundary layer thickness (aerodynamics), hence the fans on CPU coolers, rather than just relying on case airflow...
(That reminds me; Try a bigger fan on the DRAM and increase TREF a lot!)

The HVAC rule of thumb is to never have a fan closer to a perpendicular to airflow surface than the fan's diameter.
That means your evap rad is a bit too close to 'the wall' for optimal airflow through it.
Lets say said air is only in contact with the evap rad for half the time, but it'll be 'in contact' twice as often...

 

[karma0013]

Nice!


I'm wondering if you aren't freezing the cooling liquid in the CPU cooler's heatpipes, leading to high temps!??
Sounds crazy right! Bear with me:

Heatpipes work be boiling off liquid at the heat source (CPU)
The steam then flows to the cold side due to the suction caused by its condensing there, then the liquid flows back to the heat source by capillary action.
But
If the cold side is too cold; all the liquid will freeze there and capillary action will stop.
On the hot side there simply wont be any liquid to boil off.

That would explain your observations pretty well wouldn't it..?


Maybe a stock loop (Glycol excepted) is the better option...
The water volume will add to your 'Cold Battery'...
I'd add and extra pump or 2, in series, to keep the sludge moving/ impinging on the WB base.
Perhaps water in direct contact with end part of the evap is a better option than a stock rad..?
"end part":
You want to keep the air @ a lower temp than the water, just in case of condensation on the WB...

Oh ye;
I think moving the air is more important than you realize:
You want to reduce boundary layer thickness (aerodynamics), hence the fans on CPU coolers, rather than just relying on case airflow...
(That reminds me; Try a bigger fan on the DRAM and increase TREF a lot!)

The HVAC rule of thumb is to never have a fan closer to a perpendicular to airflow surface than the fan's diameter.
That means your evap rad is a bit too close to 'the wall' for optimal airflow through it.
Lets say said air is only in contact with the evap rad for half the time, but it'll be 'in contact' twice as often...

While I’m waiting on my custom loop components and tinkering with a full air over system, air movement became much more important.

when I was cooling liquid, I was ok with a 5° split as it would keep up with the slow change of liquid temp.

the air over configuration required much more airflow across the Evap to maintain low temps longer.

still hit a wall at about -30c (-22f) where there simply is not enough refrigerant in the circuit to absorb the heat if the components all ramp up to 100. The txv will lose control and regain it around the -10 mark and draw the box back down to -15.

for the custom loop, I will be reducing airflow back to normal, using 3 rads and I’m planning on placing my reservoir down in the cold storage.

also ordered some noctua industrial fans for more consistent airflow once it is complete.

I’ll add some pics when I get home, but you are correct about increasing airflow with allair over system. The space behind the Evap is adequate,but I made changes that I think you’ll approve 👍 pics to come

 

[karma0013]

Some updates and pics:

for air over testing air movement across the coil needs to be equal to or greater than air moving over the heat source or remove heat at a greater rate.

Again with 404a and this setup this becomes problematic at -20f. Not enough refrigerant in the circuit to make a maintain the temp.

With liquid, the change in box temp is much slower allowing much more of the radiant heat to be absorbed by the cold storage. This is not the case with air over cooling.

this required some minor modifications to be able to achieve and maintain -20f box temps without issue.

pics from bottom up:
Minor updates to sensor panel. CPU temp is cpu package temp (it is the only one that reports negative and is typically warmer anyways) GPU temp is GPU hotspot as this is the only temp that will report in Aida under -10c and is typically 10-15c hotter as well. Some graphic updates.

Top acrylic panel has been removed from the evaporator coil. NZXT fans installed on top of coil in “push” position. Front coil fans still in “pull” position. Bottom coil acrylic panel still in place. TXV adjusted to accommodate additional airflow.

all surrounding case fans are pulling the air from below to circulate the air across the cpu/GPU.

snapshot of afterburner when I got home today. Pretty “cool” coming home to my GPU idling at -25c.

so far the box has been very reliable and has only really posed issues when pushing it beyond its expected physical limitations (-40c I bust!!!)

I didn’t get to mess with it much today so it’s on just over a day uptime with no issues. (Since I’ve built it I’m typically restarting it a couple of times a day to make changes or tweak things.) longest uptime so far. Box set to -15°f with 3° differential and has maintained that with no variance. This is to be expected as the system is only at idle most of this time. But, uptime and reliability is important if I’m going to let this thing run 24/7 so it’s a good thing to see

 

[karma0013]

A few things came in.
PWM 4 pin powered fan hub. (No longer using salvaged case fan controller to power interior fans. Also ditched some spliced wires I had hidden 😬)

external NVMe drive housing (for the SSD that froze to death, she’s working well now although currently plugged into a usb 3.0 port) I didn’t know these even existed. Pretty handy and comes with both usb-c and usb 3.0 cables.

HDMI male to female extensions (red). The HDMI by the back of the box (by itself) is from the onboard HDMI out which will be used for the sensor panel.

Display port male to female extensions (gray)

I shut the box down and opened it up to work on it for about 25 min. I took some pics of what happens when you break vacuum and expose the components to ambient air. Frost accumulated instantly on the GPU and CPU (pictured above). Sealed everything back up. Pulled the box back into vacuum and started it up no issues.

The ssd drive on the video input side of the box will eventually move once I decide on a thunderbolt hub that suits the box best.

Also, having a hard time choosing a pump/ pumps for the custom loop setup. Just not very familiar with the different brands and reliability. I’m looking for 2 D5 PWM pumps that will be reliable in -30c environment. Any recommendation’s or suggestions are appreciated, especially if you run your pumps below freezing, let me know if you’ve had any issues.

 

[karma0013]

Temp sensors came in. I plugged them into w_in and w_out headers on the motherboard. I placed one near the return air of the evaporator coil (box temp) and one near the supply fans on the evaporator For a Delta temp.

Bios was able to read them in the negative no issue. After some frustration I found that asus’ AI suite will read these sensors in the negative. so now I have a software reference of the delta temp in the box.

Currently running Xmp II with core ratio on auto. CPU-x reference vs 10900k. The difference isn’t huge but it’s a reference point to how much better the cpu performs at stock settings in the box.

 

[Logic11]

Looking good!

I shoulda warned you the SSD would die.
They like to be hot while writing data and cold while storing it.
Most practical is to move the ctlr chip heat to the NAND with a heat pipe, then cool that end of the pipe with fins.

You want to go server U.2 as USB ads a lot of latency which kills random 4K performance.
R4K is over 60% of Windows I/O. The large sequential #s advertisers wave around are less than 1% of Windows I/O.

So you want a M.2 to U.2 adapter and cable, then U.2 to M.2 adapter to put the SSD outside the cold box.

If you stay with USB yo want to go UASP.
The ctlr chip in the external enclosure needs to be UASP capable.
Yours may be already.
You can check for SCSI driver in Device Manager.


DRAM Refresh:
As DRAM is a grid of nano capacitors, like normal caps, they discharge MUCH slower at low temps.
ie: DRAM remains usable for much longer before the data needs refreshing where its unusable for the duration of the refresh.
That means you can get a LARGE perf boost by increasing the intervals between refreshes.
Used to be called TREF. It's called something else (similar) now.

Charge times will also be a bit lower, so you should be able to decrease most all those.

DRAM termination values:
My thinking is that cold should decrease those termination resistances, leading to signal bounce errors.
So perhaps increase those if you go all out on the DRAM tweaking. (worth it @ low temps: BIG perf increase)