Novec 7000 Project [immersive evaporating cooling]

[fran2k]

This would be interesting to cheap if not were so so expensive!

[user27]

Sorry, cross posting with rig photos thread...

I think he is running Novec 7000 which changes state at 35? Very difficult to condense the vapour on something like that unless you are in very low ambient temps or you have a crap load of below ambient cooling (a single pelt uses as much power as it transfers so hardly an economical option). Novec 7100 boils at 65 which is a perfectly acceptable temp for most chips to run at and a lot easier to condense the vapour from 65 than 35.

The thing with HFE's is that they solve all your the chip cooling problems and move them along the chain; if your asics are pulling 1000w then you then need at least 1000w of cooling to condense the vapour. 1000w of cooling is not so hard from 65 but it is quite a challenge from 35 depending on you ambient temps.

The main benefit of HFE is that it makes the heat usable; you can condense vapour from 65 with a radiator and water at 60 if you have high enough flow so that water is hot enough to be useful to heat water and such like. If your chips are happy to run at higher temps you can use 7200 or greater which allow the cooling water to be heated to higher temps which are even more useful.

smracer looks to be rocking the 7100 which is easier to work with and still yeilds chip temps of 65 for almost unlimited heat dissipation.

I'm not sure how I will get on with it at this early stage but my intention is to use Neptune at a 2kW water heater by the time winter comes to England.

u27

[klondike_bar]

Sorry, cross posting with rig photos thread...

I think he is running Novec 7000 which changes state at 35? Very difficult to condense the vapour on something like that unless you are in very low ambient temps or you have a crap load of below ambient cooling (a single pelt uses as much power as it transfers so hardly an economical option). Novec 7100 boils at 65 which is a perfectly acceptable temp for most chips to run at and a lot easier to condense the vapour from 65 than 35.

The thing with HFE's is that they solve all your the chip cooling problems and move them along the chain; if your asics are pulling 1000w then you then need at least 1000w of cooling to condense the vapour. 1000w of cooling is not so hard from 65 but it is quite a challenge from 35 depending on you ambient temps.

its a lot easier and cheaper to cool a closed-loop coil within the vapour area, and then extract the heat of that loop outside the building or with fewer fans than would be needed for air cooling.



Below is a pump and heat exchanger that moves heat outside so it can be dissipated by chillers.

[DeathAndTaxes]

its a lot easier and cheaper to cool a closed-loop coil within the vapour area, and then extract the heat of that loop outside the building or with fewer fans than would be needed for air cooling.

Both 7000 and 7100 can be used with the same closed loop system.  The point is that the larger the difference between ambient (i.e. outside) temp and the condensation point of the working fluid the more efficient the operation.   Now obviously a fluid which boils at 400C would be very efficient but it would also mean the processors would be 400C as well which wouldn't be good.   So the optimal working fluid in a two phase immersion cooling system is one with the HIGHEST boiling point which is still maintains the processors at an acceptable temp.

[user27]

Both 7000 and 7100 can be used with the same closed loop system.  The point is that the larger the difference between ambient (i.e. outside) temp and the condensation point of the working fluid the more efficient the operation.   Now obviously a fluid which boils at 400C would be very efficient but it would also mean the processors would be 400C as well which wouldn't be good.   So the optimal working fluid in a two phase immersion cooling system is one with the HIGHEST boiling point which is still maintains the processors at an acceptable temp.

Yep, 34 (N/7000) is not really workable but 61 (N/7100) is fairly easy to condense and also a good working temp for chips; I would not be comfortable at 98 (N/7300) but 76 (N/7200) might be workable once the cooling is dialed in.

Higher boiling point also increases the usefulness of the hot water returned.

u27

[RoadStress]

Both 7000 and 7100 can be used with the same closed loop system.  The point is that the larger the difference between ambient (i.e. outside) temp and the condensation point of the working fluid the more efficient the operation.   Now obviously a fluid which boils at 400C would be very efficient but it would also mean the processors would be 400C as well which wouldn't be good.   So the optimal working fluid in a two phase immersion cooling system is one with the HIGHEST boiling point which is still maintains the processors at an acceptable temp.

Yep, 34 (N/7000) is not really workable but 61 (N/7100) is fairly easy to condense and also a good working temp for chips; I would not be comfortable at 98 (N/7300) but 76 (N/7200) might be workable once the cooling is dialed in.

Higher boiling point also increases the usefulness of the hot water returned.

u27

What's the price of the N7200?

[user27]

What's the price of the N7200?

Last I checked £248 for 5.4kg / 4L.

[lkthomas]

I just wondering, in a close loop system, if bottom is the miner board, top is radiator/condenser, would it work this way? or I need to have two separated system (Novec + miner = first system; Radiator + fan module = second system) and link up with an aluminum plate to do heat exchange?

[user27]

I just wondering, in a close loop system, if bottom is the miner board, top is radiator/condenser, would it work this way? or I need to have two separated system (Novec + miner = first system; Radiator + fan module = second system) and link up with an aluminum plate to do heat exchange?

One system is fine; water loop just moves the heat from the condensing rad to the cooling rad.

EDIT:
A friend has managed to get a Neptune board up and running in Novec 7100 now:

[user27]

https://bitcointalk.org/index.php?topic=313978.msg7865759#msg7865759

[mangodream]

I'm just waiting my FREE (yay !!!!) Novec 7000 and 7100 delivery, i'll put 9 Gridseed Blades in it, so 18 PCBs very close to each others. I'll mix the 7000 and 7100 together to have a perfect boiling point at 47,6 °C. 34°C is nice for the GC3355 chips but too low for coils and things, and 61°C is too hot for the chips. So i think we can get the best results when mixed up

I'll post some pics when making all the stuff

[jimmothy]

I'm just waiting my FREE (yay !!!!) Novec 7000 and 7100 delivery, i'll put 9 Gridseed Blades in it, so 18 PCBs very close to each others. I'll mix the 7000 and 7100 together to have a perfect boiling point at 47,6 °C. 34°C is nice for the GC3355 chips but too low for coils and things, and 61°C is too hot for the chips. So i think we can get the best results when mixed up

How did you get it free??

Also are you sure you can just mix the fluids like that? Why not use novec 649 which boils at 49C?

[ThomasCrowne]

I'm just waiting my FREE (yay !!!!) Novec 7000 and 7100 delivery, i'll put 9 Gridseed Blades in it, so 18 PCBs very close to each others. I'll mix the 7000 and 7100 together to have a perfect boiling point at 47,6 °C. 34°C is nice for the GC3355 chips but too low for coils and things, and 61°C is too hot for the chips. So i think we can get the best results when mixed up

How did you get it free??

Also are you sure you can just mix the fluids like that? Why not use novec 649 which boils at 49C?

Yes please...do tell!

[sang]

I'm just waiting my FREE (yay !!!!) Novec 7000 and 7100 delivery, i'll put 9 Gridseed Blades in it, so 18 PCBs very close to each others. I'll mix the 7000 and 7100 together to have a perfect boiling point at 47,6 °C. 34°C is nice for the GC3355 chips but too low for coils and things, and 61°C is too hot for the chips. So i think we can get the best results when mixed up

I'll post some pics when making all the stuff

I would strongly advise not mixing these fluids until you speak to a 3m engineer. They may very well not mix at all and just settle on top of each other - depending on the properties.

Also, you really suck. :-)

[user27]

Yes, please explain free; I have only ever been able to get hold of very small samples.

Mixing them is very unlikely to work; I imagine you would just boil off the 7000 first even if they do mix. Same as distilling alcohol from mash; even when mixed you can cause the alcohol to evaporate leaving the water behind.

u27

[mangodream]

The 3M guy told us that we can mix them together since they use the same chimical elements. But you can't mix a 649 with any 7000 series liquid, it's not the same chemical
It's designed that way so you can select the perfect boiling point for what you want to cool.
But the guy wasn't an engineer, just a representant. Wait and see, i sendt a mail to the 3M staff, my Novec will be delivered within one week, so i have the time to verify if it's true or not I'll give you the truth from them
If it's false, idk, i'll have some liters for something else, maybe my Zeus miners.

I have them for free because a friend of mine who owns a datacenter told them that he must do some tests of this kind of application and if it works like expected and not a pain to install and set up all the entire system, he will order a lot more of Novec for cooling his entire datacenter... so, yeah, they're interested as fuck for the deal

[user27]

The 3M guy told us that we can mix them together since they use the same chimical elements. But you can't mix a 649 with any 7000 series liquid, it's not the same chemical
It's designed that way so you can select the perfect boiling point for what you want to cool.
But the guy wasn't an engineer, just a representant. Wait and see, i sendt a mail to the 3M staff, my Novec will be delivered within one week, so i have the time to verify if it's true or not I'll give you the truth from them
If it's false, idk, i'll have some liters for something else, maybe my Zeus miners.

I have them for free because a friend of mine who owns a datacenter told them that he must do some tests of this kind of application and if it works like expected and not a pain to install and set up all the entire system, he will order a lot more of Novec for cooling his entire datacenter... so, yeah, they're interested as fuck for the deal

He could well be right but I would be wary of this advice; if it were so simple why do 3M sell      34    61    76    98    128      boiling points and not say a 50 which would save a lot of people mixing?

u27

[aerobatic]

I'm just waiting my FREE (yay !!!!) Novec 7000 and 7100 delivery, i'll put 9 Gridseed Blades in it, so 18 PCBs very close to each others. I'll mix the 7000 and 7100 together to have a perfect boiling point at 47,6 °C. 34°C is nice for the GC3355 chips but too low for coils and things, and 61°C is too hot for the chips. So i think we can get the best results when mixed up

I'll post some pics when making all the stuff

I would strongly advise not mixing these fluids until you speak to a 3m engineer. They may very well not mix at all and just settle on top of each other - depending on the properties.

Also, you really suck. :-)

I was told by a 3m engineer you can mix them to dial in any boiling temp you want.

[cozk]

Mega waste of time and money.

Gratz

[jimmothy]

Mega waste of time and money.

Gratz

I don't think you're understanding this technology fully.

The whole point is to save time and money.

Cooling the electronics better than any other method is just an added bonus.