Quote from: jimmothy on September 10, 2014, 08:14:27 AM

Quote from: ltorsini on September 10, 2014, 05:41:18 AM

Quote

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.

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.

Quote

**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.

That's the problem. 3-4 years is ~50-60 years in bitcoin time.

**Keep in mind the energy for performing computation was already just being used just for heat so performing the computation (from an energy perspective) is free. That dramatically changes the economics.

Quote

Quote

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.

Thinking about it from a non-bitcoin mining point of view I can actually see some use for your idea.

With bitcoin mining you need giant a warehouse in the middle of nowhere with dirt cheap electricity and maximum cooling. It would take a massive refrigerator to reuse 10+MW worth of heat.

**Consider the average Joe isn't going to the Arctic circle for free cooling and there isn't any cheap electricity to be had there at any rate. We're talking about distributing that 10MW worth of heat (computing actually) to multiple average Joe's house. This is a model where he could use the energy he would have used anyway where he is today to to create computation. Like I mentioned, mining is only one potential revenue for the waste computation. Even if you make a fraction of what you might with the latest rig, its still revenue.

Quote

Quote

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.

Quote

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.

I meant 90% of the energy used for heating is used for space heating. The graphs on your website would suggest about 60% of the energy used for heating is used for space heating in 2009. I'm sure the efficiency of absorption cooling has improved in the past 5 years and will continue to improve but heating technology will not so I think we will eventually see 90% being used for heating.

**if we have our way 90% of it will be used to do all those things at once, stacking the energy benefits on top of each other. The concept behind Exergy is not saving energy, it is getting more work of the energy you are already using... its tricky to wrap the head around.

Quote

Quote

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.

No problem. Another question for you:

How will you be heating water up to ~90C while keeping the asics/cpus/gpus cool? Water cooling normally has a ~10C dT between the water and the chip so if you need 90C water you will have a 100C chip. They are capable of working at that temp but depending on the design it will have an efficiency loss of ~0.25-1%/C.

So if you need to heat the water an extra 40C you could be using 40%-10% more power meaning you will need more hardware. The production costs of the most advanced bitcoin asics (starting from wafers) are ~$0.4/w so the costs are not insignificant. (although the costs are guaranteed to go down as process nodes shrink)

**We're more interested in keeping them hot, not cool. Remember the benefit is derived by using these components as primary heaters and looking at the computation as a secondary benefit or 'waste stream' from the heat we need to produce anyway. I think I mentioned in a post above that Moore's law will take care of some of the inefficiency - we're asking you to fundamentally change the way you think about computing. Just like automobiles moved through their adoption cycle from the toys of the wealthy to Henry Fords car for the common man, computation is and will continue to become less expensive and more integrated into everything we do - we're saying lets jump start that (like Ford did) by building it into the most important things that consume the most raw energy. Marrying the things we need to survive (heat/cold) to the things we need to thrive... massive scale, distributed computation.

We're actually boiling water with off the shelf parts today - how much heat do you think we can generate when we design our own chips that are meant to get hot? Really hot?

Quote from: ltorsini on September 10, 2014, 05:48:44 AM

Quote from: jimmothy on September 10, 2014, 04:39:54 AM

Quote from: adesanctis on September 09, 2014, 11:06:52 PM

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.

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.

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.

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!