Ethanol cooling a ghetto loop.

[GruntBlender]

I updated my cooling setup calculator. If anyone wants to test it by comparing its predictions with their setup, that'd be really cool.
Github page: github.com/novikovech/liquid-cooling-calc
Download link: github.com/novikovech/liquid-cooling-calc/archive/master.zip

 

[Karadjgne]

My issue is that ethanol is highly corrosive to rubber, aluminium and plastic. So much so that standard gasoline cars cannot use ethanol gas (just a 10% mix). To get a flex fuel car, all the standard gas lines need to be replaced with stainless steel tubing. The insides, diaphragm, seals of the pumps are doubtfully rated to handle pure ethanol, yet distilled water is dirt cheap. I like the idea of experiments, but the hazards of an ethanol system seem to outweigh the benefits.

 

[GruntBlender]

My issue is that ethanol is highly corrosive to rubber, aluminium and plastic. So much so that standard gasoline cars cannot use ethanol gas (just a 10% mix). To get a flex fuel car, all the standard gas lines need to be replaced with stainless steel tubing. The insides, diaphragm, seals of the pumps are doubtfully rated to handle pure ethanol, yet distilled water is dirt cheap. I like the idea of experiments, but the hazards of an ethanol system seem to outweigh the benefits.

Ethanol is not that bad. It's bad for SOME plastics and rubbers. Aluminium is just fine with it. So is PVC, PET, and whatever else the pump is made from. The seal in the pump MIGHT be a problem in the long run, but if it leaks, it's not really an issue. Either way, if my system fails, I'll let everyone know and combined knowledge will rise.

 

[GruntBlender]

Tested some components.
Rated pump performance:
240 l/h
3m head
Measured pump performance:
190l/h
1.8m head

Water blocks:
at 156 l/h flow, the pressure difference was approximately 0.39mH2O
That's on par with lowest restriction name brands, but leaves me worried for the thermal performance. I'll test that later if I can.

Radiator:
160 l/h, around 0.33mH2O pressure. A bit dubious because that's quite close to the water block result.

Flow was measured, pump flow and pressure were measured , block and radiator pressures were calculated from assumed P-Q pump curve.

Flow was measured by timing the filling of a 2l vessel.
Pump head was measured by measuring volume compression of a section of air in a closed pipe.

 

[the nerd 389]

The difference in density will reduce the output capacity of centrifugal pumps like the DDC and D5.

If you take the density of ethanol (0.8 kg/L) vs water (1.0 kg/L), pressure should be about 3m * (0.8/1.0)^2 or 1.94 m. Flow will be 240 * (0.8/1.0) or 193 l/h.

That's remarkably consistent with your measurements. If you put another pump in series, you can get the pressure up to about 2.75 m in theory. In practice, it will be closer to 2.55 m.

 

[GruntBlender]

The difference in density will reduce the output capacity of centrifugal pumps like the DDC and D5.

If you take the density of ethanol (0.8 kg/L) vs water (1.0 kg/L), pressure should be about 3m * (0.8/1.0)^2 or 1.94 m. Flow will be 240 * (0.8/1.0) or 193 l/h.

That's remarkably consistent with your measurements. If you put another pump in series, you can get the pressure up to about 2.75 m in theory. In practice, it will be closer to 2.55 m.

Forgot to mention, measurements were done using tap water. I'll look into how much the density and viscosity affects flow rates, thanks for reminding me.

 

[the nerd 389]

My issue is that ethanol is highly corrosive to rubber, aluminium and plastic. So much so that standard gasoline cars cannot use ethanol gas (just a 10% mix). To get a flex fuel car, all the standard gas lines need to be replaced with stainless steel tubing. The insides, diaphragm, seals of the pumps are doubtfully rated to handle pure ethanol, yet distilled water is dirt cheap.

There are two reasons cars can't run on pure ethanol. The first is that ethanol has very different flammability limits, and in order to burn properly, the computer must supply significantly more fuel to the engine than it does with gasoline. While some engines can technically run on ethanol, the air-fuel ratio will be extremely lean. If the car uses a mechanical throttle, this results in stalling when the accelerator is pressed. It also results in drastically reduced power, higher NOX emissions, and a number of other issues.

The second reason is that ethanol lacks the lubricating properties of gasoline. Fuel pumps are usually designed to take advantage of those properties, and ethanol can cause them to wear out rapidly. That's why E85 fuel exists. It's mostly ethanol, but has enough gasoline to properly lubricate the fuel pump.

Gasoline is more corrosive to plastics than ethanol. Gasoline is a non-polar solvent and readily dissolves many polymers. Alcohol, on the other hand, is a polar solvent. It doesn't dissolve most polymers, but does dissolve some fillers and smaller polymers. The two can both damage plastics, but they do so in very different ways. Generally, gasoline will cause damage more quickly than gasoline.

Finally, stainless steel fuel lines are not necessary for E85 vehicles. Usually, stainless steel fuel lines are used to increase the stiffness of the fuel system in an effort to allow the computer to change fuel pressures more quickly, at least with non-DI fuel systems. On DI systems, the lines are used to allow the much higher fuel pressures necessary to inject fuel into an already pressurized cylinder.

 

[GruntBlender]

Did a thermal conductivity test on the water blocks using tap water. Preliminary results indicate C/W between 0.2(not great) and 0.27(not acceptable). I have yet to account for several factors of the testing apparatus and expect this value to drop.

Edit 1:
Test 1:
C/W min 0.191
C/W max 0.273
C/W mean 0.232
Test 2:
C/W min 0.117
C/W max 0.190
C/W mean 0.153
Test 3:
C/W min 0.187
C/W max 0.238
C/W mean 0.212
All tests average: 0.199

Looks like the C/W for the block is hovering around 0.2. That's a lot worse than name brand blocks which can go as low as 0.11, but it's acceptable(barely) for my purpose, and for $3, not bad at all

 

[tom42]

Having just upgraded to an R9 290, my PC sounds like a jet, to the point where people behind walls are complaining. This won't do, so I'm going to to liquid cool the beast to reduce noise (and hopefully temperature.)
Now, I'm an idiot, engineer, genius, scientist, with a controversial dash of arrogance, so I decided to ignore common practice. Being poor, the loop will have the cheapest parts eBay has to offer (modified where necessary). Being mad, and this being my first liquid cooled setup, I've decided to go with ethanol instead of water. This means that in case of a leak, I don't have to worry about shorting, and if there's a very slow leak it'll evaporate with no problems.
I'm aware of the lower boiling point of ethanol, but if the temp gets to that anywhere in the loop, it'll mean bigger problems than just that. I'm also aware of the lower (about half) heat capacity of ethanol, and that's where I might run into problems. At worst, I'd be looking at double the DeltaT compared to water, but heat capacity isn't the only determinant of DeltaT.
I'll be using 2 240L/H pumps running in parallel to ensure redundancy in case of cheap components failing on me. 240L/H =~ 67ml/s, at least according to what's on the box. I'll be happy with 30mL from the parallel setup. Now, I'm planning on dumping around 300W through this system, so I'll have around 10J at most being transferred by each ml. If that were the only factor, that would give me around 3 degrees DeltaT with water and around 5 with ethanol. Not bad either way.
Ethanol also has about a quarter the thermal conductivity of water, but I'm hoping forced convection plays a much larger role in this sort of system than conduction does. With any luck I'll only see a few degrees of difference between what this system does and what it could do with water. Time will tell (eBay orders can take a month to arrive here).

2x 240L/H, 3m max head, water pumps running parallel for redundancy.
2x 40x40mm, $5 aluminium blocks for CPU & GPU, probably have to sand the bottom to get good contact.
200ml acrylic reservoir (easy to change if ethanol misbehaves with it chemically)
Silicone tubing, PVC fittings, other assorted junk.
120mm radiator (I know, I know. If it tests to be insufficient I have an old ATF cooler radiator lying about as backup).

Ethanol pros:
- No corrosion
- Non-conductive
- Quick evaporation
Cons:
- Low heat capacity
- Low conductivity
- Potential chemical interaction with acrylic components

I guess I'm posting this so you all can laugh at my naivete, and in hopes that someone might have experience with this sort of heresy and could give me a bit of advice or point out what I've missed.

 

[tom42]

Ethanol is also highly flammable, with an almost invisible flame.

 

[tom42]

Ethanol sounds like a bit of a scary proposition due to flammability. Parts on a pc may not be 300c but there is a chance of spark. Not to mention the gpu backplate can often get pretty hot and it's right below the cpu waterblock. I can appreciate the desire of avoiding short risks with water though distilled water is pretty inert in terms of shorting. Still not a good idea to have leaks regardless.

There may be a corrosion risk of the tubing or pump internals running ethanol if they're not designed to handle it, eating away at the tubing (which would promote leaking) or damaging pump seals and causing premature pump failure.

Even if the pc itself may not be a risk, if anything else were to somehow cause a flame or catch fire and move to the pc, igniting the ethanol could be a bad day. Fingers crossed everything works out well, I just have safety concerns. Even without catching fire, ethanol boils at less than 80c which can be a somewhat common component temp. If it's a sealed system, what's going to happen when the ethanol begins to boil and pressure builds in the system? Even if it doesn't fail with a 'bang' the increased pressure is going to strain the fittings, pump, rad etc.

Just further things to consider.

 

[tom42]

Is straight ethylene glycol ( automotive anti-freeze) conductive?

 

[Karadjgne]

I would have to say that it was since antifreeze isn't pure, it does contain water, diethylene glycol and other anti-corrosion stuff. And blue dye #3 and yellow #6 naturally

 

[GruntBlender]

Almost done. imgur.com/gallery/BmxC1

 

[GruntBlender]

Ran a quick test of the completed system. GPU got to around 87C and CPU to 57C at 18C ambient. Idles are GPU at 35C, CPU at 25C. Didn't run the test long enough to see stability, VRAM got a bit too hot for my liking and I'll outfit some airflow through the case before trying again. Also only one of the fans on the rad is currently on, and even that is running at under 700 RPM for some stupid reason. I'll fix these issues and report back with images and data. It's late, so tomorrow at best. Also I shorted out my graphics VRMs, but that's another story. Good night.

Edit:
The story with the short is that the body of the VRM is actually conductive, and my heatsink connected all the modules together. There was smoke, I'm hoping the smoke was from the tape I used to insulate the chokes from the heatsink. Potentially, one of the GPU phases of VRM1 may be gone, I won't know till later. Hoping against hope, I took off the heatsink and tried the card again (there was no visible damage). It worked, so the card is still functional. I repurposed some of the VRAM thermal pads from the old reference cooler to insulate the VRMs, and it's working. Under benchmark, the VRMs got to 102C before I stopped ​the test. Hopefully that will come down a little when I have some airflow over the VRAM and VRM heatsinks.
Incidentally, does anyone know how to get the VRAM temperature of the 290? GPU-Z is only showing core and VRMs.
Cheers.

 

[GruntBlender]

OK. I did a 15 min burn in on the GPU today. Got some temps and pictures.

Pictures : imgur.com/a/noZzY
Ambient temp was 21C

Idle: (T1)
CPU: 21C
GPU: 30C
VRM1: 24C
VRM2: 30C

GPU burn in with FurMark: (T2)
CPU: 41C
GPU: 90C
VRM1: 99C
VRM2: 71C

Both GPU and CPU stressed to max: (T3)
CPU: 62C
GPU: 90C
VRM1: 99C
VRM2: 71C

Based on comparing T1 and T2, where the CPU idles in both cases, I'd say my coolant gets to around 20C as expected. Getting a bigger radiator would shave about 10C off the load temps.
CPU power usage at load is ~23W and <1W at idle, with a temperature difference of 21C. That implies the waterblocks are around 1C/W.
GPU put out anything from 150-200W with a delta T to coolant of just under 50C, giving a C/W of between 0.33 and 0.25. I wonder why the two are so different, the blocks are identical...
Anyway, the acrylic reservoir seems to be holding fine with ethanol. No leaks, no cracks, no tubing problems.

I noticed the VRAM heatsinks were a bit too hot to touch. 14x14x6mm is what I used. I have a single 120mm fan blowing on the card from the VRM side, but it doesn't seem enough. I might make custom blocks or bigger sinks for them, or just add another fan to blow more directly on them.

The pumps are a bit louder than I expected. Not really ideal, but not bad for the price. They'll do for now but I'll continue looking for super cheap pumps, maybe make my own. I might get away with submerging them in a soundproof box... Thoughts for later.


Comments/Questions/Insults? (Other than telling me to switch to water)

Edit: pumps were too loud so I switched them to 5V. Doesn't appear to affect temperatures that much.

 

[GruntBlender]

Approximate costs for the project, that is price of some parts:

120mm LED fan $5.71
120mm Fan $3.87
20pc 14mm heatsinks $5.76
100x41 heatsink that became the VRM sinks $2.73
Thermal Grease $1.38
Silicon Thermal Adhesive $1.71
Water Reservoir $9.52
Alum. Water Blocks (Pair) $9.86
G1/4 120mm Radiator $14.78
ABS 12V Pump (Pair) $14.34 (half of that was shipping)
And a few other bits and bobs for the plumbing.
Total
80.12 NZD
55.28 USD

If I were doing this again, I'd opt for a slightly dearer, quieter pump, a copper block for the GPU, and a 240mm radiator. These upgrades are planned for the future. In the mean time, a working loop of questionable performance for 55USD? Fine by me.
Things I already had: Hose, ethanol, metal and spokes for the mounting plates, another fan, screws, etc.