Scratch Build – In Progress The Mineral Oil PC - 3/24/2011 - More leaks :(

I realize that my temps will be a bit higher than air cooling. In air cooling, heat is removed from the CPU with air that's ~20-25C. My oil is 30C-35C, so I'm already 5-10C higher.

But turning on the stress test in prime95 instantly pegs the CPU to almost 95C, it doesn't take 10 minutes, hell, it doesn't take 10 seconds; it's 95C instantly. In a cross fire setup, the slave card is clocked way down when not in a game. It idles at 157MHz at 75C. I simply don't understand.

As for the bubbles, yes I've seen something about that as well! I might have to give that a shot to help lower my temps if maybe reseating the HSF doesn't fix anything.

Word of advice: stay the hell away from PVC. And watch out for rubber bases on your caps! Good luck

 

The temps might be because the oil is dissolving the TIM.

Maybe you could try something like these metal pads. Its a solid piece of metal so it won't dissolve in the oil.

Just a thought.

 

I rewatched some of your videos and it looks like you have the PSU fan facing the CPU heatsink from the right side. Every other site I found did that too but they generally had the PSU facing up from the bottom of the rig. Heat rises in air, but what about oil? I...guess I could try Google real quick. >_> (10 minutes later) Well, there wasn't a whole lot about that but a few of the journal abstracts I found implied that, while not as drastic as air heat transfer, oil should still be cooler at the bottom of the tank.
Maybe add a fan down there to push oil up towards the CPU?

Another thing I saw was putting the outtake tube near the CPU heatsink so that the "fresh" oil from the reservoir gets pushed over the heatsink.

Your mod log is the most detailed I've found so I can't speak for if any of those suggestions actually made much difference but they seem good in theory. Most everyone else who has posted oil rigs on the interwebs just did it and said "lulz look what I did!!1!" without actually explaining why they made certain changes.


And yeah, I already learned the PVC problem on another site, but thanks for the caps tip! No one else seemed to have that problem documented. I might have to break down and get that MSI board when the time comes. Hopefully it'll be cheaper.

 

Hm.. interesting. However, I think something like this: http://www.skinneelabs.com/indigoxtreme.html would have been ideal. I wish I had learned about it before I dunked everything in oil the first time. The problem now is that I can't turn my motherboard horizontal without taking it completely out of the case and making a huge damned oil mess (same with the video cards). The have to be flat when melting the TIM.

I sealed around the GPUs and CPU with silicone after I put the TIM on it. It's possible that some of the silicone around the CPU socket actually got sandwiched between the CPU and the base of the heatsink; I had a lot of trouble getting the HSF bolted down so I may have smeared silicone everywhere. But I didn't have that problem with the GPUs, so I don't have any idea why they're so hot.

Heat rises everywhere regardless of the medium (well, as long as it's not a solid, hehe). You're correct, the CPU is pushing oil into the heatsink. I was thinking about adding a second fan on the other side of the heat sink so I'll have a push/pull configuration.

My cool oil exhaust from the radiators is directly into the CPU heat sink. The hot oil intake is near the top of the tank right next to the PSU. I was thinking about moving the hot oil intake to the GPUs, with 2 branches so each 5850 gets its hot oil removed.

But honestly, the difference between hot oil and cool oil is only 1.5C, so I don't imagine this making a huge difference.

 

That Indigo Xtreme looks awesome. I'll have to give it a try when I build mine.

Question about sealing the CPU. Did you think it was necessary or was it for safety's sake? I've read about 50/50 on whether sealing the CPU was actually needed. Most people seemed to do it just so they wouldn't lose TIM but the Indigo product would make that moot.

 

I think the problem you are having is due to you sealing the heatsink areas. I design and build workstations and servers so heat is always my demon. If you did get a full seal around the chip and none of the mineral oil can get in you have an air pocket. With no moving air it just sits there and gets hot. My two cents and all. Either way if you have an opening you have min oil near the heatbridge or if you dont you have an easy bake oven around the chips. Only way to find out is drain and play. I've also seen other tanks (oil filled) that didnt seal around the chips.

 

the TIM being dissolved shouldnt make a difference. We tested in school different items for TIM and oil did work and was not that much higher than proper TIM. infact we ran some with no TIM at all.

but if you got silicone inbetween the cpu and heat sink that could definitely cause proablems.

 

I had high temps 5 months in the past with my Q6600 in the oil (but they were tolerable). I had used no TIM, only mineral oil and I didn't seal it. I decided to try using a real TIM and sealing the socket with silicone. The temps are higher now, so I don't think sealing it was a good idea.

Yeah, I'm thinking about getting rid of the seals and seeing what happens.

 

Excellent...I can't wait for you to be my guinea pig.

You should get one of your sponsors to chip in for the Indigo. They're out of stock at Petra but hey, even I'd be willing to chip in a few bucks for my own knowledge.

 

What's the flow rate on that pump again? I forget.
But anyway, in my machine I used Arctic Silver Ceramique as the TIM on my CPU, and oddly enough there's no evidence of it dissolving in the oil.
Another thing to consider is that when there's no seal around the CPU (as is the case in my machine) the oil is contacting EVERY surface of the CPU, increasing the area of heat transfer over-all. Most of those who bother to seal around their CPUs are just paranoid about stray capacitance which, in my opinion, is unfounded. The fact that many people are running mineral oil rigs without sealing the CPUs and getting flawless operation out of them is proof enough for me.
Other than that, unless your pump is proving to be inadequate and not pushing enough oil I can't see any reason for a radiator setup like that not cooling well enough. Maybe it's all the elbows, I hear those can have a pretty big impact on flow rates. (but I've never dabbled in water-cooling so what do I know? :V)
All I can recommend is to turn it on and wait a while and try to observe the different viscosity levels of the oil as it comes in from the radiator loop. If there's a readily visible difference, then your rads are performing as they should.

 

Hahaha. The sponsors don't really work that way though; they provide specific hardware or discounts in exchange for publicity. IE, Crucial provided one of their new SSDs, PTS provided a radiator and a small discount on further purchases and Crystalfontz provided one of their little LCDs with the hardware controller.

However, if you want to chip in for the indigo, you're more than welcome

I calculated my flow rate using some fun math a few pages back:

Properties of oil:
Specific gravity: 0.9 g/cm^3
Heat capacity: 1.966 J/(K*g)

Idle power draw of computer: 270 W = 270 J/s
deltaT of oil In/Out temp: 1.94K

(270 J)/s * (cm^3)/(.9 g) * 1/(1.94 K) * (K*g)/(1.966 J)
Everything cancels but the cm^3/s and yields: 78.66 cm^3/s
I plugged that number in Wolfram alpha to convert it to gallons/hour and got 74.81gal/h. So the pump is indeed working, otherwise the oil temp would just go up and up until the machine overheated. However, my Crystalfontz Uber SCAB control unit keeps the oil below 40C by regulating the speeds of 6 fans.

What kind of temps did you see with AS5? Did you have a fan on the heatsink or was it just sitting in the oil relying on convection to remove the heat?

 

inefficient

I would agree with this ..the air trap inside is supper hot and that heat is transfered back to the proc.. with all this said about TIM and efficiency or should i say inefficiencies the TIM and the heat sink all compile to the transfer of heat to the liquid with all this touching in between going on.. is it possible and hear is question to mineral oil cool with out any heat sink at all? the way im thinking is that the entire case just became your water block.. what better way would there be to get rid of all the inefficiencies? now I no somebody will say .but you dont have any surface area.. I say you only need that becuse air is so inefficient that it needs a massive amount of surfice area or the chip would over heat before it had chance to cool.. so maybe if you directed a circulating pump right at the chip with nothing to get in the way, this would be the most efficient way of cooling possible. if I'm way off please tell me before I try it.

 

From what else I've read the only way no HSF would be viable is if you had a very strong jet of mineral oil pointed directly at the CPU. While that would theoretically work it would be a lot safer to at least have even the stock HSF on there. You would also get better temps with a HSF combined with the nearby jet.

 

I feel that that would be a very bad idea. With no heatsink, you'll be generating like 120W in an area about 4cm^2. I'd rather not ignite my oil

 

oil viscosity >> water viscosity ( a hundred time) => you need a gear pump. The one you use is designed for water.

I'm not even sure if the fan can force oil through the heatsink : fins are too tightly packed

I think you get high temps because what's eventualy cooling your harware is the exchange surface of the plexiglass of the case. It's passive cooling. Let it cool for the night, and try with the pump off to see if there's a difference.

 

Allow myself to quote myself:

It's not a question of oil temp. Besides, my radiators/pump are doing a perfectly adequate cooling job. The oil coming out of the rads is 1.5-2.0C cooler than the oil going in.

 

Mmmh tell me, when it happens, what's oil temperature ?


Oh I assume your pump takes hot oil from the top of the tank, right ?

 

Since the thermal capacity of oil is so high, raising the temp of all 8 gallons of oil 10C would take 150 watt hours of power.

Idle power usage: 280W
Prime95 Load: 360W
delta of 80W

Thus it would almost take 2 hours to raise my oil temp 10C by running prime95. Yet the CPU temp hits 95C in a matter of seconds. And correct, my hot oil intake is at the top of the tank. Believe me, my high temps have nothing to do with the pump or radiators.

 

So bubble it must be. A bubble traped around the cpu :

low cpu load, cpu is "cool" bubble doesn't interfer.
High cpu load, temp rises up, as does the fluid in the bubble. P.V=n.R.T kicks in => Vb increase => contact between cpu & HS is lost.

Or oil is fumbling a connection on the socket when cpu is hot => false reading.


Uh, I think your maths are wrong :
"Since the thermal capacity of oil is so high, raising the temp of all 8 gallons of oil 10C would take 150 watt hours of power.

Idle power usage: 280W
Prime95 Load: 360W
delta of 80W"

That's 150 W during 1 hour. If you provide 360W, then it takes 1 Hour *(150 /360) = 0.416 H = 25 minutes.

 

Whoops you're right. The 8 gallons + 10C taking 150Wh number is correct, but the 2 hours I quoted is incorrect. Assuming no heat was being removed from the tank, it'd take 25 minutes of 360w to raise the temp 10C like you said.