[Finished] The silent cube: Pushing passive cooling to the limit with dual GTX 980 - [Update 21: Liquid Cristal Thermometer]

[waggythegeek]

I just jumped on a new hype train yesterday: Useing peltier elements as a thermoelectric generator by takeing advantage of the Seebeck effect!

I know the efficiency is only ~1%, but it's just for fun. Because of the low heat conductivity I can't mount the elements directly on the CPU/GPU, so I will mount them on the big heat sink on the front (hot side) and cool them with the two smal heat sinks on each side. Therefor I can only expect a smal temperature difference (5 - 10 °C estimated), so I will get an oper circuit voltage of 50 - 300 mV per modul. Since I'm useing cheap 3 euro peltier modules I have no information about the thermoelectric generator characteristics.

But since I'm useing 10 modules and a low power LED needs only ~2 mW (1,8 volt, 1mA) I'm pretty confident I can at least illuminate some LEDs.

When the final measurements shows that I'm able to harvest >= 600 mW, I could even charge my smartphone on a slow rate (5V, 100mA) useing a boost converter.

Or do you know something amazing to do with up to 500 mW?

So basically, reverse peltier cooling?

[Guest]

So basically, reverse peltier cooling?

Yep.

 

In other peltier related news I have planned out a similar system as this one except with slightly more efficient parts and actual peltier cooling. Yep. Sub zero passive PC. Only issue is I have no money for parts.

[PineappleHolidays]

Yep.

 

In other peltier related news I have planned out a similar system as this one except with slightly more efficient parts and actual peltier cooling. Yep. Sub zero passive PC. Only issue is I have no money for parts.

"sub zero passive"

you will cool the hot side with giant heat sinks ? that's going to be hard .

[Darth-san]

Amazing case project

 

But, will it blend ?

 

I would have thought that anodised aluminum would actually weigh less...40 kg, looks like a very heavy duty case

[Guest]

"sub zero passive"

you will cool the hot side with giant heat sinks ? that's going to be hard .

Yep. I'm going to have to get very efficient parts, so I can use lower power peltiers and still get sub zero. And a lot of heatsink.

[Stefan1024]

@Biased Opinion

At first I thougth about that. But I don't dare to two independed PSUs in parallel mode without a load balanceing system or at least some diodes in series. With all components under full load I will draw ~530 watt. But that's a very unlikely scenario. And I'm allowed to draw 550 watts constantly and 605 watts for one minute. And like you said two fanless PSUs are expensive.

 

@Rheinwasser

The whole thermoelectric generator costs me around 50 euro including shipping. Compared to the rest of the build it's not that much. And I also have a little board here that measures the temperature and humidity every 30 seconds. It also has a display and a microcontroller. It consumes 5 mW! Now it running on a button cell that should least for 3 years. But I'm thinking about sourcing it with the peltiers.

 

@bear-in-the-air

It would be amazing if you do this one day! Pleas let me know when you start....

Technical peltier cooling is also an active cooling method. To cool something under the ambient temperature without doing something is physical impossible. But it will still be fanless and potentially noiseless.

 

@Darth-san

Thx! Since I will screw the heat sinks together and they are very rigid it should not be able to bend. And don't get fooled by the pictures. The heat sinks are huge and have a massive 15 mm thick base plate for the heat distribution.

[Guest]

@bear-in-the-air

It would be amazing if you do this one day! Pleas let me know when you start....

Technical peltier cooling is also an active cooling method. To cool something under the ambient temperature without doing something is physical impossible. But it will still be fanless and potentially noiseless.

I guess. The only thing holding me back currently is money. If I go for a low enough power chip I will be able to use an off the shelf cooler and keep the PC in a test bench form. I'm also tempted to get a low end card that only has a heatsink on the chip and put a peltier and large cooler on that too.

[Stefan1024]

Sneak peek: GPU heat transfer plate

 

I made the first heat transfer plate for the GPU so I can mount it to the heatsink.

 

 

A lot of people complaining the EVGA GTX980 SC ACX 2.0 has no backplate. I bougth this because it has a frontplate, that gives the card mechanical strength and unlike the backplate better cooling. For the final assembly I will add much more thermal pads between the frontplate and my copper plate. So the whole PCB is connected to the heat sinks and VRM as well as VRAM temperatures should be fine.

 

[Stefan1024]

Update 5: GPU 1 cooling tested: OC with passive cooling?

 

I mounted the GPU to the heat sink today to test the cooling. Don't worry I will reassemble the slot bezel again later.

 

Pics:

 

The temperatires are very pleasing. The core is only ~15°C hotter than the heat sink. Due to the big heat capacity of the heat sink the rapid temperature change of the core is limitted to 15°C. This avoids mechanical stress.

For short (5 mins) peak loads the core stays below 55°C. After one hour of stress testing the temperature settles around 73°C with peaks up to 75°C. The ambient temperature was 26°C. This measurements are taken with a an OC to 1,5 GHz and 110% power target! Personally I don't like OC that much, since the card is very powerful at default settings and the performance per watt ratio of OC is pretty bad.

The temperature in the finished system will probably be a bit lower since all heat sinks will be coupled. In this thest the heat sink runs on it's own.

 

 

 

[X_Rated_Box]

wow...

[Stefan1024]

Smal news:

I got the peltier elements and measured them. At 10°C temperature difference, they have an open circuit voltage of ~200 mV and a short circuit current of ~10 mA. Since I only have a cheap multi meter the short curcuit current is probably higher, due to the high internal measuring resistance of the device.

It looks like I can only expect ~0.5 to 1 mW of power per module. This is enougth for an LED (> 5 mW), but I won't be able to do a lot more. But I have to redo the measurement when the modules are assembled.

[0ld_Chicken]

Smal news:

I got the peltier elements and measured them. At 10°C temperature difference, they have an open circuit voltage of ~200 mV and a short circuit current of ~10 mA. Since I only have a cheap multi meter the short curcuit current is probably higher, due to the high internal measuring resistance of the device.

It looks like I can only expect ~0.5 to 1 mW of power per module. This is enougth for an LED (> 5 mW), but I won't be able to do a lot more. But I have to redo the measurement when the modules are assembled.

Never thought about using Peltier as a generator before. That's awesome! Even at.5w (however inefficient it may be in this application).

Edit: subbed!

[Stefan1024]

Redone first post...

[Uwillparish]

Redone first post...

It looks alot better than the original one, Still can't wait to see this build finished!

[skylane]

Amazing build and great concept. sub

[Stefan1024]

Update 6: Energie harvesting efficiency = 0.0005%

 

I measured the energie output of thee peltier modules in serie today. To do this i atached the second heat sink provisional.

 

 

I get 1 volt open circuit voltage and 45 mA short circuit current. This gives me a maximum power of ~12.5 mW. This is enough for the boost converter to start up, bit it only reaches 4.2 volts instead of 5 volts. Once I will add the other modules and a big capacitor at the input it will work.

 

 

The efficiency is funny:

Power from the wall: 250 watts

Power from the generator: 12.5 mW

Efficiency: 0.0005%

 

I also ordered the professional PCI-E riser and I'm working on the CPU cooling.

[Guest]

Dat effiency doe. If you power the peltiers you could probably use the smaller h/s to make the main h/s a couple of degrees cooler, but that will make the smalll H/S  alot hotter.

[Necrodead]

Wow lots of work have gone into this and it shows.

 

Also where did you order your Pci-e riser? I'm looking for one but i've heard most of the cheapo ones don't work.

[Stefan1024]

@bear-in-the-air

I could to this but only to very limited extend:

- I'm not allowed to heat per modules over 80°C. So the smal HS is maxed out at 80°C and the main HS is ~60°C now.

- I have only ~30 watts to spare in my power budget.

As you can see I could only get rid of ~10 watts more. But luckily the cooling is good enough without the additional HS.

 

@Necrodead.

Up to 15 cm the ribbon cable based once are fine. For larger distances you need impedance matched once.

I bougth them from Digikey. Over 100$ you have worldwide free shipping.

250 mm: http://www.digikey.com/product-search/en?mpart=8KC3-0726-0250&vendor=19

500 mm: http://www.digikey.com/product-detail/en/8KC3-0726-0500/3M12026-ND/3641403

[Necrodead]

 

Up to 15 cm the ribbon cable based once are fine. For larger distances you need impedance matched once.

 

Thanks, good to know as i'm only needing a couple cm extension.

[Stefan1024]

Update 7: CPU block and PCI-E riser

 

Yesterday I made the copper block to connect the CPU to the heat pipes. Drilling the 60mm deep holes with this big, old drill press is defficult because it runns slightly eccentric. This forces the drill to chafe inside the hole and I can feel the vibration in the handle despite the vice is bolted to the machine. Really don't like it.

The copper block heats up during drilling a lot. And you know it's hot when moisture and oil is evaporating from the vice :unsure:

Luckily this extends the hole a bit and reduces the chafing. Also it creates a tight fit with the heat pipes once cooled down. So I'm not sure if I should cool it during drilling.

 

You guessed it, Murphy's law struck and the drill get's stuck and broke just before the hole was finished.

 

 

But I can use the machine for free so I have to stop complaining now...

 

However the CPU block in nearly finished now, I only have to sand it.

 

 

Also the package with the PCI-E riser and other stuff arrived, so I can test the whole system. In benchmarks the system power consumption peaked around 450 watts. Considering 550 watts continuous and 605 watts peak power deliverty of the PSU the system fits perfectly into the power budget.

 

 

The PCI-E riser are expensive but worth the money. 0.5 meter riser and it works like it sits in the MoBo slot.

[Stefan1024]

Update 8: CPU cooler

 

This week I was working on the CPU cooler. To drill the long holes this time I sprayed WD-40 from time time on the drill. It works fine and the holes are much smother now.

The WD-40 evaporated quickly due to the heat and it's highly flammable but luckily it didn't caught fire.

 

Sadly my smartphne camera is pritty bad to make a picture of vapour:

 

However I mounted the cooler:

 

 

Sadly the cooling performance isn't acceptable. Yes it's 32°C in my room rigth now, but during stress test the CPU reaches 99°C after 15 minutes and during gaming it's 70°. The big heat sink isn't to blame because it stays relatively cool. I assume the CPU block isn't making proper contact to the heat pipes since the block on the CPU is hot and the block on the heat sink is only lukewarm.

 

I found a new Zalman FX70 for only 40$ so I will mount it on top of the copper block to get rid of ~20 - 30 watts. Also I will redo the CPU block mounting.

 

Also some personal thougths: The mounting force for the CPU cooler is extreme. First I used some normal springs to hold the block down. It didn't work and the past wasn't spread from the preasure. Then I used some leftover springs from an other CPU cooler. They are very hard to squeeze and the force on the CPU is considerable. But it looks like it is desinged for that and fine.

[Guest]

So badass, especially 74 degrees on the 980.

[Stefan1024]

Update 9: CPU cooler finally mastered

 

So I finally got the cooler and put it on top of the copper block.

 

Test: Bios -> set may power consumption of the CPU to 100 watts -> Prime95 -> max power consumtion preset

So the CPU consumes 96 - 102 watts permanently and I can meassure. The max. temp. on the hottest core was 91°C. Normally they where hovering around 85°C. Not to bad considering 28°C ambint temp. and a normal application only reaches ~80 watts.

 

The next two weeks the build is on halt since I'm on vacation.  B)

 

And sorry for crappy foto:

[Guest]

-snip-

GIVE ME YOUR SSD PLEASE (very good job by the way this is insane)