Topic: We'd love board feedback on our concept: Combined Heating and Computation - page 2. (Read 6850 times)

You can't extract much energy from low-grade heat.  Maximum efficiency = (T2-T1)/T1, where T2 is the temperature of the low-grade heat (exhaust from the computer system), and T1 is the ambient temperature of where you dump the heat.  All temps are absolute (K).  Typical chip surface temps are 60-80C, and 70C is about 343K. If your cold end is a  cool day (say 15C), then T1 is 288K. The maximum energy you can recover from that heat is 19% of the input. That's a theoretical upper limit. Actual heat engines don't do that well.

Systems for absorption-Rankine low-grade heat recovery have been proposed.. At 120C in, 25C out, it might be possible to get 8% efficiency from a practical system. So yes, you can get some energy out, but it's barely worth it.

It is even worse than the above calculation.  The maximum energy that can be extracted between temperatures Thot and Tcold is (Thot - Tcold)/ Thot.  The above equation should be (T2-T1)/T2 rather than (T2-T1)/T1.

For the above example of 70C and 15C the maximum possible efficiency is 16% rather than 19%.  In reality the efficiency is likely to be between 5 and 10%.

Trying to extract any power from a small temperature difference is very inefficient.  Large power stations get better efficiency because the difference between high and low temperatures is around 500C, not 50C.  Power stations need the biggest possible temperature difference - that's why they need cooling water or cooling towers.  


In summary part of the confusion in this thread seems to be because there are 2 very different ways proposed in this thread, and on the 3xergy site, to use the heat.

1)  Simple use the heat as heat.  This is efficient. Simply use the heat from a miner to heat a room or heat water.  This seems to 'first generation' proposed at the www.3xergy.com site and appears to be nothing more than using a miner's excess heating to do space (or water) heating.  There is nothing new or unusual with this proposal.

2) The second way to use the heat is to use the small temperature difference between the miner and a cool point to extract power, for example to generate electricity or to power refrigeration.  As shown above this is very inefficient with small temperature differences.  Extracting power from the heat (temperature difference) and distributing excess heat via the electrical grid, as proposed on the www.3xergy.com site,  is extremely unlikely to be cost effective.  I won't call it a scam but it may be wishful thinking.  It can be done but is rarely worth the effort.  If you want more information see Carnot's principle or the second law of thermodynamics.    

They don't use heated water do they? I'm assuming there would need to be some modifications to the house/heating elements.

**Its really just adding a water jacket around the generator (where heat makes the chemicals react) and removing all of the propane/natural gas burners.  Its much les costly to heat them with the coolant than gas from a parts perspective so the units should be less expensive at scale.


How many of those companies building absorption refrigeration plants are using datacenters as a heat source?

**None.  They run them too cold to make the kind of heat you need to run building absorption refrigeration plants.  We aim to change their minds and show them a different way of thinking about heat and computation.

I know the technology is not new but it's not removing the cooling element. You will still need fans+heatsinks (or cooling towers, etc) to cool the water as well as electricity for the two water pumps required.

**We don't need a cooling tower, the heat is consumed in the work - heating water, making cold, etc.  Our idea is to not waste a single therm of heat if possible, its not getting pumped into the air as a waste stream - it is a primary heat stream.  Also, the 'new' parts of this technology actually address some of these very issues, like pumping specifically.

Thanks again - good thinking!  

You can't extract much energy from low-grade heat.  Maximum efficiency = (T2-T1)/T1, where T2 is the temperature of the low-grade heat (exhaust from the computer system), and T1 is the ambient temperature of where you dump the heat.  All temps are absolute (K).  Typical chip surface temps are 60-80C, and 70C is about 343K. If your cold end is a  cool day (say 15C), then T1 is 288K. The maximum energy you can recover from that heat is 19% of the input. That's a theoretical upper limit. Actual heat engines don't do that well.

**The refrigerators are mass produced today - every RV, most boats and many, many off grid homes already use absorptive refrigeration... it is safe, economical and completely silent.  When you take into account that your refrigerator is the device that runs the most in your house, you'll likely rethink that last statement. 

Actually several multi billion dollar corporations build absorptive refrigeration plants every single day.  The majority of high rise buildings in every major city use absorptive chiller plants for cooling.  Absorption cooling was invented by the French scientist Ferdinand Carré in 1858 and predates the direct expansion refrigeration systems you are familiar with by a long shot.  We havent developed any new technology (yet) we're using existing technology to do something new and novel.  This really isn't rocket surgery... its more like shade tree mechanics.

Please do get into the thermodynamics. I'd love to know how you can cool something using nothing but heat. I'm sure there are also several multi-billion dollar companies wondering the same thing.

This is a scam in the same way solar roadways is a scam. It is a neat idea that doesn't make sense economically.

**Not really.  Its been reviewed by several PhDs - in fact our senior scientist has his doctorate.  There is a very sensible and viable business model to this, I'm sure you can come up with it if you spend a minute or two.


How much do you expect a modified refrigerator/air conditioner would cost/save? I really doubt you will recycle enough energy to justify the costs whatever they may be.  

**The refrigerators are mass produced today - every RV, most boats and many, many off grid homes already use absorptive refrigeration... it is safe, economical and completely silent.  When you take into account that your refrigerator is the device that runs the most in your house, you'll likely rethink that last statement.  

You might be able to save a few hundred dollars if you heated your water tank for a year using an asic, but your contraption cost would probably negate those savings. Not many people mine for more than a few months and those who do have very low electricity costs which makes recycling pointless.

**The energy you recovered alone would likely reimburse the cost of the system in under 3 years once they are in full production, remember you are using this energy already, we're just talking about using the same energy for two or more purposes - Heating/Cooling/Computation.  If you take into account Moore's law we'll very likely have good reason to replace the units every 3-4 years, as computing gets smaller and more efficient we can pack more and more of it into the same appliance.  I'm really not interested in heating water with ASIC - there is a much more robust business model in selling distributed computation that doesn't rely on a race to the bottom from a technology perspective.

The real savings would come from space heating but you don't need a rube goldberg machine for that.

**The savings can come from far more than space heating - take a look at the website and think about the energy balance of the house.  We will eventually address more than 70% of the load in the average house through space heating, water heating, air conditioning and refrigeration.  You don't have to be rube goldberg to see that.


Ignoring the fact that most modern datacenters use ~10-30% for cooling and that converting all their air cooled hardware to water cooled would cost an obscene amount of money, how do you plan on moving several hundred KW worth of heat from a datacenter to a place that needs it?

**Lets not ignore the fact that 'modern datacenters' make a very, very small portion of the overall population of datacenters which have been around since the 90s at scale.  The vast majority of datacenters are embedded in office buildings and rely on the HVAC system or CRAC units for cooling... you can bet they are not using 10-30% total energy for cooling.

I don't doubt it can be done but I really doubt it can be done cost effectively.

**That's our job to figure out, not yours.



The question is what portion? Have you done calculations on how much electricity you might be able to recycle? I'd love to see a cost/savings breakdown.

**We have - think along the lines of 70%  to start with and then think about how many of the appliances that make up the remaining 30% could be completely obviated if you actually had a house with a super computer built into its core.  We're not there... but we will get there.


But if 90% of that energy is spent on space heating, why not use a $25 heatsink+fan instead of a $300 rube goldberg machine and another ~$1000 for a modified AC+refrigerator?

**Only 40% at best is used for space heating with another 18% used for water heat.  Of course we'll start there, its the easiest and most cost effective loads to get to in a house!  But, if you are going to purpose build a house around a super computer why exactly wouldn't you install the $1000 refrigerator or the $2000 air conditioner?  Remember, the more you use these computation powered devices the more computation that has to be done in order to create the heat... if we are selling the computation to Cloud Services Aggregators & Brokerages then that makes everyone happy.

Basically you can recycle ~$2700 without modifying your hardware by space heating, or you can spend an extra ~$300 to save $3000(~$300 extra) by incorporating water heating.

If you can do that we have some friends at NREL who would love to have a talk with you.  Remember, computers aren't designed to run hot they are built to be run as cool as practicable (dont worry, NREL couldn't get their heads wrapped around the concept at first and they are ALL engineers and PhDs) and today's computers certainly aren't optimized to get to the 180-200f that we need to do refrigeration or air conditioning. 

Thanks much for the good questions, we need to get prepped for these conversations as we start to roll out the prototypes.

They pull a TON of juice dude just to keep them going. You know what else they pull juice for (on a separate input?). Cooling. It's constant and it's REALLY expensive. Do I have to get into the principles of thermodynamics for the cooling part on this? As long as we can produce the heat onsite within a certain range (and also depending on the system type), we can cool the stacks without drawing a separate load to do the same thing. Also, the remaining waste heat can be used to heat the building, water tanks, etc.

Hey there, no one is asking you to buy anything and, as a matter of fact, your reaction is exactly what we were hoping to elicit.  This isn't a scam, its a working device that heated my 1100sf apartment over the New York winter while it was mining Doge and folding for Stanford during testing.

As for the technical questions, here's a shot: what will it do when the room is heated, it'll heat water for your shower, what will it do when the room and your water is heated, it'll run your refrigerator.  What will it do when your water is heated, your refrigerator is cold and the house is too hot - it'll regenerate (heat) a desiccant air conditioning system and cools the house.  

The economic model - your a whiz, maybe this will make sense: instead of spending 60% of the energy bill in a data center somewhere cooling the machines think about moving the machines to where there is a need for heat

The cool (figuratively speaking) part is, now you aren't paying an gas/electric bill (or a portion at least) just to heat your house/water/etc

Do you think this might extend the payback of the mining rig a bit when you offset the roughly 3k a year the average residence spends a year to heat or cool water/space in your house?  Do you think it makes even more sense since you are no longer running the AC to cool the miner... like the data centers do 24/7?  There is a pony in there... think it over.  Talk about an absurd idea, sticking a miner in your house and turning on the AC to keep it cool was one of the most absurd ideas of all - yet how many people were doing that last year?

For one, everything you and your website says is worded like a scam.

For two, this is the most absurd idea I've heard in my life.

This is a rube goldberg machine disguised as a futuristic innovative cooling/heating thingy.

Subtracting all the bullshit and obfuscation this is how I understand your contraption: A watercooling loop delivers heated water to an insulated tank which then goes through a radiator to heat the air when you need it.

What happens when the tank is full of hot water and the cpu needs cool water? Will it shut off the cpu? underclock? heat the room when it's is already heated?

Why would anyone buy this contraption for presumably hundreds of dollars when a $20 cpu+fan does essentially the same thing?

Wise of you to go with kickstarter. This is exactly the type of scam they eat up.


I'm not sure what concept you are proving or even what you are trying to say. Seems like you're just circumventing the question.

Watercooling is a widely proven technology. You've just added a water tank that stores the hot water and cools it later.

Why not just heat the air by computing when you need heating instead of when you don't? (via a $20 heatsink+fan combo)

Honestly what's the point?


Please do share. Preferably without the obfuscation.

For one: Scams usually ask for stuff ($ mainly). We don't. We're just looking to talk with people. If an investor comes along, we'll talk to them too (we are). So let's settle down on that. If you are someone who doesn't want to give your money, rest assured, we don't want it.

For two: thanks for the feedback. It's why I'm here.

It's a proof of concept: not a product ready for sale, distribution, etc. The question really isn't where do you get more cold water - it's how do make sure the material components heating the water through heat exchange can handle that type of long term stress AND make the thing work over a larger distributed network so when the demand for heat is there, there is data to fire up the components to provide heat.

The concept also has really strong applications with datacenters, but I think that's a different discussion.

I actually have some good build pics and screen grabs of testing....but I'm at a loss as to how to post pics on here. Lil help? Anyone?

For one, everything you and your website says is worded like a scam.

For two, this is the most absurd idea I've heard in my life.

This is a rube goldberg machine disguised as a futuristic innovative cooling/heating thingy.

Subtracting all the bullshit and obfuscation this is how I understand your contraption: A watercooling loop delivers heated water to an insulated tank which then goes through a radiator to heat the air when you need it.

What happens when the tank is full of hot water and the cpu needs cool water? Will it shut off the cpu? underclock? heat the room when it's is already heated?

Why would anyone buy this contraption for presumably hundreds of dollars when a $20 cpu+fan does essentially the same thing?

This....is nothing like what we are doing.

Did you visit our site? Or are you knee jerking it?

We are in a very early R&D phase. Submit a contact form and I'll get you on the list to be notified when the Kickstarter sets off.