Passively cooled fish tank computer

[linuxfan66]

Once finished, I will cool 500 watts passive. It's possible, but I thougth your idea was to use the big tank, so you deal with more heat than you dissipiate at a given time....

a hidden quiet radiator could fulfil as passive as possible that said

[Stefan1024]

a hidden quiet radiator could fulfil as passive as possible that said

For this build yes.

Personally I don't use water, so I can't bridge a big distance between the components and the heat sink.

[FakeGamerGuy]

So I found this video while researching passive coolers. Looks to have much more surface area than the tank does, much better heat transfer, and it's vertical to promote airflow. Even then he's getting low 70s @ 83.6W to the CPU. My lovely AMD chip is a 140W part. If I added the radiator he's using to my setup (which is actually 2 radiators strung together) I'd likely get a similar result with no option to cool my GPUs.

So I can still run some more tests to find the heat radiated with no lid or with air moving across the tank or whatever but it's looking more and more like I should invest in standard radiators (maybe a few of them) with low speed fans. The good news is, if I do that, I'll need to get water blocks for my parts anyway so I may as well run the loop and watch the results for the sake of science.

[Stefan1024]

So I found this video while researching passive coolers. Looks to have much more surface area than the tank does, much better heat transfer, and it's vertical to promote airflow. Even then he's getting low 70s @ 83.6W to the CPU. My lovely AMD chip is a 140W part. If I added the radiator he's using to my setup (which is actually 2 radiators strung together) I'd likely get a similar result with no option to cool my GPUs.

So I can still run some more tests to find the heat radiated with no lid or with air moving across the tank or whatever but it's looking more and more like I should invest in standard radiators (maybe a few of them) with low speed fans. The good news is, if I do that, I'll need to get water blocks for my parts anyway so I may as well run the loop and watch the results for the sake of science.

Oh yes this rads will work very well for passive colling. If you can find some used ones (they can hardly ever break) for a reasonable price, grab them instantly. But new once are very expenisve as far as I know.

[FakeGamerGuy]

While building my shopping list I decided to double check my socket type out of paranoia. On AMD's page for the Phenom II X4 965 I noticed that it says the max temp is 65C. If the passive system I linked to earlier was only cooling down to 71... Unless I'm missing something?

http://products.amd.com/pages/desktopcpudetail.aspx?id=591&AspxAutoDetectCookieSupport=1

[Stefan1024]

While building my shopping list I decided to double check my socket type out of paranoia. On AMD's page for the Phenom II X4 965 I noticed that it says the max temp is 65C. If the passive system I linked to earlier was only cooling down to 71... Unless I'm missing something?

http://products.amd.com/pages/desktopcpudetail.aspx?id=591&AspxAutoDetectCookieSupport=1

Well keeping a 140 watt chip under 65°C with 30°C ambient temperature with passive cooling is pritty ambitious. But with the big tank you should be able to keep it cool for some hours. Also the CPU is quite old, isn't it?

Down the road you will update to a 95 watt, max. 80°C CPU I assume, and the cooling will be solved.

[FakeGamerGuy]

Yeah, I was considering upgrading CPUs earlier today for the sole purpose of lower TDP (perhaps underclocking/undervolting to match my current CPU with even lower TDP).

 

Last night I bought my waterblock, pump, barbs, and hoses. I should have everything by Wednesday so either then or Thursday I'd like to hook it all up and run it with the tank just to measure the temperature rise and see if it eventually reaches equilibrium. It's not my goal for a passive system anymore, and soon I'll be buying traditional rads to cool my CPU and my new GTX 560 Ti GPUs (the fans on those cards are insane at load).

[Stefan1024]

Hmmmm it's a shame if you give up your planes for a passive PC. We have to less of them on LTT.

But hey, it's your system so it's your decision!

You can try to runn the daily use passive and just turn the fanns on the rads ob for gaming.

[Meawa]

make sure you have a lot more cooling capacity than you need. or you're going to kill those poor fish

[FakeGamerGuy]

I'm still going to do passive testing to see exactly how capable the system is. Since it depends on surface area my custom L-shaped tank would theoretically cool better than my test tank but I'm not really looking to do that anymore. I'll run the test purely with the tank, then with rads, then with rads and fans and compile all the data.

[FakeGamerGuy]

Half of my parts showed up today, the rest gets here over the next two days. I've never watercooled a computer before except with AIO systems and I plan on streaming my headache tomorrow while I put the waterblock on and install the barbs and pump. Wednesday the tubing gets here so I won't be able to actually use and test the system until then. I'll monitor the temperatures over time while I'm gaming to see what happens. Anyone want to take bets on how long I can reasonably use the system with no additional cooling?

[Stefan1024]

How much water is in your tank?

[FakeGamerGuy]

It's a 10 gallon tank but I filled to it 9 or so for my last test.

 

Got my waterblock installed today. Not pictured here are the barbs which I thought I had taken a picture of after installing. Pump is here too but I don't think it's strong enough for when I install the rads later on. Just waiting for the hose to show up. Hopefully testing will begin Thursday.

 

[BeakTrots]

Yeah! Science!

[FakeGamerGuy]

Okay, so before I could start doing science and monitoring temperature rise I had to finish the loop. So I scrubbed my tank clean, wiped it down with alcohol, put the pump in the bottom, and filled it with distilled water until the pump was an inch under the surface. With a section of tube running from the outlet of the pump straight into the tank to test the pump on it's lowest setting I got pretty good flow. Turned the pump all the way up and, while the flow wasn't much better, it did flow more.

 

 

So my housemates and I had some shots.

 

 

I got some hose clamps out of the garage (I'll use compression fittings in the final build), put the hoses on, plugged the pump in, and...

 

 

PANIC!! One of the clamps wasn't tightened down enough. Luckily I pulled the plug on the pump before too much water came out, but unfortunately it took me a minute to realize that the water was still leaking. I pulled the waterblock off and drained the water back into the tank. The one good thing to come out of this is that I found out that I should have put a bit more thermal paste on, so that's... a thing...

 

 

So now I'm doing the leak test with the block OUT of the computer and after tightening the clamps again it seems to be just fine. I even gave it a bit of a tug and twist and not a drop came out. Just waiting for the water to dry before I can do my actual testing but I'm really busy until Monday so it might have to wait until then.

[FakeGamerGuy]

Everything seemed dry and additional blasts of air into the hiding spots on the motherboard didn't produce any drops so everything seemed okay. Let it dry for a few more hours still while playing Mario Party with the house mates. Reconnected all the components and... nothing. I get lights on the motherboard, at one point the fins were spinning and the drives were spinning but I had no signal to my monitor. Switched from DVI to VGA, nothing.Eventually it stopped powering altogether. I still get the green power LED on my laptop but the power button/jumpers do nothing. I'm wondering if maybe some water got behind the CPU? Nothing should have made it's way into the PSU. Guess I'll completely tear the system apart, put it together in a few days, and... hope? Otherwise I'll take this as a sign that I needed to build a new machine.

[0ld_Chicken]

Everything seemed dry

Snip

put it together in a few days, and... hope? Otherwise I'll take this as a sign that I needed to build a new machine.

Oh no! I'll be hoping for good things to happen! That's rough man

[FakeGamerGuy]

Well it wasn't water. I tore my whole system apart to let it dry completely and when I removed the CPU from the socket I saw there was a glob of thermal compound bridging two pins. The compound that came with the waterblock was strangely runny compared to the Arctic Silver I normally use, so maybe when the water leaked it pulled some rogue compound down into the socket? Now I have to figure out how to clean between the pins and remove the compound from the socket. FML.

 

 

 

Update: That's exactly what was wrong. Cleaned both the CPU and the socket, put the main parts back together as a test bench, and... well, first I got no video, then I moved the video card up to PCI-E 1 and it POSTed.

 

 

 

Windows is running, CPU is running 26-28C during Fire Strike which I'm pretty happy with. Granted, it's unlikely to sustain those temperatures but the goal to see how useable the system is. I'll do actual testing and benchmarking on Monday, or maybe Sunday night.

[Stefan1024]

Nice to see you get it running again!

I'm on vacation for the next two weeks so i might not be able to coment the benchmark results, but keep going

[Whotos]

The temperature and noise levels in the benchmarks are going to be interesting. It might make me consider a similar project

[FakeGamerGuy]

The temperature and noise levels in the benchmarks are going to be interesting. It might make me consider a similar project

I don't have a DB meter but maybe worth picking one up for science after I get the system back in a (new) case.

 

Last night I had some free time so I ran Prime95 for a couple hours. I'll try to make nice graphs and stuff later when I have more time and I'll do a much more comprehensive (longer) test Monday or Tuesday. Temperatures below are from internal monitor and two different thermometers so results aren't 1:1 but should at least give an idea of deltas. Water thermometer is analog so it's hard to get an exact reading.

 

10:00 PM

CPU temp: 27 C immediately after boot, 32 C after Prime95 had been running for a minute or two.

Water temp: ~22.5 C

Ambient temp: 23.5 C

 

10:20 PM

CPU temp: 34 C

Water temp: ~23.5 C

Ambient temp: 23.3 C

 

10:45 PM

CPU temp: 34 C

Water temp: ~24.5C

Ambient temp: 23.6 C

 

12:20 PM

CPU temp: 37 C

Water temp: ~27.5 C

Ambient temp: 24.4 C

 

I thought I took a picture of the temperature drop when I stopped the test but I guess not. Over the course of maybe a minute the CPU temp dropped from 37 C to 28 C. This is actually really promising I think and I didn't even use the full 9 gallons. I had only poured in 6.5 so the thermal capacity is actually higher.

 

It should also be  noted that I had a 120mm fan connected to the CPU fan jumpers to blow air over the motherboard to cool the northbridge and whatever and to simulate having a large, slow spinning case fan. It was spinning at 1690 RPM. Next time I'll go into the BIOS and turn it down considerably. Aside from that the pump, despite being underwater, was still audible though any sounds in the room easily overpowered it. The PSU would only spin it's fan occasionally and was completely inaudible. The only other source of noise was the dual fans on each of my two graphics cards. Prime95 obviously didn't touch those but under gaming or graphics benchmarking loads they get pretty damn loud and will be the next components to get water cooled.

 

Times are from my camera's metadata which is about 20 minutes off and AM/PM was reversed so I wouldn't have to enter dates to make the chart not freak out.

[FakeGamerGuy]

Note: I've rewritten this post so it's coherent.

 

The other night I did a longer test and used OBS to record the display and a webcam to record the thermometers allowing me to get good data without having to write it down every n minutes.

 

 

This shows the temperature of the CPU, water in the tank, and air immediately next to the tank. For this test the tank was open. Prime95 had been running for a couple minutes so the first data point is CPU under load with room temperature water. At the 5 hour mark I stopped Prime95 and left the system to idle until the next morning. To avoid having a 20 gig video file I stopped the recording when I stopped Prime95 so there are no data points until the final one at the 14 hour mark. It's extremely likely that the CPU temperature would have dropped to the water temperature after a couple minutes given that when I woke up the two were identical.

 

Average delta between CPU and water temperature during load was 9.4 C and the average delta between CPU and ambient temperature (as measured next to the tank) was 12.7 C.

 

Noise from the cooling loop was very low. The pump made an audible hum but it was also a very inexpensive pump working at it's maximum setting. I'll experiment with lowering the pump speed but eventually it will be replaced when the video cards are water cooled. The biggest source of noise was the 120 mm fan that I had plugged into the CPU fan header to keep air moving across the board and to prevent the system from freaking out that there is no CPU fan. The fan was spinning at 1261 rpm which I believe is 100%. I attempted to lower the fan speed but it maintained it's high RPM so I'll have to tinker with it some more. PSU was perfectly silent as it only occasionally spun its fan and even then it was very slow and totally inaudible. The 4 GPU fans spun the whole time but I'm sure at their lowest settings and were also inaudible. The one time I ran Fire Strike the fans may have become somewhat audible but were massively overpowered by the fan plugged into the CPU fan header.

 

Right now I'm feeling pretty confident about the build. It's proven that it can maintain extremely low temperatures with very little noise under loads that no video game would produce (5 hours of 100% load on all 4 cores). I'm currently looking for a new case that includes holes in the back for the water lines so I can move the system up to my room, add the remaining 2.5 gallons of water, and start using it regularly again.

[FakeGamerGuy]

Found the BIOS settings to make it not freak out when there's no CPU can so now it's much, much quieter. The pump is now the dominate source of noise making an obnoxious rattling sound. While running 3D Mark the video cards stay silent but running the Tomb Raider benchmark they start to make weird noise (from the caps?) and the fans start to spin up a LOT more than while running 3D Mark.

 

Speaking of 3D mark, at stock clock settings @ 3.4 GHZ I'm getting 5149 on Fire Strike (http://www.3dmark.com/fs/5535806) and 11917 on Sky Diver (http://www.3dmark.com/sd/3229124). 

 

Also, I bought an NZXT N440 case to put all of this nonsense back together in. I also ordered compression fittings so I can get rid of these mega-ugly hose clamps.

[Vivilacqua1]

This is pretty cool, but wouldn't it be bad for the fish to have temperature fluctuations? Say from off to on and gaming?

 

EDIT- +1 for the h-440

[FakeGamerGuy]

So I got a new thingie in the mail today.

 

 

 

A Shiny new AMD FX-8350. 8-cores, 4.0 GHz stock, same 125w TDP. Probably should have picked up a E-version at 95 watts but I'm dumb so *shrug*.

 

 

BIOS recognizes the chip so that's good!

 

 

And since people have been asking me for pictures of the setup, here it is. It's just a workbench until my case comes in next week but you can see how ridiculously simple this setup is. Tank, pump, waterblock, tank. The glass rocks will be added to the tank for aesthetic reasons once it's upstairs. The fan is just for chipset cooling isn't actually necessary. I ran Prime95 for 10 hours while I was at work with no fan and when I came home the CPU temp was up into the 50s and the chipset was in the 40s IIRC. While certainly not in the danger zone of CPU temps it was warmer than I would have preferred. The warm weather certainly didn't help but at least I know that it can run at unrealistic loads for unrealistic periods of time. I recorded all the data and will try to get a graph up some time soon.

 

Anyway, I ran Fire Strike with the new CPU and my score only increased from 5150 to 5400. I'm investigating it now but I think it has something to do with the multiplier moving all over the place. Physics test ran in the mid- to high-20s where before it was in the 9s so I know it's doing it's job.

 

UPDATE: I guess I was looking at the wrong CPU or wrong video card earlier. My benchmark results fall inline with what they should be. My Cinebench R15 score was 687 for the CPU test, slightly better than a Core i7 3770. OpenGL score was 96.68 FPS.