Topic: Overclocking predictions using Novec 2 phase immersion cooling (Read 1100 times)

Yes there are a bunch of different ones and SOME of them can be combined to get the ideal temp you are looking to get.

But there are Novecs with different boiling temperatures. Wouldn't it work if you choose liquid with something like 50 -60 C boiling boint?  Ambient is well below that and 50 -60C is ok with boards.

No it's not, because you need a temperature differential between ambiant and Novec boiling point.
With high ambient, you'll have to use chillers or huge amount of air to cool the fluid in the condenser.

Novec won't really help reduce costs, it'll allow you to have a more dense power usage, reducing some fixed costs.

If you need to dissipate a certain amount of heat (watts/BTU), wether it's in 2 phase immersion cooling or air cooling, it'll cost the same (it will be even more in Novec).
The advantage of Novec is to allow you to densely pack those heat generating unit.

Maybe you can ask planetcrypto since they did some tests

https://www.youtube.com/watch?v=gcb9TyQP5ZA

BTW, with a proper condenser, you could squeeze at least 4 hashing boards in the tank they used, and if the spacing allows it, maybe even 6.
15-20mm between boards is usually fine.

Thanks Kilo17 for sharing your findings about 2PIC applied to home mining. Not being able to further overclock/overvolt in Novec pretty much kill profitability of such development. Novec may be a better option when you live in hot climate and need to use complex costly system to cool your farm, which is the case here only a few days a year.

Matias, there is probably chemicals that are able to deeply clean oil from miners that would get you full resell value.


I'll have to rethink about the next step. Probably a sound proof box or maybe oil cooling.

I did not notice this thread before starting my own.

https://bitcointalksearch.org/topic/immersion-cooling-1323283


Mods, feel free to merge!

I think immersion cooling in oil would be easier than  two-phase immersion cooling, because with latter you would need air tight box, but with oil immersion fish tank is enough.

Problem with oil immersion is, that your miner resale value would be close to zero after oil bath.

There are a lot of problems with your plan, imho.  I have done considerable research (and some development) of a small immersion tank for small farms.  There are also numerous threads here on the forum about it as well. 

The biggest savings with 2PIC is the savings from not spending as much on moving large amounts of air in data centers or farms.  It is NOT the gain from overclocking or overvolting.  The overall gain to the system is probably in the 5% range (A guesstimate).  So overvolting or overclocking the chip is not going to gain much more than water or air cooling.

I think you are confusing CPU cooling with ASIC cooling which are different beasts.  I was heavy into overclocking CPU's and ran DI and LN on a regular basis but if you know much about that then you would realize that those are short runs and there are inherent problems with it.  Have you ever poured LN or even DI on a CPU and let it sit there and cool off for about 2-3 minutes and tried starting the rig- well it doesn't work very well. (I won't go into detail).

Cooling like that allows you to put in tons of power to reach high speeds but Asics are different, try cooling the Avalon6 chips to about -30 and see how well they perform - now try doing it on a single miner with 80 chips and then 100 miners with 80 chips each. 

Now comes the real issue with 2PIC- It cools only to the boiling point and doesn't super cool anything, limiting how much power you can actually put to it, and then you start running into surface area problems.

Can it be done to some degree - yes, but the overclocking will not offset the costs of the cooling system for a few miners

That is my opinion and I could be wrong.

My system is similar to having my miners inside the HVAC intake duct. Miners preheat air before HVAC.

Right now are you just ducting your miners outtake fans to your hvac furnace using something like dryer cents?

You seem to have made the maths. I'm very interested about your opinion. I'll refine my spreadsheet about exact effect of lower efficiency as you overclock further.

Having lower efficiency may still yield better profit. Would you find it interesting to invest let’s say 250$ per S7 to convert it to 2PIC?

Oil cooling is an interesting technology also. My understanding is that it's a more complex system to design as you need to ensure proper flow of cold oil around all your boards. Tank output temperatures are also lower at around 40°C which is more difficult to cool in summer. Huge advantage of lower initial price.

Immersion in a 2 phase fluid is way more expansive. Tank is simpler but you have to control it well to ensure no precious gas is lost. Overclocking potential to be discussed. My feeling is that it would be in between water cooling and liquid nitrogen cooling.

tldr - did you look into 1 phase oil immersion cooling?

I doubt you will get anywhere near 7 Th and even if you do the efficiency will be crap.  I am also interested in immersion cooling but I fear you are wrong on the calculations.  It is great for cooling and you can increase the density of miners but the amount of overclocking/overvolting will not offset the price of the Novec for 6 or even 10 miners

I have 6 miners for about 2 months now. I plugged them all to my central HVAC system to distribute the heat generated through my whole house. I’ve tuned each miner’s frequency and fan speed to the limit of my noise level tolerance.

Now as the next step, I’m considering building a sound proof box around my little farm to allow greater overclocking. Note that I believe that the price of that box is lower than the price of buying low noise fans. Except that you have to design the sound box properly so that it does not overheat your miners. Being a professional engineer, I’m not afraid about that.

While designing this sound box, I’ve come across a more advanced cooling method: 2 phase immersion cooling which after hours of reading and calculations seems an ever more interesting challenge to tackle. The price of building this “professional quality” cooling tank is high, especially for the Novec fluid itself. And I wonder what overclocking would be feasible with this extremely high cooling potential.

Not being a pro of overclocking limits, I turn to you guys to help me predict what can be achieved with 2PIC. I’ve heard that overclocking and overvolting of an S7 would give Hash rate of about 7 TH/sec which is an improvement of +45%. If such improvement of hash rate can be assumed safe using 2PIC, than this tank development would make sense money wise.

Again, please help me predict what % of hash rate improvement could be safely achieved with this cooling method.

Thanks !