Bitcoin Forum
March 29, 2024, 01:09:30 AM *
News: Latest Bitcoin Core release: 26.0 [Torrent]
 
   Home   Help Search Login Register More  
Pages: [1]
  Print  
Author Topic: How efficient will ASICs get?  (Read 1058 times)
mustyoshi (OP)
Sr. Member
****
Offline Offline

Activity: 287
Merit: 250



View Profile
May 11, 2013, 07:27:46 PM
 #1

It is my understanding that the current generation of ASICs all get around 100MH/joule give or take a few million hashes. While that is upwards of 20x more efficient than GPU mining, how much more can be pushed out? Given we understand exactly the operation needed, does anybody know what the minimum amount of energy that could possibly be expended to perform SHA256?
According to NIST and ECRYPT II, the cryptographic algorithms used in Bitcoin are expected to be strong until at least 2030. (After that, it will not be too difficult to transition to different algorithms.)
Advertised sites are not endorsed by the Bitcoin Forum. They may be unsafe, untrustworthy, or illegal in your jurisdiction.
1711674570
Hero Member
*
Offline Offline

Posts: 1711674570

View Profile Personal Message (Offline)

Ignore
1711674570
Reply with quote  #2

1711674570
Report to moderator
1711674570
Hero Member
*
Offline Offline

Posts: 1711674570

View Profile Personal Message (Offline)

Ignore
1711674570
Reply with quote  #2

1711674570
Report to moderator
erk
Hero Member
*****
Offline Offline

Activity: 826
Merit: 500



View Profile
May 11, 2013, 07:29:18 PM
 #2

It is my understanding that the current generation of ASICs all get around 100MH/joule give or take a few million hashes. While that is upwards of 20x more efficient than GPU mining, how much more can be pushed out? Given we understand exactly the operation needed, does anybody know what the minimum amount of energy that could possibly be expended to perform SHA256?


Do you know what Moore's law is?

tvbcof
Legendary
*
Offline Offline

Activity: 4564
Merit: 1276


View Profile
May 11, 2013, 08:04:46 PM
 #3


Do you know what Moore's law is?


It's associated with Wirth's law.  With permutations such as "Gate's law":

  "The speed of software halves every 18 months."

Now, to be fair, that does not really apply to custom hashing chips.  But the whole goal is to support Bitcoin (or that is the way most people look at it) and Bitcoin as a system displays such behavior in a big way due to the persistent nature of the baggage it needs to pack around.


sig spam anywhere and self-moderated threads on the pol&soc board are for losers.
ecliptic
Sr. Member
****
Offline Offline

Activity: 322
Merit: 250


View Profile
May 11, 2013, 08:51:18 PM
 #4

It is my understanding that the current generation of ASICs all get around 100MH/joule give or take a few million hashes. While that is upwards of 20x more efficient than GPU mining, how much more can be pushed out? Given we understand exactly the operation needed, does anybody know what the minimum amount of energy that could possibly be expended to perform SHA256?


Do you know what Moore's law is?


Something that hardly applies to bitcoins.  Makes sense for the microprocesser market worth tens of billions of dollars (maybe into the hundreds of billions)

When your NRE for an ASIC runs several million or tens of millions of dollars, you have to make a financial decision if it's worth the investment, risk, and design/lead times for a marginal improvement over what already exists

Look at it this way : Asic Miner is a 100% capitalist, maximize profit type of organization.  They used, what, 110 or 130nm node for their fab?  You could get much better power efficency and hashrates with a smaller process, but it's far more expensive and complicated.  They have no plans to change node, they're bringing 262 THash/sec online with their 110nm in the next 3 months.  A massive investment that shows they're dedicated to what they've got.
Littleshop
Legendary
*
Offline Offline

Activity: 1386
Merit: 1003



View Profile WWW
May 11, 2013, 09:17:04 PM
 #5

It is my understanding that the current generation of ASICs all get around 100MH/joule give or take a few million hashes. While that is upwards of 20x more efficient than GPU mining, how much more can be pushed out? Given we understand exactly the operation needed, does anybody know what the minimum amount of energy that could possibly be expended to perform SHA256?

ASIC's could be easily 2x - 4x more power efficient as the best chips out today even using the same NM manufacturing process because as far as I know none of the chips out are true CLEAN SHEET designs.  You can get several orders of a magnitude by having lower NM manufacturing processes as well as a more efficiency from multiple but slower cores per chip.  All of this costs money and lots of it but then makes for a lower cost product in the end.

dogie
Legendary
*
Offline Offline

Activity: 1666
Merit: 1183


dogiecoin.com


View Profile WWW
May 11, 2013, 10:15:53 PM
 #6

It is my understanding that the current generation of ASICs all get around 100MH/joule give or take a few million hashes. While that is upwards of 20x more efficient than GPU mining, how much more can be pushed out? Given we understand exactly the operation needed, does anybody know what the minimum amount of energy that could possibly be expended to perform SHA256?


Do you know what Moore's law is?


Something that hardly applies to bitcoins.  Makes sense for the microprocesser market worth tens of billions of dollars (maybe into the hundreds of billions)

When your NRE for an ASIC runs several million or tens of millions of dollars, you have to make a financial decision if it's worth the investment, risk, and design/lead times for a marginal improvement over what already exists

Look at it this way : Asic Miner is a 100% capitalist, maximize profit type of organization.  They used, what, 110 or 130nm node for their fab?  You could get much better power efficency and hashrates with a smaller process, but it's far more expensive and complicated.  They have no plans to change node, they're bringing 262 THash/sec online with their 110nm in the next 3 months.  A massive investment that shows they're dedicated to what they've got.

You really believe they're making 26,200 blades?

mustyoshi (OP)
Sr. Member
****
Offline Offline

Activity: 287
Merit: 250



View Profile
May 12, 2013, 02:55:54 AM
 #7

It is my understanding that the current generation of ASICs all get around 100MH/joule give or take a few million hashes. While that is upwards of 20x more efficient than GPU mining, how much more can be pushed out? Given we understand exactly the operation needed, does anybody know what the minimum amount of energy that could possibly be expended to perform SHA256?


Do you know what Moore's law is?


Something that hardly applies to bitcoins.  Makes sense for the microprocesser market worth tens of billions of dollars (maybe into the hundreds of billions)

When your NRE for an ASIC runs several million or tens of millions of dollars, you have to make a financial decision if it's worth the investment, risk, and design/lead times for a marginal improvement over what already exists

Look at it this way : Asic Miner is a 100% capitalist, maximize profit type of organization.  They used, what, 110 or 130nm node for their fab?  You could get much better power efficency and hashrates with a smaller process, but it's far more expensive and complicated.  They have no plans to change node, they're bringing 262 THash/sec online with their 110nm in the next 3 months.  A massive investment that shows they're dedicated to what they've got.

You really believe they're making 26,200 blades?
Why wouldn't they? They stand to not only increase the hashrate, but effectively ensure their product will be bought at the same time.
For every blade they sell, another one will be sold to keep somebody else in the mining game.
jspielberg
Sr. Member
****
Offline Offline

Activity: 490
Merit: 255



View Profile
May 12, 2013, 12:58:58 PM
 #8

I would venture to say (armed with absolutely no information at all) that the 50 BTC they are charging for the 10Gh/s block eruptor blades allows them to make 4 new ones to replace it... i.e. it costs them 10BTC to make.  I fully expect them to deploy their 260 Th/s farm, though I think they will move to a more advanced technology/process node whatever, so as to simplify their deployments/logistics.

Deployments for them is expensive as they need a person (or people) to manage the deployment.  The less number of people they need, the less overhead, the more profit.  AM is governed by its ability to be efficient.
Pages: [1]
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2006-2009, Simple Machines Valid XHTML 1.0! Valid CSS!