OK, well I guess this is one of those topics that really gets people going. I did not expect this much response this quickly, so this will be a very long update:
To start, I intend to make this a detailed study of the effect on temperature of introducing a water bath to the oil cooled computer concept. We will be starting with a fully operational system. I am donating my PVR to the cause, and will start by taking temp readings over a week long period, with various system loads, on a stock air-cooled system. We will then set up the oil cooled computer, and perform the same tasks and temp readings on a non-circulating system. We will then set up a radiator circulation system and repeat the tests. Following that, we will set the system in the water tank, without circulation, and repeat. I would have said that last we would then reconnect the radiator and check the temps, but I am now considering a final modification, and adding a water radiator, and measuring the effect of actively cooling the water through a radiator on the overall temps. We will be allowing the system to cool completely between tests, as one of the theories I want to test is whether the water bath, and the resultant Oil-Water and Water-Air heat transfer can arrest the very slow heat curve of the oil at a lower level than without it. I also may go back, and test the effect of submerging the oil cooled system when it is at full operating temp, to see if it stabilizes at a different temp than starting it from cool. This is intended to be an actual study of the effect of the water on the oil as an additional cooling medium. There have been numerous people that have done a straight oil-cooled system, and I am interested in whether this will alleviate some of the heating issues they had. On that note, here are a few of the things that are simply accepted:
- Oil is not going to cool as well as a well set up water rig. That's a given, as water has much beter thermodynamic properties for heat dissipation than oil does. However, I also should not have to flush my system every 6 months.
- Oil is messy: Actually, this isn't a given, just a factor of how much you have to operate in the oil tank. If done properly, you should have tested that your system works before submerging it, in which case, you should not need to remove it from the oil again, and therefore, this is a non-issue.
- This is a lot of work for lackluster cooling. Again, we understand there may be better ways to do cooling, but how do you think we found those? This is an academic exercise first, so a little work is expected.
Now, all of that aside, I would like to respond to some of the specific points raised, and maybe get some additional input:
Pinhedd:
Demonstrating an oil cooled PC is very cool, as would be a heat exchanger between an oil loop and an open water reservoir but after a period of time it will heat up to the point where it poses a danger to your fish
The oil cooling itself stabilizes in the 40°C range, which, while a little hot for a computer, actually remains very stable, as the oil heats all of the components on the system equally, which seems to keep it stable in what most of us would consider a high "liquid cooled" running temp. Now, that being said, I am also cognizant of the fact that the water could in fact have absolutely no effect on the temp, and as such, we are planning on having tropical fish in the tank, which are happy with temps of 80-85°F.
An aquarium that big would have a water pump. If not, the water gets too stagnant and most fish can't live in that
Also, the tank will be set up as a standard fish tank, including water circulation and air infusion. I am now considering trying an additional radiator on the water to measure its effect. That will be a purely budgetary consideration, as my DOE almost had a conniption when I gave her my first prelim budget. But, as I already own almost everything we will need to try this, I might be able to swing that part as well. I am also considering using an oversize pump for the fish tank even if we don't set up a second rad just to increase the water flow throughout the tank.
House70 -
The question would be how do you plan to circulate the water through the oil radiator. I don't believe the convection current created within the radiator fins will be powerful enough to move an adequate amount of water in order to achieve good cooling of the radiator. Water has a high specific heat and will take a long time to get a good transfer from the oil-heated radiator to it. I am afraid you will cook the CPU well before you get to cook the fish in the tank, if you see what I'm saying.
Would it be connected to a radiator, as well? Just circulating the water is good for the fish, but will not do much for the temperature of the water (unless is a thermally regulated aquarium, as well; IDK if these exist and if so, how common they are). Besides, the pump itself creates a small amount of heat that needs to be dissipated. In a fish-only scenario, the heat is dealt with by the water's cooling properties, but if you add a significant water heater like a PC in it, it might not suffice.
The radiator is intended for the oil, not the water, although that now has me thinking about a second radiator. Even without a radiator, an oil cooled system stabilizes below the failure threshold of just about all computer components (about 40°C or so), so I am not concerned with frying the system. However, I would like to have the system remain perfectly stable, and hopefully do so while overclocking it to degrees not possible otherwise. That will be one of the things we will test while doing our temp measurements. I am really hoping that the Oil-Water surface area will transfer a great deal of the heat from the oil, and that the Air-Water surface area will be sufficient to draw the majority of that heat back out of the water faster than the oil is dumping it in. I am confident that even if my theory proves incorrect, the addition of the oil radiator will ensure the oil temp remains low enough that if the oil and water stabilize at the same temp, that should only be around the 40°C level, which tropical fish are OK with. I am hoping that I am correct about the amount of thermal transfer I can achieve, however, and that I can get the system to stabilize in the 30-35°C range, which would be more that comfortable for your common goldfish.
Now, I did like your idea of using a separate radiator for the water, and as I stated above, I do intend to explore this as a possible addition. I am also wondering how much affect our AC will have on the results. The setup will be in the hallway of our school, and our students are often complaining that the school is too cold. This should help the water temperature remain lower, and I do not know how much that might skew the results from somewhere the ambient temp fluctuates up more. However, if anything, that would be a question to ponder after we get the results for what we have to work with now.
Rubix_1011 -
You are talking about dissipating several hundred watts of heat into an aquarium? I don't foresee how your fish are going to like this unless it's a tropical tank and the fish like very warm water temps.
Many people have done oil cooling and it's not as great as they originally thought. Oil isn't a very good medium when trying to cool computer components...specific heat of mineral oil is about 60% lower than that of water. It takes longer to absorb heat and takes longer to dissipate heat, making it a poor conductor or coolant.
For that reason, we are planning to start with tropical fish, unless our tests show we get a significant drop in the stable temp from what other OCCs have shown. I am actually counting on the long heating curve of the oil to help in keeping the overall temp lower. I am hoping the oil cannot absorb heat faster than the water can remove it and dump it. But, that is why we are trying this. Hopefully I am right, and we will end up with a much lower stable operating temp.
I'm not sure I follow your complete workflow, but trying to submerge an acrylic case into water is just asking for leaks. I'm not sure how good you are with acrylic, but of everyone I've read about who's tried this, the acrylic always finds a way to leak massive amounts of mineral oil...one guy had his apartment floor flooded with it. You can definitely try this, but 2 things are bound to happen, here: 1. the oil leaks out, meaning that medium that is acting as your coolant interface is now gone, or less effective depending on how much is lost. 2. You have the entire enclosure under water, meaning now water is displacing the mineral oil. I doubt you are using 100% distilled or anything that is non-conductive, and even if it was, as soon as the water is exposed to anything, it starts becoming conductive from particles being introduced. This water now finds it's way onto electrically charged component PCB, and causes shorts.
If you take the time ans attention to seal your acrylic correctly, there should be no reason you cannot produce a perfectly oil tight case. After all, all of the joints will be sealed with silicone, and neither mineral oil or water is reactive to silicone. That's why they use it to seal fish tanks. Most of the OCC mods I have seen have been built using hand-made Acrylic tanks. Puget has a video showing their tank build, and that system has run for five years without needing to be topped off, and without a leak.
I've never understood the appeal of mineral oil submersion...it's very impractical, provides lackluster cooling performance and makes a large mess. I can understand the science project concept, but even then I don't completely follow what the demonstration would help prove. I would really like to know the questions you are trying to answer or disprove with this, but I don't know that right-off I can see any that are relevant.
I've seen a lot of watercooling and a lot of mineral oil submersion on many different forums and when it comes to oil submersion, it is always from people who have found those Puget computers they made or they've just Googled their way onto something similar.
I'm not saying it isn't possible, I just think it's far more work and effort into something that will perform less than a normal watercooling loop on its own
I trust I answered this one above in the beginning. While oil cools less than water, it allows greater stability in a system at higher temps, because it heats the entire system uniformly. Most systems become thermally unstable because one component is dramatically hotter than the components working with it, and therefore it's electrical properties will be significantly different, or it just becomes too hot to perform properly at all. If all components are around the same temp, then their elec. properties should be the same. That being said, I am hoping that the water will introduce a dramatic temp decrease, while maintaining all of the advantages the oil provides.
Maxinexus -
Sorry to give it to you but the whole thing is just big waste of time. As cool as it may seems it is not practical to do oil because it will make upgrades almost impossible and painful for the owner and not to mention smell and maintenance. There is enough work and maintenance to do on simple water cooling loop.
OK. a few things. First, Mineral Oil is odorless. Second, there is no water loop. I don't expect to have to maintain the water in any way I would not have to for owning a fish tank. and lastly, I will address the whole "hard to upgrade" argument that always seems to come up in these discussions. Other than the time you need to give the base tray to fully drain when removing it from the oil tank, and the fact that you will have to work on it on an absorbent surface, this system is no harder to upgrade than any other. When is the last time you purchased a faster processor (ONLY) for your computer? 99% of all computer upgrades involve the replacement of the MB/Proc/Memory bundle, because you are usually getting a new MB that won't work with your old processor or memory. Yes, the old stuff will have Min. Oil on it forever probably, but if you are upgrading, then you are probably planning to retire the old ones anyway. And if not, then you can consider the time required to alcohol the whole thing clean in your upgrade considerations. I intend to make the base tray accept a standard ATX size motherboard, and have mounting for a standard power supply. If I really need to upgrade, it will not be "difficult" but rather will mean complete replacement, and probable retirement of the existing parts. Oh, and by the way, I AM the owner, so I am accepting this as a given ahead of time.
I do appreciate all of your input. Please, review the original post, and the replies, and if you have any other insight, it is welcome and appreciated.
Thanks