Oil submersion optimization

[ivanlabrie]

It's definitely an interesting proposition...gonna be fun to watch.

 

[Black C5 Z06]

Oil immersion is feasible albeit not practical; power utilities have been using oil immersed transformers for decades. Some considerations would include:

1. The oil used: You would need to use an oil that won't retain water easily. Power company transformers use a special oil that doesn't absorb water as readily as most oils. Since the smallest quantity of that oil I’ve ever seen is in 55 GL drums (I used to work in the industry), you might have a bit of trouble finding it. A possible source might be your friendly neighborhood power company since you probably won’t need more than a gallon or two.

2. You will need to have a closed system to exclude contaminants, including moisture.

3. You won’t be able to use computer fans to circulate the oil. Oil, no matter how light (transformer oil is almost as light as kerosene), is still far more viscous than air. The increased resistance on the impellers will burn up a computer fan in nothing flat.

4. You will need to use the biggest, baddest heat sink you can find to cool the CPU since you will not be able to use fans to move oil through it. Same for cooling GPUs since their fans will be useless. An option might be to use water cooling for the CPU and GPUs independent of the oil bath.

5. Since you will need to depend on convection to establish circulation (oil pumps would be a royal pain to implement), you would probably be better off mounting the MOBO vertically so you can get more depth for your oil bath and to get oil movement over the board from convection. For the same reason, the big, bad CPU heat sink should be oriented for vertical circulation.

6. Current HDDs are vented and need air inside to operate (oil inside create too much drag on the platters, heads, etc.) so they will need to be mounted outside the oil bath. The proposed helium filled HDDs might be sealed well enough to withstand immersion. SSDs might also be able to withstand immersion.

Again, please watch the first video, and atleast a few minutes of the second video I linked to earlier in the thread. Some of these claims are simply not true, as evidenced in these videos.

 

[Anonaru]

Again, please watch the first video, and atleast a few minutes of the second video I linked to earlier in the thread. Some of these claims are simply not true, as evidenced in these videos.

I think I put the fears of that quoted post to rest by saying a stock intel HSF and fan ran and kept the CPU very cool for years of 24/7 demonstration use in an aquarium tank in my shop haha

 

[Black C5 Z06]

I think I put the fears of that quoted post to rest by saying a stock intel HSF and fan ran and kept the CPU very cool for years of 24/7 demonstration use in an aquarium tank in my shop haha

Maybe he just missed your post. Because, yes, I would have thought that would have put them to rest as well.

 

[pinky33]

I have built a few oil cooled computers and I have yet to see a fan last longer then a month.

I also feel it is only good for the cool/neat factor.

 

[Anonaru]

I have built a few oil cooled computers and I have yet to see a fan last longer then a month.

I also feel it is only good for the cool/neat factor.

I must have been lucky

Really, I built mine for this reason-- and because I wanted to see how well it worked. It seems like an amazing cooling solution, but it loses its effective edge after the oil starts warming up. Even on the low-powered build I had, it wasn't very easy to keep the oil cool. Maybe because I used an aquarium tank

I wonder if using a huge, low RPM fan and a custom-made HS with direct contact to the oil could make for a possible solution-- shrug

 

[bluezero5]

I had several experiments setup submerged in liquid, (non-computer, but scientific)

a few concerns I will raise:

1, Fan will not be very good at keeping flow, you might move the oil near the fan, but the will stop after a few inches. To keep it flowing, you need near/lower than water viscosity. and follow convection patterns, meaning, pushing UP through the heatsink will give you a lot more flow. That being said, use a PUMP instead, PUMPs are designed to move liquid.

2, many mineral oils have surprisely low specific heat, meaning it heats up VERY fast, and the heat can be quite localized,

3, SSD 'should' work, HDD won't for sure.

4, I will use a 'deep container, and submerge 'deeper'. As heat reduces the density of the mineral oil, it will raise, so a deep container will naturally give you better convection.

GOOD LUCK, keep up posted.

 

[Anonaru]

I had several experiments setup submerged in liquid, (non-computer, but scientific)

a few concerns I will raise:

1, Fan will not be very good at keeping flow, you might move the oil near the fan, but the will stop after a few inches. To keep it flowing, you need near/lower than water viscosity. and follow convection patterns, meaning, pushing UP through the heatsink will give you a lot more flow. That being said, use a PUMP instead, PUMPs are designed to move liquid.

2, many mineral oils have surprisely low specific heat, meaning it heats up VERY fast, and the heat can be quite localized,

3, SSD 'should' work, HDD won't for sure.

4, I will use a 'deep container, and submerge 'deeper'. As heat reduces the density of the mineral oil, it will raise, so a deep container will naturally give you better convection.

GOOD LUCK, keep up posted.

This is golden. I used fans because it was a basic project-- I'm sure there's plenty of pumps that are inexpensive / will get the job done.

HDDs aren't airtight.. There's a few exceptions to this, but even then I'm not sure if they need to have some interaction with outside air. SSDs should be completely fine, I can't imagine them not working properly for any reason

 

[ivanlabrie]

But also, ssd's don't need cooling. I think you'd be better off leaving the hdd and ssd in the bottom tray and have a big slow fan blow air accross them (hdd's do heat up)

 

[EarthDog]

I wouldnt worry about cooling your HDD directly either... I have never done that.

 

[wagex]

It would be nothing like that at all. The engine (more or less) would not be damaged at all.
If using an automatic, the torque converter would be the first to go. It would burn up because you'd be spinning the engine at maximum speed while the transmission is barely moving, therefore heating up the fluid in the torque converter until the TC failed.
If using a manual, the half shafts, driveshaft, clutch, or transmission would fail, which ever had the lowest manufacturing specs and depending how you tried to drive it (slipping the clutch, dumping the clutch, using too high a gear, etc). Again, this is all because of mechanical interactions.
The engine, on the other hand, as long as it has sufficient water and oil cooling, will be perfectly fine.

But I was just trying to relate to the fan. I'm assuming the fan wants to get to a certain voltage drop and current draw. I can't imagine why it would try to "ramp up." It has no micro-controller to try and get it to a certain speed. Fans don't control their speed. Either your controller or the mobo does by regulating voltage (3 pin) or switching the circuit on or off (4 pin PWM). Current is steady, that's built into the hub circuitry.


Yes, I already linked to both videos. One of Linus showing Slick's PC, and the other with Slick explaining his PC some.

it does not take that much to pull a semi lol trust me ive done it with a lawnmower(had to resort to that to pull one into the shop luckily our lot is pretty flat),it wouldnt snap anything in the drive line. i was a mechanic from halfway through my sr year of high school till i landed the it job ive only had for almost a year,and i still go in the shop here and do stuff on occasion.

the car could pull it, no busted shafts or other funzies. would take it a long time to get up to speed if it could even if it was only 20mph lol but the stress on the engine maintaining speed with that much of a load on it with alot more friction and not enough air flow over the radiator (even with fans) would indeed make it overheat, causing warped heads, blown headgaskets, all that fun stuff. even pickups have towing packages for a reason not only do they get the receiver in the back, but they get a transmission cooler, higher capacity radiator and a beefier suspension. ive seen a pickup overheat just from pulling another truck (alebit with a flat tire[dont ask.. drunk rednecks are dumb sometimes]) we had to stop because his tranny started burping tranny fluid out of the vent tube, from not being able to go very fast, his engine was pretty warm too. but hey i dont know anything about vehicles and this is a computer discussion so who cares.


and why do fans pull almost double the amperage when they are spinning up? i didnt say they had a microctroller or anything that controlled them they are going to have. any electric motor if you try to slow it down it will get warm/hot and pull more amperage. either way resistance = heat+stress

lots of information about how loads on electric motors cause them to pull more amperage. lots of info

i can see why its understandable why they could fail eventually, although
http://www.pugetsystems.com/submerged.php
the faq's show they have had fans running for years submerged and they are just fine.

i wasnt saying they would certainly fail, i could understand what earthdog was saying though, kind of made me wonder the same. we can talk theoretical all day but when it comes down to it results are the only thing that matter i guess lol.

 

[Black C5 Z06]

it does not take that much to pull a semi lol trust me ive done it with a lawnmower(had to resort to that to pull one into the shop luckily our lot is pretty flat),it wouldnt snap anything in the drive line. i was a mechanic from halfway through my sr year of high school till i landed the it job ive only had for almost a year,and i still go in the shop here and do stuff on occasion.

the car could pull it, no busted shafts or other funzies. would take it a long time to get up to speed if it could even if it was only 20mph lol but the stress on the engine maintaining speed with that much of a load on it with alot more friction and not enough air flow over the radiator (even with fans) would indeed make it overheat, causing warped heads, blown headgaskets, all that fun stuff. even pickups have towing packages for a reason not only do they get the receiver in the back, but they get a transmission cooler, higher capacity radiator and a beefier suspension. ive seen a pickup overheat just from pulling another truck (alebit with a flat tire[dont ask.. drunk rednecks are dumb sometimes]) we had to stop because his tranny started burping tranny fluid out of the vent tube, from not being able to go very fast, his engine was pretty warm too. but hey i dont know anything about vehicles and this is a computer discussion so who cares.


and why do fans pull almost double the amperage when they are spinning up? i didnt say they had a microctroller or anything that controlled them they are going to have. any electric motor if you try to slow it down it will get warm/hot and pull more amperage. either way resistance = heat+stress

lots of information about how loads on electric motors cause them to pull more amperage. lots of info

i can see why its understandable why they could fail eventually, although
http://www.pugetsystems.com/submerged.php
the faq's show they have had fans running for years submerged and they are just fine.

i wasnt saying they would certainly fail, i could understand what earthdog was saying though, kind of made me wonder the same. we can talk theoretical all day but when it comes down to it results are the only thing that matter.

Point taken. I was just kind of thinking out loud. Like I said, it's just my thought process.
There's a reason they call this a *discussion* forum

 

[wagex]

Point taken. I was just kind of thinking out loud. Like I said, it's just my thought process.
There's a reason they call this a *discussion* forum

it is all good it happens to me too.
apologies if i was too blunt as well.

 

[ivanlabrie]

I still think the fans would fail at moving oil, and op should be looking at some kind of pump.
perhaps he can use some sections of oil connected by tubes and each section could be a radiator of its own.

 

[wagex]

I still think the fans would fail at moving oil, and op should be looking at some kind of pump.
perhaps he can use some sections of oil connected by tubes and each section could be a radiator of its own.

i agree with this 100% weather the fans failed or not i do wonder how well they would even move any fluid let alone oil such a viscosity it has, and agree a pump would be able move it better and more acurately as well since you can put tubes anywhere although care with the tubing you use because most tubing will be dissolved by oil / hardened. tygon makes some very good chemical resistant tubing that can even handle battery acid

 

[bluezero5]

I think a simply directional flow like this might be easy to set up.
let the tubing to the work, and fasten the tubing using a metal rig so the tubing won't be jetting around.



just thought of something positive....

since air and condensation is not involved here...
why not run the radiator through the AC or something.
Just add insulating tubing from the radiator tube to the submersion, and then no worries about water condensing.

potential result: sub-ambient oil flowing through the CPU heatsink.

 

[ivanlabrie]

Bluezero5, you're the man
that actually sounds GREAT.

 

[Anonaru]

I think a simply directional flow like this might be easy to set up.
let the tubing to the work, and fasten the tubing using a metal rig so the tubing won't be jetting around.

View attachment 121333

just thought of something positive....

since air and condensation is not involved here...
why not run the radiator through the AC or something.
Just add insulating tubing from the radiator tube to the submersion, and then no worries about water condensing.

potential result: sub-ambient oil flowing through the CPU heatsink.

Genius! Strong @$%# pump made me chuckle

 

[EarthDog]

Yaaaaaaaaaaaaaaaaaaaaawn... someone PM me when this gets built, K?

 

[abarth_1200]

How about using a water pond pump or even the pumps used in water cooling loops to move the oil instead, you could even route some tubing to suck in cooled oil within the tank and expel it directly into the cpu heatsink etc, this way will keep the system fan free.

I also dont think you will need the four 200mm fans.

This looks like a well thought out and interesting project and I would love to see it go ahead, good luck to you.