If formfactors and power levels were standardized so swapping boards on waterblocks wasn't basically a customized rebuild with every successive generation - like how S3 waterblocks would fit on an S5 - it'd be a lot easier.
Saying "it's too difficult right now" should be a vote in the direction of "make this less difficult" rather than "don't do it at all" because, quite frankly speaking, bitcoin mining is turning into an ecological disaster and there really needs to be a way to head that off. Harvesting waste heat for other projects would really help. To do something like heat water for sous vide brisket or desalination or a Turkish Bath, or heating air for greenhousing in cold climates or whatever else - it's worth doing, and just because the infrastructure isn't trivial right now doesn't mean it should stay that way. The effluent of shortsighted gain has been wrecking the planet for quite long enough, and recycling energy is a good way to counter that.
Also, and I'm with Phil on this one - brisket. Brisket, guys.
Topic: Anyone building hotter exhaust temp ASICs? (Read 207 times)
Heat desalination uses partial vacuums to reduce the boiling point. By decreasing the pressure in consecutive stages, you can heat water once to 80-90C and it'll boil off in one stage, then boil off again in the next, all the way to the last (lowest temperature/pressure) stage.
Didn't know that, the last one I saw (which was back in the 1980s) just boiled the water off, collected and condensed the steam back to water and that was it. Guess things change in 40+ years :-)
Since you and jstefanop build these for a living I'll take you at your word that it's doable. I just think doing all of this to do something with the waste heat other then general heat the office / home / whatever is going to be more difficult then it's really worth. More plumbing, more things to go wrong and so on. A fan either works or does not and it's quick and simple to replace.
-Dave
Heat desalination uses partial vacuums to reduce the boiling point. By decreasing the pressure in consecutive stages, you can heat water once to 80-90C and it'll boil off in one stage, then boil off again in the next, all the way to the last (lowest temperature/pressure) stage.
You wouldn't be able to get much higher exhaust heat from air cooled devices even if you ran ASIC temps higher, air is such a bad carrier of heat it requires ALOT of it to move significant heat (ie high RPM fans) which means that air will be significantly cooler than what its trying to cool.
There are some silicon formulations in fabs that can be run over 100c continually, and here it gets interesting because you can use distilled water to cool the ASICs directly and actually boil the water, steam is a much more useful wast heat. We are playing around with some of these types of ASICs.
There are some silicon formulations in fabs that can be run over 100c continually, and here it gets interesting because you can use distilled water to cool the ASICs directly and actually boil the water, steam is a much more useful wast heat. We are playing around with some of these types of ASICs.
Could do desalination as well. A lot of desalination plants are set up next to industrial facilities to utilize waste heat.
Something for California to consider.
But they won't
Yes, but remember desalination requires the water to boil, so you are still probably going to still have to add some more energy to get the water that hot.
And, side note, most modern desalinization plants use filters instead.
Back to this.
Getting miners to run hotter would start to limit a lot of things.
Right now, you can repurpose older data centers for mining and when you are done still use them as regular data centers.
Or you can put a bunch of miners in a large enough warehouse so long as you can cool them.
And so on.
If you start setting them up where they REQUIRE all this extra cooling because of running at such high temperatures, you are going to be forcing people into only specific builds for the equipment.
That could kill a lot of business plans, since the question would be asked, what do we do with the building after....
-Dave
Could do desalination as well. A lot of desalination plants are set up next to industrial facilities to utilize waste heat.
Something for California to consider.
But they won't
Could do desalination as well. A lot of desalination plants are set up next to industrial facilities to utilize waste heat.
I suppose that I should point out that the OP's ideas regarding using the heat from miners to generate power is not something new. Using comparatively low temperature waste process heat to generate power has been done for decades. Just do a search for 'waste heat co-generation'.
Those Bitcoin Maxi Carnivores are going to love this Brisket idea!
But I was thinking more along the lines of converting thermal energy into mechanical or electrical if the exhaust was hotter.
Actual implementation of anything like this is going to take 100s of millions of dollar or even more, so not really expecting anyone to just come up with a solution.
But if we can motivate an ASIC hardware manufacturer to have recoverable heat that can be converted to work, that will be huge.
It will also address some of the environmentalist's concerns who think Proof of work mining is doing useless work and just converting electricity to heat.
But I was thinking more along the lines of converting thermal energy into mechanical or electrical if the exhaust was hotter.
Actual implementation of anything like this is going to take 100s of millions of dollar or even more, so not really expecting anyone to just come up with a solution.
But if we can motivate an ASIC hardware manufacturer to have recoverable heat that can be converted to work, that will be huge.
It will also address some of the environmentalist's concerns who think Proof of work mining is doing useless work and just converting electricity to heat.
you could slow cook tons of food in an industrial plant.
you could heat apartment house hot water.
But getting chips to do 450c or more is not very likely .
that would be cool if you could since 450c would do a turbine.
but solar power plant and chips heat water to 100c. then run the 100c water through mirror array would give you superheated steam for 4 or 5 of 24 hours.
https://www.ebay.com/itm/394348147736?
So while the sun is up you could generate a steam driven turbine.
hmm actually sounds like a fun idea.
Those Bitcoin Maxi Carnivores are going to love this Brisket idea!
But I was thinking more along the lines of converting thermal energy into mechanical or electrical if the exhaust was hotter.
Actual implementation of anything like this is going to take 100s of millions of dollar or even more, so not really expecting anyone to just come up with a solution.
But if we can motivate an ASIC hardware manufacturer to have recoverable heat that can be converted to work, that will be huge.
It will also address some of the environmentalist's concerns who think Proof of work mining is doing useless work and just converting electricity to heat.
But I was thinking more along the lines of converting thermal energy into mechanical or electrical if the exhaust was hotter.
Actual implementation of anything like this is going to take 100s of millions of dollar or even more, so not really expecting anyone to just come up with a solution.
But if we can motivate an ASIC hardware manufacturer to have recoverable heat that can be converted to work, that will be huge.
It will also address some of the environmentalist's concerns who think Proof of work mining is doing useless work and just converting electricity to heat.
Actually spondollies did 120c if IIRC.
...
Along that line, mining chips are designed to run most efficiently in a fairly narrow range of temperatures. In the case of mining chips it is very hot compared to other kinds of chips - the only others designed for high temperature applications are for automotive & aerospace uses. Back to miners, the old Spondollies are a perfect example of that. Their chips were designed to run at such a high temp because the miner form factor dictated a 2U high rack mount which meant cooling issues. They worked quite well BUT very often needed a pre-heat from an external source (a hair dryer) before they would even start......
That narrow temp constraint is still true today. It's why most miners start with a minute or 2 of high power operation to get them warmed up then restart to begin applying the configured speed & Vcore parameters. The other way would be like how Kano's drivers for Sidehacks Compac's does it: They force the miner to start at a fixed low speed and ramp up in steps with each having to satisfy performance checks before going higher to finally reach configured speed giving the chips the needed pre-heat levels.
On cooking, remember for medium rare beef target temp is 140F (60C) so using a hot-side fluid temp of around the 70-80C temp chips want to run at put through a secondary heat exchanger loop to suvee meats would be quite doable...
I understand that the efficiency of current ASIC chips drops drastically at higher temperatures, so increasing exhaust fluid temp seems like a bad idea.
Current ASICs designs are all essentially treating exhaust heat as a nuisance that needs to be discharged.
While this has worked in the past, as this space evolves, anyone who is able to reuse the 'waste' heat will be more competitive than other miners.
The relatively low exhaust temps that we see coming out of ASICS, has restricted the use of this energy to space heating, drying type of uses.
If the exhaust temps were much higher, you could run an engine off of it, to produce mechanical or electrical energy from it. Those are forms that are easier to transmit and store.
Current Silicon based semiconductor and PCB designs can handle much higher temps than <85C temp that's recommended on Antminer S19. In fact, Antminer S9s used to run much hotter than the new-gen miners.
New semiconductor substrates like GaN, SiC & GaAs might be able to sustain even higher temperatures.
Current ASICs designs are all essentially treating exhaust heat as a nuisance that needs to be discharged.
While this has worked in the past, as this space evolves, anyone who is able to reuse the 'waste' heat will be more competitive than other miners.
The relatively low exhaust temps that we see coming out of ASICS, has restricted the use of this energy to space heating, drying type of uses.
If the exhaust temps were much higher, you could run an engine off of it, to produce mechanical or electrical energy from it. Those are forms that are easier to transmit and store.
Current Silicon based semiconductor and PCB designs can handle much higher temps than <85C temp that's recommended on Antminer S19. In fact, Antminer S9s used to run much hotter than the new-gen miners.
New semiconductor substrates like GaN, SiC & GaAs might be able to sustain even higher temperatures.
Actually spondollies did 120c if IIRC.
But that is not superheated steam it would be more or less slightly overheated steam at best.
So pretend you do that what good is 100-110c hot water/liquid? Is the benefit of having a hotter liquid (105c) outweigh the better power use at 80c for a mining chip?
edit
I did come up with this idea
slow cooked brisket. you can cook brisket at 210 f to 225 f. 10 to 14 hours it is very good.
So liquid cool the chips and you build a heat pump to radiator oven setup that is on view you slow cook the best brisket possible.
Of course it would need to be tons of brisket if you have a big farm
you need a bigger scale than this one
I understand that the efficiency of current ASIC chips drops drastically at higher temperatures, so increasing exhaust fluid temp seems like a bad idea.
Current ASICs designs are all essentially treating exhaust heat as a nuisance that needs to be discharged.
While this has worked in the past, as this space evolves, anyone who is able to reuse the 'waste' heat will be more competitive than other miners.
The relatively low exhaust temps that we see coming out of ASICS, has restricted the use of this energy to space heating, drying type of uses.
If the exhaust temps were much higher, you could run an engine off of it, to produce mechanical or electrical energy from it. Those are forms that are easier to transmit and store.
Current Silicon based semiconductor and PCB designs can handle much higher temps than <85C temp that's recommended on Antminer S19. In fact, Antminer S9s used to run much hotter than the new-gen miners.
New semiconductor substrates like GaN, SiC & GaAs might be able to sustain even higher temperatures.
Current ASICs designs are all essentially treating exhaust heat as a nuisance that needs to be discharged.
While this has worked in the past, as this space evolves, anyone who is able to reuse the 'waste' heat will be more competitive than other miners.
The relatively low exhaust temps that we see coming out of ASICS, has restricted the use of this energy to space heating, drying type of uses.
If the exhaust temps were much higher, you could run an engine off of it, to produce mechanical or electrical energy from it. Those are forms that are easier to transmit and store.
Current Silicon based semiconductor and PCB designs can handle much higher temps than <85C temp that's recommended on Antminer S19. In fact, Antminer S9s used to run much hotter than the new-gen miners.
New semiconductor substrates like GaN, SiC & GaAs might be able to sustain even higher temperatures.
...
What I don’t understand in your equation is, that there is not the input power needed. A miner will put out 99% of the input power out as heat. A steam or combustion engine will not convert 99% of input power into heat.
Wrong. Even the minuscule amount of energy that 'leaks out' as data over the LAN becomes heat.What I don’t understand in your equation is, that there is not the input power needed. A miner will put out 99% of the input power out as heat. A steam or combustion engine will not convert 99% of input power into heat.
100% of the power feeding a miner is turned into heat. Even the energy transferred to the air by the fans increasing the air pressure so it moves also slightly raises it temperature.
The energy feeding any device cannot just disappear. Even when used to cause motion or whatever, at the end of the chain it eventually it becomes heat.
As for the OP's equation - that deals with the amount of usable work energy that can be recovered from the energy put in. 'Usable' being defined as being able to produce work be it powering a pump, moving a vehicle, etc. or generating electricity so the recovered energy can be used elsewhere. A miner does not do that. It just sucks in energy and uses it to power circuitry for the calculations & runs other sundry things like the fans. Since aside from running the fans no physical effort is created all the input energy consumed is directly transformed into heat
You can run the exhaust heat of the miner into the radiator of a heat pump and get much higher temperatures if you put in relatively low energy into the heat pump. Not shure however if this information will benefit you in any way, since steam engines are at another level in comparison.
What I don’t understand in your equation is, that there is not the input power needed. A miner will put out 99% of the input power out as heat. A steam or combustion engine will not convert 99% of input power into heat.
What I don’t understand in your equation is, that there is not the input power needed. A miner will put out 99% of the input power out as heat. A steam or combustion engine will not convert 99% of input power into heat.
You increase the heat output by using more power - so it makes no sense to do that.
Electronics / ICs tend to perform worse the hotter they get so making them run hotter for the most part does not help.
Adding complexity through things like water cooling / oil immersion are all designed to get the heat away from the chips as quickly as possible.
Making them run hotter is not the goal. Running them cooler is the goal.
You can use the exhaust heat now for heating rooms and such, but to generate enough to do something else does not work with this kind of equipment.
Not to mention, it adds complexity and expense.
-Dave
Adding complexity through things like water cooling / oil immersion are all designed to get the heat away from the chips as quickly as possible.
Making them run hotter is not the goal. Running them cooler is the goal.
You can use the exhaust heat now for heating rooms and such, but to generate enough to do something else does not work with this kind of equipment.
Not to mention, it adds complexity and expense.
-Dave
Recovering exhaust heat from ASICs can add additional revenue streams for Bitcoin miners.
As mining becomes more competitive, this becomes increasingly important.
Carnot Efficiency is the theoretical maximum % of the energy that can be recovered from a heat source.
Current ASICs exhaust heat at a much lower temperature when compared to Steam Power or Internal Combustion engines.
This limits the Carnot efficiency & usefulness of exhaust heat.
Is anyone building ASIC miners with a hotter exhaust temp & better heat recovery potential?
-Aviral
As mining becomes more competitive, this becomes increasingly important.
Carnot Efficiency is the theoretical maximum % of the energy that can be recovered from a heat source.
Current ASICs exhaust heat at a much lower temperature when compared to Steam Power or Internal Combustion engines.
This limits the Carnot efficiency & usefulness of exhaust heat.
Is anyone building ASIC miners with a hotter exhaust temp & better heat recovery potential?
-Aviral
Jump to: