Whole Room Water Cooling Part 6

[Counter-Strike Player]

Does anyone have the link to the WAN show that Linus first thought of the idea on camera? I wan't to see that part again.

 

Well from what I recall, part 1 was uploaded 7 December 2014 and even THAT was a long delay from the initial start of the shooting. I think that's why they had parts 2, 3, and 4 in queue fairly quickly, but then nothing happened for months and hence the sporadic episode release. I noticed they didn't even include part 6 in the playlist, lol.

 

I'm guessing pre-August 2014 because this wan show had "teasers" (again for a video that didn't get released until 4 months later, lol):

[Counter-Strike Player]

Oh Luke...https://www.youtube.com/watch?v=nl8nhDHH07A&feature=youtu.be&t=56m40s

[Counter-Strike Player]

@ThatAngryGnome

Here is the "official" announcement from 25 July, 2014: https://youtu.be/Xe0mitLNRhQ?t=16m It sounds like Linus wanted everything sponsored from the start (Swiftech, Nvidia, and probably A LOT more emails sent out to other companies that never got back) and hence his excitement from the start. Luke @17:58: "this is his solution to not buying AC." Linus's (more potential dollar-saving) reaction to the idea that Luke's dad can hard-pipe the copper a few minutes later: https://youtu.be/Xe0mitLNRhQ?t=21m12s so obviously this was about money from the start, even though they explicitly said it was a "100% crazy" idea @20:58.

[Counter-Strike Player]

@ThatAngryGnome

More snicker-worthyness and corroboration: https://youtu.be/3fPSk5TSxTk?t=27m16s

[ThatAngryGnome]

Well from what I recall, part 1 was uploaded 7 December 2014 and even THAT was a long delay from the initial start of the shooting. I think that's why they had parts 2, 3, and 4 in queue fairly quickly, but then nothing happened for months and hence the sporadic episode release. I noticed they didn't even include part 6 in the playlist, lol.

 

I'm guessing pre-August 2014 because this wan show had "teasers" (again for a video that didn't get released until 4 months later, lol):

Thanks bro!

[Moejoe]

Linus should read the manual of his infrared thermometer. You cannot determine the temperature of reflective (or even rough) metal surfaces with a thermometer like that. You will always measure the temperature of the room, since metal is really bad at emitting radiation and very good at reflecting it.

 

The Leslie Cube is a very good illustration of this problem: https://upload.wikimedia.org/wikipedia/commons/7/7f/LesliesCube.png 

 

It is a cube that is filled with boiling (or very hot) water. So all surfaces of the cube will have the same temperature more or less. But if you look at it with a infrared camera, the metal surfaces appear to be cold, even though they are not.

 

So don't measure copper pipes with that ok ?

[pk-man]

Although watching your progress was fun and sometimes cringeworthy, the project only sounded cool on paper. I bet Linus only did it because he tried to cheap out on A/C.  I hope he learned his lesson haha.  There were two problems with this project. One is that it lacks expandability and flexibility.  The setup he had wouldn't allow him to add more computers or reposition them around should the need arises. Another problem is that the room is still piping hot. Even if the computers are cooled, the employees are not.  Image sitting in a hot room most of the day sweating your ass off trying to work. What if someone is emitting bad body odor? Not fun.

 

All in all, it was a fun series to watch but the whole thing didn't have much practicality.

[TheInfamousMrD]

Top marks Linus for having the vision to try doing something epic like this. It is an amazing achievement to have stuck with it for 12 months, let alone get it to the point where it works as it should. Regardless of whether you or anyone else consider it a success, you should be really damn proud that you made it happen.

I have only recently discovered your channels but I am enjoying watching the videos you guys are making. Keep it up & keep pushing the envelope.

[Jumper118]

Did you have any freezing or condensation problems in winter?

[MastuhWaffles]

This is cool, but what happens when you guys move to the new office?

Was this really worth it? Your going to have to move it or leave it later on.

[Snadzies]

Months and months of Linus refusing to get air conditioning installed, instead coming up with the whole room water cooling solution and, in the end, finally admitting that air conditioning would have been the better solution.

 

While I am sure there are many who feel vindicated by the conclusion (some undoubtedly working at the LTT office), thanks to Linus' stubbornness we have not only some very entertaining videos but a good lesson / warning to the community about whole room water cooling.

 

I'm sure that, just like air conditioning, he will have to begrudgingly admit his refusal to install a fire pole was wrong and the LTT office will finally get the fire pole it needs.

[Turbo_Steve]

This was a really entertaining video - one of the best (and the one to finally get my lurking butt to sign up) but the engineering errors have had me properly shouting at the TV

 

As far as I can see, there are three primary heat sources in that room (in reverse order):

3) The PCs (~400watts per unit)

2) The humans (~120watts of heat output per-person)

1) Solar Gains (generally measured in kW)

 

Assuming we can't fix the humans (though putting them all on a low-calorie diet would reduce their heat-output, I'm not sure Linus would last long!) the first and foremost fix would have been to tint the windows - or black them out completely.  This makes a huge difference.

 

The watercooling loop is a neat enough idea (as suggested, running the pipes at waist height, and raising the reservoir above this level would help with bleeding, would allow the system to gravity-fill and would also mean that the reservoir needn't be pressurized:  a simple PVC bowl would suffice!) but a far more elegant solution (though potentially a little more expensive) would be for each PC to use a water\water heat exchanger internally.  Not only would this force each PC to run isolated from the main loop (mitigating any contamination or poor chemical selection issues - each PC can run it's own suitable coolant internally) but would act as a "buffer" smoothing out any hot-spikes caused by poor flow in the main-loop.   The catch is that the temperature differential does need to be somewhat larger for this to be truly efficient (so the coldest temperature attainable inside a computer will rise compared to a a directly connected loop) but the benefits generally outweigh this.   Insulating the pipes is obviously a must, too:  even painting them would have made a significant difference here (especially if they were black & white for the appropriate side of the loop!).

 

Finally, arguably one of the main heat generators in a PC is the power-supply.   Most of these are 85%-90% efficient, so at 450watts of draw, that's 45watts (or more) of heat being released into the room.    There are probably a few more watts in some of the motherboard sources too.  So perhaps moving all those power-supplies into another room would be a cheap way of reducing the heat output further - especially if you're not keen to watercool the PSU (which is entirely understandable!).

 

It's an awesome experiment, and I would love to see a do-over, but as with all Businessmen, Linus has (quite correctly) chosen to adopt an industry-standard and recognised method of cooling his revenue-generating tools (the team and their computers) and to leave the experiments to the lab, or to a specialist.   A shame, but totally understandable.

 

My only remaining concern is that the HVAC engineer in the new office has been able to do the correct calculations :  The LTT office is NOT likely to be a typical office (with computers outputting at least twice as much heat as an average office-users machine, and potentially twice as many machines) and I am a little concerned that the "server room" is not running independant aircon - 24U of servers i& switches is likely to work out to be around 12kW of heat ouput when running full-load:  that is some serious heat-output!  Unless the plan is to use a high-flowing exhaust fan to constantly expel a significant number of CFM from the server room, drawing cooler (office temp) air into the server room through a carefully selected grille (tuning the amount of flow to keep negative pressure in the room) then it's going to get hot in there:  business-impactingly-hot.    

 

If Mr.HVAC has a proper understanding of what is going in these rooms - it's all sorted.   If not - potential problem.

 

Lastly - presumably the new, larger studio is going to have some pretty awesome lights in a variety of colour temperatures - if they are all LED and Fluro then no problemo, but tungstens and halides and the like can chuck out some serious heat too.   

[Counter-Strike Player]

I bet Linus only did it because he tried to cheap out on A/C.

 

Did you not see the timestamped link I made 3 posts before yours? Watch it for just 15 seconds...your head might explode though.

[Counter-Strike Player]

...one of the main heat generators in a PC is the power-supply.   Most of these are 85%-90% efficient, so at 450watts of draw, that's 45watts (or more) of heat being released into the room.

 

I find your lack of sources disturbing.

[Turbo_Steve]

I find your lack of sources disturbing.

 

http://www.antec.com/Believe_it/index.php?page=80Plus_info

 

They generally all have stickers on them bragging if they're better than 80% efficient these days - all my numbers are "wet finger" assumptions.

 

Oh, and just for laughs - http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story&reid=48

 

Which got me to thinking - would the best way of extracting ALL the heat from a PC from the room be to submerse the whole assemblage into a bath of inert oil, and to then have a heat exchanger from this fluid to the rooms main cooling loop - linus' copper coil from his DIY watercooling project could work, here.       http://mygaming.co.za/news/hardware/44281-deep-fried-pc-cooling-anyone.html

[ace10301]

This series was the only reason I started watching "LinusTechTips" videos, and now I watch the daily video every day at breakfast.  Sweet.

[NumLock21]

MSi like Linus's whole room watercooling idea?

 

Credit: Ian Cutress @Anandtech.com

[nutral]

I'm not sure the calculations where done on this, but i think the main loop was a bit too small in diameter, so then even if the pump is large enough it can't pump enough through because of the too small pipe it has to go through.

 

Second, the system should have been calibrated for all the computers with simple valves for the intake of every computer. You can open the valves a bit more or less to make sure that all the computers get the same amount of flow going through, couple that with an in series watercooling setup in the computer and the coolant should always stay clean at the computers.

[EposVox]

Was he even hitting the copper pipes with that temperature gun at the beginning? lol

[Pmacca]

Like the broken paino the pump setup was doomed from the start one pump should have fed each end of the system for better flow an valves on the down pipes to regulate the flow like some one else said.

[BigE]

I hate to tell you this since it happened so long ago, but the issues you had with the system weren't due to corrosion or organic material. Luke's dad installed the copper pipe using standard home plumbing practices which includes using flux on all of the joints he soldered. If you notice in the video, he uses a copper colored flux. This is the orange discoloration you saw in the pipes. When flux is heated, it melts turning into a greasy paste. This is the floaters you observed. It's also the source of the oily deposits you observed in the graphics card water block window. As I mentioned, Luke's dad used standard house plumbing practices which assumes that anything in the plumbing lines will be flushed out because home plumbing systems are open loop. Water comes in from the street and exits through drains. No water recirculates. Your system is a closed system that recirculates the same water which only spread the unmelted and melted flux throughout the system. The way to prevent this would have been to flush the copper pipes with a cleaner prior to hooking up the computers or pumps. Alternatively, you could have used a product similar to Pex piping as it wouldn't have required flux or soldering. I know this long after the fact but I just recently discovered your YouTube channel. I hope it at least provides some peace of mind to know the source of the issues.

[NavyCuda]

I've been thinking about this more and I think if you ever consider doing it again, look to some industrial choices.

 

Instead of a large radiator it might be more efficient to use a chiller, which can bring coolant temps down to around 4-10 celcius. 

 

There needs to be a bettet manifold design as well.