The conventional wisdom is that liquid cooling is not cost-effective and is only used in "money is no object" applications like mainframes and Cray supercomputers.
Bitfury is not a dumb company. And the power consumption per square foot of their equipment is drastically higher than almost any high performance computing project that preceded it
Bitcoin mining is so odd.. BitFury's going to add a power plant's worth of mining to secure the blockchain and nothing will change at all -- blocks will come at the same pace, the transaction limit won't increase, the network won't be appreciably more secure... what a waste of energy. 40MW of super clean power that could be manufacturing solar panels or smelting aluminum or any of a million better options is instead performing useless calculations to pay back the cost of building this silly infrastructure.
The energy will go towards securing $6B+ of electronic currency. The more players like BitFury that work to secure the blockchain, the harder it is for a single player to attack the chain with a 51% attack.
Anyway, they're going to spend some $10M to build this data center.. High end cards can do something like 0.6GH/W, so this place will add 24M gigahash/S to the network.
The current hashrate is something like 650M GH/S, so we can use a literal powerplant worth of energy to increase the hashrate / security of the blockchain by something like 3.5% -- And the hamster wheel keeps on turning.
There are some others at ~1.6 GH/W, which is the inverse of 0.6 W/GH, so I'm assuming you just got the units backwards.
This is a next generation facility so it's probably closer to 3.65 than 1.6 (if not >4), but even 1.6 would put it at 64 M GH/s, which would be a ~10% increase. 3.65 is 22%
What is a bitcoin if not a certificate that you've increased entropy?
ie: a bitcoin is proof that you converted a bunch of energy into heat via meaningless calculations.
"The ability to create densely packed circuit boards paired with high-performance during overclocking made 2PIC a natural fit for Blockchain transaction processing, an industry with a need for processing power that is growing at unprecedented rates."
Er, no. "Blockchain transaction processing" for Bitcoin requires a low-end laptop with a modest database to handle the 7 TPS transaction rate. If some group of banks sets up a blockchain to track bond transfers, the compute power required will be modest. They don't need to "mine", just limit membership to a finite set of organizations. Remember, if anybody fakes a transaction, everybody knows about it immediately. The purpose of mining is to provide enforcement in an environment where you can't sue anybody because you're not sure who they are.
Also, Bitcoin mining systems aren't really "data centers". There's very little data flowing around. Once you start up a miner, you don't need to talk to it again unless it finds a hit, or a new block comes along and you have to restart all the miners from a new starting point. Some of the earlier Bitcoin mining operations had tons of cabling, with a cable connecting each miners to some master machine, but the newer ones seem to have the cabling problem under control.[1] This makes liquid cooling easier, because there's not a lot of wiring coming out of the liquid cooled components.
They say it's still cheaper and less energy intensive than credit card processing. So it's a step in a greener direction. We don't have to get to the final solution in the first try. Dogecoin, etc have their own shortcomings that make them far worse than bitcoin for exchange and storing value.
I'm curious about the "two phase immersion cooling technology". So, they're putting the circuit boards into some kind of liquid, and the waste heat from the boards is causing that liquid to boil? What kind of liquid are they using?
16 comments
[ 2.9 ms ] story [ 37.0 ms ] threadThe Cray-2 was liquid cooled in 1985: https://en.wikipedia.org/wiki/Cray-2#Packed_circuit_boards_a...
The energy will go towards securing $6B+ of electronic currency. The more players like BitFury that work to secure the blockchain, the harder it is for a single player to attack the chain with a 51% attack.
Anyway, they're going to spend some $10M to build this data center.. High end cards can do something like 0.6GH/W, so this place will add 24M gigahash/S to the network.
The current hashrate is something like 650M GH/S, so we can use a literal powerplant worth of energy to increase the hashrate / security of the blockchain by something like 3.5% -- And the hamster wheel keeps on turning.
3.65 GH/W: https://bitmaintech.com/productDetail.htm?pid=00020151117034...
There are some others at ~1.6 GH/W, which is the inverse of 0.6 W/GH, so I'm assuming you just got the units backwards.
This is a next generation facility so it's probably closer to 3.65 than 1.6 (if not >4), but even 1.6 would put it at 64 M GH/s, which would be a ~10% increase. 3.65 is 22%
It's the opposite of a carbon tax.
Er, no. "Blockchain transaction processing" for Bitcoin requires a low-end laptop with a modest database to handle the 7 TPS transaction rate. If some group of banks sets up a blockchain to track bond transfers, the compute power required will be modest. They don't need to "mine", just limit membership to a finite set of organizations. Remember, if anybody fakes a transaction, everybody knows about it immediately. The purpose of mining is to provide enforcement in an environment where you can't sue anybody because you're not sure who they are.
Also, Bitcoin mining systems aren't really "data centers". There's very little data flowing around. Once you start up a miner, you don't need to talk to it again unless it finds a hit, or a new block comes along and you have to restart all the miners from a new starting point. Some of the earlier Bitcoin mining operations had tons of cabling, with a cable connecting each miners to some master machine, but the newer ones seem to have the cabling problem under control.[1] This makes liquid cooling easier, because there's not a lot of wiring coming out of the liquid cooled components.
[1] https://bitcointalk.org/index.php?topic=1072474.0