Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.forth > #27831 > unrolled thread

Green Array energy consumption: a reality check.

Back to article view | Back to comp.lang.forth

Contents

  Green Array energy consumption: a reality check. albert@spenarnc.xs4all.nl (Albert van der Horst) - 2014-01-12 15:18 +0000
    Re: Green Array energy consumption: a reality check. Tristan Plumb <firth@trstn.net> - 2014-01-12 17:02 +0000
      Re: Green Array energy consumption: a reality check. albert@spenarnc.xs4all.nl (Albert van der Horst) - 2014-01-12 17:27 +0000
        Re: Green Array energy consumption: a reality check. Tristan Plumb <firth@trstn.net> - 2014-01-12 22:40 +0000
          Re: Green Array energy consumption: a reality check. albert@spenarnc.xs4all.nl (Albert van der Horst) - 2014-01-12 23:03 +0000
            Re: Green Array energy consumption: a reality check. Paul Rubin <no.email@nospam.invalid> - 2014-01-12 16:54 -0800
    Re: Green Array energy consumption: a reality check. "Rod Pemberton" <dont_use_email@xnohavenotit.cnm> - 2014-01-13 02:45 -0500
      Re: Green Array energy consumption: a reality check. "Rod Pemberton" <dont_use_email@xnohavenotit.cnm> - 2014-01-13 02:54 -0500
        Re: Green Array energy consumption: a reality check. "Rod Pemberton" <dont_use_email@xnohavenotit.cnm> - 2014-01-13 02:57 -0500
      Re: Green Array energy consumption: a reality check. albert@spenarnc.xs4all.nl (Albert van der Horst) - 2014-01-13 12:32 +0000
        Re: Green Array energy consumption: a reality check. "Rod Pemberton" <dont_use_email@xnohavenotit.cnm> - 2014-01-13 09:52 -0500
          Re: Green Array energy consumption: a reality check. mszepien@gmail.com - 2014-01-19 11:01 -0800
            Re: Green Array energy consumption: a reality check. albert@spenarnc.xs4all.nl (Albert van der Horst) - 2014-01-19 19:38 +0000
              Re: Green Array energy consumption: a reality check. mszepien@gmail.com - 2014-01-19 13:40 -0800
                Re: Green Array energy consumption: a reality check. "Rod Pemberton" <dont_use_email@xnohavenotit.cnm> - 2014-01-19 17:50 -0500
                  Re: Green Array energy consumption: a reality check. mszepien@gmail.com - 2014-01-20 12:55 -0800
                    Re: Green Array energy consumption: a reality check. "Rod Pemberton" <dont_use_email@xnohavenotit.cnm> - 2014-01-20 16:34 -0500
                      Re: Green Array energy consumption: a reality check. mszepien@gmail.com - 2014-01-21 04:49 -0800
                        Re: Green Array energy consumption: a reality check. "Rod Pemberton" <dont_use_email@xnohavenotit.cnm> - 2014-01-21 10:48 -0500
                          Re: Green Array energy consumption: a reality check. mszepien@gmail.com - 2014-01-21 14:37 -0800
                            Re: Green Array energy consumption: a reality check. albert@spenarnc.xs4all.nl (Albert van der Horst) - 2014-01-22 19:30 +0000
      Re: Green Array energy consumption: a reality check. "Rod Pemberton" <dont_use_email@xnohavenotit.cnm> - 2014-01-27 16:26 -0500

Page 1 of 2  [1] 2  Next page →

#27831 — Green Array energy consumption: a reality check.

I've recently looked at the Green Array chips with bitcoins
in mind. With the current state of affairs a bitcoin is
approximately worth the expense in energy consumption to
generate it. This looks like a promissing area for the GA144.

There is special hardware to do 300 GigaHash/second at 175 Watts.
Not even considering whether it is even possible to harness
the GA144 to arrive at this level of bulk performance, lets
look at the energy picture.

With a device like that one hash is calculated at the tune of
600 pJ. This is all in, for a device that you can buy right now.
(butterfly bitcoin miner).

A SHA256 costs 64 rounds with about 16 operations, not counting
the rotations (that promise to be very expensive on the GA144).

If we are optimistic and
- totally ignore communication overhead,
- ignore cross wiring involved in the rotations,
- assume that a 32 bit operation costs merely two GA144 cycles
- assume the minimum energy Green arrays mentions for an operation (7pJ).

we arrive at 2*1000 * 7 is 14,000 pJ.

This looses out at a fat order of magnitude.

Groetjes Albert
-- 
Albert van der Horst, UTRECHT,THE NETHERLANDS
Economic growth -- being exponential -- ultimately falters.
albert@spe&ar&c.xs4all.nl &=n http://home.hccnet.nl/a.w.m.van.der.horst

[toc] | [next] | [standalone]

#27834

On 2014-01-12, Albert van der Horst <albert@spenarnc.xs4all.nl> wrote:
> I've recently looked at the Green Array chips with bitcoins
> in mind. With the current state of affairs a bitcoin is
> approximately worth the expense in energy consumption to
> generate it. This looks like a promissing area for the GA144.

> There is special hardware to do 300 GigaHash/second at 175 Watts.
> Not even considering whether it is even possible to harness
> the GA144 to arrive at this level of bulk performance, lets
> look at the energy picture.

> With a device like that one hash is calculated at the tune of
> 600 pJ. This is all in, for a device that you can buy right now.
> (butterfly bitcoin miner).

From the Butterfly Labs FAQ <http://www.butterflylabs.com/faq/>:

| What is the power consumption of the Bitforce SC (ASIC based) units?
|
| Our 65nm chips consume between 4 and 5 watts per gigahash.
| ...
| Our 50 gigahash machine will consume approximately 250 to 300 watts.

5 watt-seconds per gigahash is 5 nJ/hash, 5000 pJ/hash.

> A SHA256 costs 64 rounds with about 16 operations, not counting
> the rotations (that promise to be very expensive on the GA144).

> If we are optimistic and
> - totally ignore communication overhead,
> - ignore cross wiring involved in the rotations,
> - assume that a 32 bit operation costs merely two GA144 cycles
> - assume the minimum energy Green arrays mentions for an operation (7pJ).

> we arrive at 2*1000 * 7 is 14,000 pJ.

> This looses out at a fat order of magnitude.

You're comparing the power consumption of a general purpose computer with
that of ASIC (application specific ic). Fixed rotations and shifts cost
_no_ energy in hardware, and there's no overhead. ASICs win at everything
except flexibility and small quantity cost...

Tristan

[toc] | [prev] | [next] | [standalone]

#27835

In article <slrnld5im2.3h4.st@tumtum.plumbweb.net>,
Tristan Plumb  <firth@trstn.net> wrote:
>On 2014-01-12, Albert van der Horst <albert@spenarnc.xs4all.nl> wrote:
>> I've recently looked at the Green Array chips with bitcoins
>> in mind. With the current state of affairs a bitcoin is
>> approximately worth the expense in energy consumption to
>> generate it. This looks like a promissing area for the GA144.
>
>> There is special hardware to do 300 GigaHash/second at 175 Watts.
>> Not even considering whether it is even possible to harness
>> the GA144 to arrive at this level of bulk performance, lets
>> look at the energy picture.
>
>> With a device like that one hash is calculated at the tune of
>> 600 pJ. This is all in, for a device that you can buy right now.
>> (butterfly bitcoin miner).
>
>From the Butterfly Labs FAQ <http://www.butterflylabs.com/faq/>:
>
>| What is the power consumption of the Bitforce SC (ASIC based) units?
>|
>| Our 65nm chips consume between 4 and 5 watts per gigahash.
>| ...
>| Our 50 gigahash machine will consume approximately 250 to 300 watts.
>
>5 watt-seconds per gigahash is 5 nJ/hash, 5000 pJ/hash.

Well, I thought it wasn't unfair to compare with their best machine,
it may not be in stock, but it is in production.

>
>> A SHA256 costs 64 rounds with about 16 operations, not counting
>> the rotations (that promise to be very expensive on the GA144).
>
>> If we are optimistic and
>> - totally ignore communication overhead,
>> - ignore cross wiring involved in the rotations,
>> - assume that a 32 bit operation costs merely two GA144 cycles
>> - assume the minimum energy Green arrays mentions for an operation (7pJ).
>
>> we arrive at 2*1000 * 7 is 14,000 pJ.
>
>> This looses out at a fat order of magnitude.
>
>You're comparing the power consumption of a general purpose computer with
>that of ASIC (application specific ic). Fixed rotations and shifts cost
>_no_ energy in hardware, and there's no overhead. ASICs win at everything
>except flexibility and small quantity cost...

I know. I'm desperate to find a niche where the GA144 shines. Not here.

>
>Tristan
>

Groetjes Albert
-- 
Albert van der Horst, UTRECHT,THE NETHERLANDS
Economic growth -- being exponential -- ultimately falters.
albert@spe&ar&c.xs4all.nl &=n http://home.hccnet.nl/a.w.m.van.der.horst

[toc] | [prev] | [next] | [standalone]

#27838

On 2014-01-12, Albert van der Horst <albert@spenarnc.xs4all.nl> wrote:
> Tristan Plumb  <firth@trstn.net> wrote:
>>From the Butterfly Labs FAQ <http://www.butterflylabs.com/faq/>:
>>
>>| What is the power consumption of the Bitforce SC (ASIC based) units?
>>|
>>| Our 65nm chips consume between 4 and 5 watts per gigahash.
>>| ...
>>| Our 50 gigahash machine will consume approximately 250 to 300 watts.
>>
>>5 watt-seconds per gigahash is 5 nJ/hash, 5000 pJ/hash.
>
> Well, I thought it wasn't unfair to compare with their best machine,
> it may not be in stock, but it is in production.

All of their 28nm products appear to be in pre-order, and by my reading
of their "Launch Schedule" it's awaiting production, not in production.
(Or perhaps you know more?)

And actually, that is another major factor. The GA144 is fabricated at
180nm, which is a much larger than either 65 or 28nm.

Tristan

[toc] | [prev] | [next] | [standalone]

#27839

In article <slrnld66fl.3h4.st@tumtum.plumbweb.net>,
Tristan Plumb  <firth@trstn.net> wrote:
>On 2014-01-12, Albert van der Horst <albert@spenarnc.xs4all.nl> wrote:
>> Tristan Plumb  <firth@trstn.net> wrote:
>>>From the Butterfly Labs FAQ <http://www.butterflylabs.com/faq/>:
>>>
>>>| What is the power consumption of the Bitforce SC (ASIC based) units?
>>>|
>>>| Our 65nm chips consume between 4 and 5 watts per gigahash.
>>>| ...
>>>| Our 50 gigahash machine will consume approximately 250 to 300 watts.
>>>
>>>5 watt-seconds per gigahash is 5 nJ/hash, 5000 pJ/hash.
>>
>> Well, I thought it wasn't unfair to compare with their best machine,
>> it may not be in stock, but it is in production.
>
>All of their 28nm products appear to be in pre-order, and by my reading
>of their "Launch Schedule" it's awaiting production, not in production.
>(Or perhaps you know more?)

Preorder, that doesn't mean that they aren't being produced. They are just
in big demand. Plus the seller is in a position to accept no bullshitting.
If you want it, you pay now, they'll tell you when you'll get it,
and no way to cancel your order.

As I pointed out the older machines are not power savy. So it is probably
unwise to buy them. That may mean they'll never sell those anymore.
I wouldn't be surprised that is at the heart of their
"Pay now, or bugger off. " policy.


>
>And actually, that is another major factor. The GA144 is fabricated at
>180nm, which is a much larger than either 65 or 28nm.

Of course. It is interesting that they don't seem to worry using
3 generations behind technology.

>
>Tristan
>

Groetjes Albert
-- 
Albert van der Horst, UTRECHT,THE NETHERLANDS
Economic growth -- being exponential -- ultimately falters.
albert@spe&ar&c.xs4all.nl &=n http://home.hccnet.nl/a.w.m.van.der.horst

[toc] | [prev] | [next] | [standalone]

#27843

albert@spenarnc.xs4all.nl (Albert van der Horst) writes:
>>The GA144 is fabricated at 180nm
> Of course. It is interesting that they don't seem to worry using
> 3 generations behind technology.

Supposedly 180nm is where leakage current is the lowest.  They want to
minimize idle power even at a cost in active power, since their approach
involves keeping the cpu cores idle almost all the time, and switching
on and off very quickly.

[toc] | [prev] | [next] | [standalone]

#27852

On Sun, 12 Jan 2014 10:18:17 -0500, Albert van der Horst
<albert@spenarnc.xs4all.nl> wrote:

> I've recently looked at the Green Array chips with bitcoins
> in mind. [...]  This looks like a promissing area for the GA144.

Ok, distantly Forth related...

<OT: Bitcoins>

> With the current state of affairs a bitcoin is
> approximately worth the expense in energy consumption to
> generate it.

FYI, that assumption is only true you're paying an "expense
in energy consumption" to someone else in the first place...

I.e., if you're generating your own power cost effectively, e.g.,
highly efficient diesel generator or solar power generation, or
you're generating power at wholesale prices instead of retail,
you could maximize that energy's value by mining.

As long as you're paying someone else for power and paying
someone else for special processing circuitry, you're unlikely
to recover your financial inputs, at least currently and
into the future.  One or both of those costs needs to be made
neglible to make a profit.

> There is special hardware to do 300 GigaHash/second at 175 Watts.
> Not even considering whether it is even possible to harness
> the GA144 to arrive at this level of bulk performance, lets
> look at the energy picture.

If you were able to (legally) obtain a large quantity of processors
without paying much for them, and you have a cheap source of
electricity, and/or you have a cheap supply of cooling, i.e.,
iceberg or a mine or caves, then it's at least worth looking into
Bitcoin mining for profit.  That's assuming their value will
continue to increase, which may happen, but is doubtful.

Otherwise, I see no point.  Bitcoin has numerous technical flaws,
e.g., exponentially growing public ledger or blockchain, legal
issues and challenges, i.e., simple use constitutes money laundering
under some government laws, e.g., U.S.A., and it doesn't provide
the user with any advantage over the current banking system or
credit/debit cards.  They can be lost or stolen just like cash and
must be converted to/from fiat currency or legal tender, allowing
governments to track activity.  Transactions aren't reversible which
allows for fraud and theft.  Lack of irreversible transactions was
the main reason people converted from cash to traveller's checks (cheques)
and credit cards.  They are anonymous in name only since the "wallet"
or account number (public key) can't easily be linked to you.  But,
the accounts and all transactions are known and public, forever, so
legal agencies can piece together information from standard investigative
methods and confiscate the assets.

There is just too much to cover here.  So, if you're still interested,
i.e., passion, obsession, delusional, moron, gambler, etc.  Here are
some links:

http://www.bitcoin.org
https://bitcointalk.org
https://blockchain.info/
https://blockchain.info/address/"address"
https://github.com/znort987/blockparser
http://bitcoinrichlist.com/top100
http://bitcoincharts.com/
https://en.bitcoin.it/wiki/Bitcoin_Ladder
http://www.bitcoinx.com/profit/
https://en.bitcoin.it/wiki/Power_Calc

P.S.  Yes, it's true that simple usage of Bitcoin constitutes money
laundering under U.S. Federal laws.  It's a non-legal tender transaction
with indeterminable source and destination that didn't occur within
the financial system.  The U.S. has two definitions of money laundering.
That's the second one.  The first is that money comes from an unlawful
activity.  The second one needs *NO* illegal activity other than use.
That's why you can be charged with money laundering from using poker
chips, Canadian dollars used in the US, (etc.) to "transfer" money...


Rod Pemberton

[toc] | [prev] | [next] | [standalone]

#27853

On Mon, 13 Jan 2014 02:45:42 -0500, Rod Pemberton  
<dont_use_email@xnohavenotit.cnm> wrote:

<OT: money laundering due to usage of Bitcoins>

> P.S.  Yes, it's true that simple usage of Bitcoin constitutes money
> laundering under U.S. Federal laws.  It's a non-legal tender transaction
> with indeterminable source and destination that didn't occur within
> the financial system.  The U.S. has two definitions of money laundering.
> That's the second one.  The first is that money comes from an unlawful
> activity.  The second one needs *NO* illegal activity other than usage.
> That's why you can be charged with money laundering from using poker
> chips, Canadian dollars used in the US, (etc.) to "transfer" money...
>

P.P.S.  Someone is likely to bring up bartering as a counter argument.
It's not.  You're supposed to file IRS paperwork to provide the necessary
information (source, destination, value, etc.) for each bartered
transaction to not be considered money laundering under U.S. laws.


Rod Pemberton

[toc] | [prev] | [next] | [standalone]

#27854

On Mon, 13 Jan 2014 02:54:09 -0500, Rod Pemberton  
<dont_use_email@xnohavenotit.cnm> wrote:

> On Mon, 13 Jan 2014 02:45:42 -0500, Rod Pemberton  
> <dont_use_email@xnohavenotit.cnm> wrote:
>
> <OT: money laundering due to usage of Bitcoins>
>
>> P.S.  Yes, it's true that simple usage of Bitcoin constitutes money
>> laundering under U.S. Federal laws.  It's a non-legal tender transaction
>> with indeterminable source and destination that didn't occur within
>> the financial system.  The U.S. has two definitions of money laundering.
>> That's the second one.  The first is that money comes from an unlawful
>> activity.  The second one needs *NO* illegal activity other than usage.
>> That's why you can be charged with money laundering from using poker
>> chips, Canadian dollars used in the US, (etc.) to "transfer" money...
>>
>
> P.P.S.  Someone is likely to bring up bartering as a counter argument.
> It's not.  You're supposed to file IRS paperwork to provide the necessary
> information (source, destination, value, etc.) for each bartered
> transaction to not be considered money laundering under U.S. laws.
>

P.P.P.S  Someone else is likely to bring up U.S. cash as a counter  
argument.
It's not.  It's (mostly) exempt since it's legal tender in the U.S.


Rod Pemberton

[toc] | [prev] | [next] | [standalone]

#27860

In article <op.w9l7ugxd5zc71u@localhost>,
Rod Pemberton <dont_use_email@xnohavenotit.cnm> wrote:
>On Sun, 12 Jan 2014 10:18:17 -0500, Albert van der Horst
><albert@spenarnc.xs4all.nl> wrote:
>
>> I've recently looked at the Green Array chips with bitcoins
>> in mind. [...]  This looks like a promissing area for the GA144.
>
>Ok, distantly Forth related...
>
><OT: Bitcoins>
>
>> With the current state of affairs a bitcoin is
>> approximately worth the expense in energy consumption to
>> generate it.
>
>FYI, that assumption is only true you're paying an "expense
>in energy consumption" to someone else in the first place...
>
>I.e., if you're generating your own power cost effectively, e.g.,
>highly efficient diesel generator or solar power generation, or
>you're generating power at wholesale prices instead of retail,
>you could maximize that energy's value by mining.
>
>As long as you're paying someone else for power and paying
>someone else for special processing circuitry, you're unlikely
>to recover your financial inputs, at least currently and
>into the future.  One or both of those costs needs to be made
>neglible to make a profit.

You say nothing new. Of course standard economic laws apply.

>
>> There is special hardware to do 300 GigaHash/second at 175 Watts.
>> Not even considering whether it is even possible to harness
>> the GA144 to arrive at this level of bulk performance, lets
>> look at the energy picture.
>
>If you were able to (legally) obtain a large quantity of processors
>without paying much for them, and you have a cheap source of
>electricity, and/or you have a cheap supply of cooling, i.e.,
>iceberg or a mine or caves, then it's at least worth looking into
>Bitcoin mining for profit.  That's assuming their value will
>continue to increase, which may happen, but is doubtful.

If you're clever enough to get the gold that is present in
seawater you'll be rich. So ?
If you have a cheap source of electricity, cheaper than the market,
sell it and get rich. No need for a bit coins detour.

<SNIP>
>P.S.  Yes, it's true that simple usage of Bitcoin constitutes money
>laundering under U.S. Federal laws.  It's a non-legal tender transaction
>with indeterminable source and destination that didn't occur within
>the financial system.  The U.S. has two definitions of money laundering.
>That's the second one.  The first is that money comes from an unlawful
>activity.  The second one needs *NO* illegal activity other than use.
>That's why you can be charged with money laundering from using poker
>chips, Canadian dollars used in the US, (etc.) to "transfer" money...

Interesting. The rest of the world is going to use bitcoins and the dollar
has no use but toilet paper.

>
>
>Rod Pemberton

Groetjes Albert
-- 
Albert van der Horst, UTRECHT,THE NETHERLANDS
Economic growth -- being exponential -- ultimately falters.
albert@spe&ar&c.xs4all.nl &=n http://home.hccnet.nl/a.w.m.van.der.horst

[toc] | [prev] | [next] | [standalone]

#27863

On Mon, 13 Jan 2014 07:32:38 -0500, Albert van der Horst  
<albert@spenarnc.xs4all.nl> wrote:
> In article <op.w9l7ugxd5zc71u@localhost>,
> Rod Pemberton <dont_use_email@xnohavenotit.cnm> wrote:
>> On Sun, 12 Jan 2014 10:18:17 -0500, Albert van der Horst
>> <albert@spenarnc.xs4all.nl> wrote:

<OT: Bitcoins>

>>> With the current state of affairs a bitcoin is
>>> approximately worth the expense in energy consumption to
>>> generate it.
>>
>> FYI, that assumption is only true you're paying an "expense
>> in energy consumption" to someone else in the first place...
>>
>> I.e., if you're generating your own power cost effectively, e.g.,
>> highly efficient diesel generator or solar power generation, or
>> you're generating power at wholesale prices instead of retail,
>> you could maximize that energy's value by mining.
>>
>> As long as you're paying someone else for power and paying
>> someone else for special processing circuitry, you're unlikely
>> to recover your financial inputs, at least currently and
>> into the future.  One or both of those costs needs to be made
>> neglible to make a profit.
>
> You say nothing new. Of course standard economic laws apply.
>

You said nothing new either.  I've seen at least four articles
saying a Bitcoin can no longer be mined for less than it's worth.

>>> There is special hardware to do 300 GigaHash/second at 175 Watts.
>>> Not even considering whether it is even possible to harness
>>> the GA144 to arrive at this level of bulk performance, lets
>>> look at the energy picture.
>>
>> If you were able to (legally) obtain a large quantity of processors
>> without paying much for them, and you have a cheap source of
>> electricity, and/or you have a cheap supply of cooling, i.e.,
>> iceberg or a mine or caves, then it's at least worth looking into
>> Bitcoin mining for profit.  That's assuming their value will
>> continue to increase, which may happen, but is doubtful.
>
> If you're clever enough to get the gold that is present in
> seawater you'll be rich. So ?

If you're clever enough to get the salt that is present in
seawater you'll be very rich too.  So?

Both salt and pure water are valuable too.  Both our suggestions
imply the cost of doing so is neglible relative to the return.

> If you have a cheap source of electricity, cheaper than the market,
> sell it and get rich. No need for a bit coins detour.

It's a matter of the relative level of rich.  Think of maize.

If the farmer can grow maize cheaper than competitors, he could
sell it for slightly more of a profit, but will he become rich?
No.  He'll make a slightly larger, but otherwise neglible profit.
Now, if the farmer can convert the grain to a liquor and sell it,
he'll earn 65 times as much per bushel.  The liquor is a far more
valuable product.  A Bitcoin in the future could be the more
valuable product.  A Bitcoin in the future could be worth more
than the electricity cost in the present.  Given that expectation,
it even makes sense to mine Bitcoin when the price of electricity
for mining a coin exceeds the coin's current value...  Of course,
there is no guarantee that Bitcoin won't deflate or stagnate.  It
depends on demand.  The eventual total supply of Bitcoins is fixed
and therefore so are fractional coins.

> <SNIP>
>> P.S.  Yes, it's true that simple usage of Bitcoin constitutes money
>> laundering under U.S. Federal laws.  It's a non-legal tender transaction
>> with indeterminable source and destination that didn't occur within
>> the financial system.  The U.S. has two definitions of money laundering.
>> That's the second one.  The first is that money comes from an unlawful
>> activity.  The second one needs *NO* illegal activity other than use.
>> That's why you can be charged with money laundering from using poker
>> chips, Canadian dollars used in the US, (etc.) to "transfer" money...
>
> Interesting.

I thought so.

> The rest of the world is going to use bitcoins

Unlikely.

I see Bitcoins as just the first of many, many solutions.  In time,
only two or so are likely to be used, and neither will likely be
Bitcoin.  It's very likely alternate solutions will fix technical
implementation issues, fit user demands better such as providing
reversible transactions, and work with governments and banking systems.
IIRC, one U.S. bank has already filed patents on technology very
similar to Bitcoin.

It's very likely in my opinion that governments and banks will use
their full authority to control any useful virtual currencies and/or
the development of them.  They're not going to relinquish monetary
policy to ordinary people of the world any time soon even if the
technology allows for it.  They're not going to forfit their ability
to control crime by preventing money laundering any time soon either.
If a government can't control crime, of what use is it?

> and the dollar has no use but toilet paper.

Does the euro fair any better in your dystopian world?


Rod Pemberton

[toc] | [prev] | [next] | [standalone]

#27955

On Monday, January 13, 2014 3:52:22 PM UTC+1, Rod Pemberton wrote:
> On Mon, 13 Jan 2014 07:32:38 -0500, Albert van der Horst  
> 
> <albert@spenarnc.xs4all.nl> wrote:
> 
> > In article <op.w9l7ugxd5zc71u@localhost>,
> 
> > Rod Pemberton <dont_use_email@xnohavenotit.cnm> wrote:
> 
> >> On Sun, 12 Jan 2014 10:18:17 -0500, Albert van der Horst
> 
> >> <albert@spenarnc.xs4all.nl> wrote:
> 
> 
> 
> <OT: Bitcoins>
> 
> 
> 
> >>> With the current state of affairs a bitcoin is
> 
> >>> approximately worth the expense in energy consumption to
> 
> >>> generate it.
> 
> >>
> 
> >> FYI, that assumption is only true you're paying an "expense
> 
> >> in energy consumption" to someone else in the first place...
> 
> >>
> 
> >> I.e., if you're generating your own power cost effectively, e.g.,
> 
> >> highly efficient diesel generator or solar power generation, or
> 
> >> you're generating power at wholesale prices instead of retail,
> 
> >> you could maximize that energy's value by mining.
> 
> >>
> 
> >> As long as you're paying someone else for power and paying
> 
> >> someone else for special processing circuitry, you're unlikely
> 
> >> to recover your financial inputs, at least currently and
> 
> >> into the future.  One or both of those costs needs to be made
> 
> >> neglible to make a profit.
> 
> >
> 
> > You say nothing new. Of course standard economic laws apply.
> 
> >
> 
> 
> 
> You said nothing new either.  I've seen at least four articles
> 
> saying a Bitcoin can no longer be mined for less than it's worth.
> 
> 
> 
> >>> There is special hardware to do 300 GigaHash/second at 175 Watts.
> 
> >>> Not even considering whether it is even possible to harness
> 
> >>> the GA144 to arrive at this level of bulk performance, lets
> 
> >>> look at the energy picture.
> 
> >>
> 
> >> If you were able to (legally) obtain a large quantity of processors
> 
> >> without paying much for them, and you have a cheap source of
> 
> >> electricity, and/or you have a cheap supply of cooling, i.e.,
> 
> >> iceberg or a mine or caves, then it's at least worth looking into
> 
> >> Bitcoin mining for profit.  That's assuming their value will
> 
> >> continue to increase, which may happen, but is doubtful.
> 
> >
> 
> > If you're clever enough to get the gold that is present in
> 
> > seawater you'll be rich. So ?
> 
> 
> 
> If you're clever enough to get the salt that is present in
> 
> seawater you'll be very rich too.  So?
> 
> 
> 
> Both salt and pure water are valuable too.  Both our suggestions
> 
> imply the cost of doing so is neglible relative to the return.
> 
> 
> 
> > If you have a cheap source of electricity, cheaper than the market,
> 
> > sell it and get rich. No need for a bit coins detour.
> 
> 
> 
> It's a matter of the relative level of rich.  Think of maize.
> 
> 
> 
> If the farmer can grow maize cheaper than competitors, he could
> 
> sell it for slightly more of a profit, but will he become rich?
> 
> No.  He'll make a slightly larger, but otherwise neglible profit.
> 
> Now, if the farmer can convert the grain to a liquor and sell it,
> 
> he'll earn 65 times as much per bushel.  The liquor is a far more
> 
> valuable product.  A Bitcoin in the future could be the more
> 
> valuable product.  A Bitcoin in the future could be worth more
> 
> than the electricity cost in the present.  Given that expectation,
> 
> it even makes sense to mine Bitcoin when the price of electricity
> 
> for mining a coin exceeds the coin's current value...  Of course,
> 
> there is no guarantee that Bitcoin won't deflate or stagnate.  It
> 
> depends on demand.  The eventual total supply of Bitcoins is fixed
> 
> and therefore so are fractional coins.
> 
> 
> 
> > <SNIP>
> 
> >> P.S.  Yes, it's true that simple usage of Bitcoin constitutes money
> 
> >> laundering under U.S. Federal laws.  It's a non-legal tender transaction
> 
> >> with indeterminable source and destination that didn't occur within
> 
> >> the financial system.  The U.S. has two definitions of money laundering.
> 
> >> That's the second one.  The first is that money comes from an unlawful
> 
> >> activity.  The second one needs *NO* illegal activity other than use.
> 
> >> That's why you can be charged with money laundering from using poker
> 
> >> chips, Canadian dollars used in the US, (etc.) to "transfer" money...
> 
> >
> 
> > Interesting.
> 
> 
> 
> I thought so.
> 
> 
> 
> > The rest of the world is going to use bitcoins
> 
> 
> 
> Unlikely.
> 
> 
> 
> I see Bitcoins as just the first of many, many solutions.  In time,
> 
> only two or so are likely to be used, and neither will likely be
> 
> Bitcoin.  It's very likely alternate solutions will fix technical
> 
> implementation issues, fit user demands better such as providing
> 
> reversible transactions, and work with governments and banking systems.
> 
> IIRC, one U.S. bank has already filed patents on technology very
> 
> similar to Bitcoin.
> 
> 
> 
> It's very likely in my opinion that governments and banks will use
> 
> their full authority to control any useful virtual currencies and/or
> 
> the development of them.  They're not going to relinquish monetary
> 
> policy to ordinary people of the world any time soon even if the
> 
> technology allows for it.  They're not going to forfit their ability
> 
> to control crime by preventing money laundering any time soon either.
> 
> If a government can't control crime, of what use is it?
> 
> 
> 
> > and the dollar has no use but toilet paper.
> 
> 
> 
> Does the euro fair any better in your dystopian world?
> 
> 
> 
> 
> 
> Rod Pemberton

To get back on topic, litecoin [https://litecoin.info/] and some of the other altcoins use the scrypt hash function [http://www.tarsnap.com/scrypt.html] which relies on fast acces to a substantial amount of memory as well as processing power. The memory requirement makes it so that ASICs have less of an advantage over CPUs when it comes to generating hashes as fast as possible. Perhaps this could be a better target for the GreenArrays chip, possibly in conjunction with an external memory module?

[toc] | [prev] | [next] | [standalone]

#27956

In article <baf16dab-2d99-4bdf-bbc9-dd3d775a749b@googlegroups.com>,
 <mszepien@gmail.com> wrote:

<SNIP> In the discussion about bitcoins.
>> Rod Pemberton
>
>To get back on topic, litecoin [https://litecoin.info/] and some of the
>other altcoins use the scrypt hash function
>[http://www.tarsnap.com/scrypt.html] which relies on fast acces to a
>substantial amount of memory as well as processing power. The memory
>requirement makes it so that ASICs have less of an advantage over CPUs
>when it comes to generating hashes as fast as possible. Perhaps this
>could be a better target for the GreenArrays chip, possibly in
>conjunction with an external memory module?

Thanks for the reference. The amount of memory in the GA144 is pathetic.
An interface with external memory would require several hops, which
probably make the algorithm unfeasible regards speed.
Even with external interfacing Green Arrays aims for 64 kbyte (!).

[I've pondering the use of Mbyte memory chips, where the cas and ras
lines are not demultiplexed before connecting them to the GA144. That
would allow up to 256 Mbytes, but it is a hardware accomplishment
beyond my capabilities.]

Groetjes Albert
-- 
Albert van der Horst, UTRECHT,THE NETHERLANDS
Economic growth -- being exponential -- ultimately falters.
albert@spe&ar&c.xs4all.nl &=n http://home.hccnet.nl/a.w.m.van.der.horst

[toc] | [prev] | [next] | [standalone]

#27959

On Sunday, January 19, 2014 8:38:26 PM UTC+1, Albert van der Horst wrote:
> In article <baf16dab-2d99-4bdf-bbc9-dd3d775a749b@googlegroups.com>,
> 
>  <mszepien@gmail.com> wrote:
> 
> 
> 
> <SNIP> In the discussion about bitcoins.
> 
> >> Rod Pemberton
> 
> >
> 
> >To get back on topic, litecoin [https://litecoin.info/] and some of the
> 
> >other altcoins use the scrypt hash function
> 
> >[http://www.tarsnap.com/scrypt.html] which relies on fast acces to a
> 
> >substantial amount of memory as well as processing power. The memory
> 
> >requirement makes it so that ASICs have less of an advantage over CPUs
> 
> >when it comes to generating hashes as fast as possible. Perhaps this
> 
> >could be a better target for the GreenArrays chip, possibly in
> 
> >conjunction with an external memory module?
> 
> 
> 
> Thanks for the reference. The amount of memory in the GA144 is pathetic.
> 
> An interface with external memory would require several hops, which
> 
> probably make the algorithm unfeasible regards speed.
> 
> Even with external interfacing Green Arrays aims for 64 kbyte (!).
> 
> 
> 
> [I've pondering the use of Mbyte memory chips, where the cas and ras
> 
> lines are not demultiplexed before connecting them to the GA144. That
> 
> would allow up to 256 Mbytes, but it is a hardware accomplishment
> 
> beyond my capabilities.]
> 
> 
> 
> Groetjes Albert
> 
> -- 
> 
> Albert van der Horst, UTRECHT,THE NETHERLANDS
> 
> Economic growth -- being exponential -- ultimately falters.
> 
> albert@spe&ar&c.xs4all.nl &=n http://home.hccnet.nl/a.w.m.van.der.horst

Well actually, as you alluded in your first post, it's not so much a matter of speed as of energy consumption. Also, I'm not aware of any scrypt ASICS, so the competition would be with CPUs, against which the GA chip should have the upper hand, if it can address sufficient memory.

According to [https://litecoin.info/Comparison_between_Litecoin_and_Bitcoin], the litecoin clients use a scrypt which is configured to use 128kB, to stay within the L2 cache of modern CPUs. This indeed exceeds the total memory on a GreenArrays chip, by my quick calculations (128 words * 18 bits/word * 144 cores / 8 bits/Byte = 41kB) so one would need to use external memory.

I don't understand your 64 kbyte remark, is there any reason you couldn't just interface the chip with, say, a "normal" SDRAM module? (Disclaimer: I know next to nothing about memory hardware) Would that be too complicated to even consider?

[toc] | [prev] | [next] | [standalone]

#27962

On Sun, 19 Jan 2014 16:40:07 -0500, <mszepien@gmail.com> wrote:
> On Sunday, January 19, 2014 8:38:26 PM UTC+1, Albert van der Horst wrote:
>> In article <baf16dab-2d99-4bdf-bbc9-dd3d775a749b@googlegroups.com>,
>>  <mszepien@gmail.com> wrote:

> Well actually, as you alluded in your first post, it's not so much a  
> matter of speed as of energy consumption.

Bitcoins are first come.  I.e., whomever solves a "proof of work" problem
first, gets the coins for it.  So, I'd say speed is very important...


Rod Pemberton

[toc] | [prev] | [next] | [standalone]

#27978

On Sunday, January 19, 2014 11:50:18 PM UTC+1, Rod Pemberton wrote:
> On Sun, 19 Jan 2014 16:40:07 -0500, <mszepien@gmail.com> wrote:
> 
> > On Sunday, January 19, 2014 8:38:26 PM UTC+1, Albert van der Horst wrote:
> 
> >> In article <baf16dab-2d99-4bdf-bbc9-dd3d775a749b@googlegroups.com>,
> 
> >>  <mszepien@gmail.com> wrote:
> 
> 
> 
> > Well actually, as you alluded in your first post, it's not so much a  
> 
> > matter of speed as of energy consumption.
> 
> 
> 
> Bitcoins are first come.  I.e., whomever solves a "proof of work" problem
> 
> first, gets the coins for it.  So, I'd say speed is very important...
> 
> 
> 
> 
> 
> Rod Pemberton

I think that's actually not true, since the proof of work problem involves brute-forcing a hash function and not some sequential algorithm. I think speed of a single hash computation doesn't matter, as long as it stays considerably shorter than the average block confirmation time on the bitcoin network (calibrated to be 10 minutes).

Suppose I have two machines at my disposal. The first consumes an energy E1 to produce 10^6 hashes in time T1; the second takes energy E2 < E1 to produce the same amount of hashes in time T2 > T1.

Assuming that the only cost of the machines is the energy they consume (ie ignoring the acquisition cost), suppose I feed each of them the same total amount of energy, and let them calculate hashes until they have exhausted that resource. At the end of my experiment, machine 2 will have mined more (higher expected value of) bitcoins, by virtue of having calculated more hashes. Therefore, after factoring in energy costs (which are the same for the two machines), depending on the price of bitcoins, it is possible that machine 2 will have turned a profit, while machine 1 did not, even though it calculates hases faster than machine 2.

Accordingly, as far as I can see, any design trade-off which decreases energy consumption per hash at the expense of hash rate is worth taking (again, neglecting the need to recuperate the acquisition cost, and assuming that the hash time stays well below 10 minutes), because it would increase the machine's profitability in terms of bitcoins per kWh.

From Albert van der Horst's initial analysis it seems that GA chips can't compete with ASICs in terms of SHA256 hashes per kWh, but IIRC the ASICs are producing, in USD terms, a multiple of what they cost to run, so I'd say that the GA chips may not necessarily be so energy-inefficient as to be unprofitable.

That said, the GA device probably stands a better chance of being competitive in mining scrypt-based altcoins, where ASICs (and I believe also GPUs) don't have as much of an advantage.

[toc] | [prev] | [next] | [standalone]

#27981

On Mon, 20 Jan 2014 15:55:42 -0500, <mszepien@gmail.com> wrote:
> On Sunday, January 19, 2014 11:50:18 PM UTC+1, Rod Pemberton wrote:
>> On Sun, 19 Jan 2014 16:40:07 -0500, <mszepien@gmail.com> wrote:

>> > Well actually, as you alluded in your first post, it's not so much a
>> > matter of speed as of energy consumption.
>>
>> Bitcoins are first come.  I.e., whomever solves a "proof of work"  
>> problem first, gets the coins for it.  So, I'd say speed is very
>> important...
>
> I think that's actually not true, since the proof of work problem  
> involves brute-forcing a hash function and not some sequential  
> algorithm. I think speed of a single hash computation doesn't matter, as  
> long as it stays considerably shorter than the average block  
> confirmation time on the bitcoin network (calibrated to be 10 minutes).
>
> Suppose I have two machines at my disposal. The first consumes an energy  
> E1 to produce 10^6 hashes in time T1; the second takes energy E2 < E1 to  
> produce the same amount of hashes in time T2 > T1.
>

AIUI, you don't get to choose who your competition is.  Multiple computers,
or groups of them, attempt to solve the exact same proof-of-work at the  
same
time.  I.e., if this was a foot race, it's a sprint with many losers
and one winner.

If my machines are attempting to solve the same proof-of-work as your
machines, and my machines solve it first, you don't get any bitcoins.
I get the bitcoins.  Once a proof-of-work has been solved, any computers
that were attempting to solve that proof of work but failed stop working
that proof and start another.  Now, if your is working a different proof
than my machine, you might have a chance at earning some coin.  But, if
my faster machine is competing against your slower machine, you lose again,
and again, ...

Does it matter to me how much energy I consume?  Yes, but only once
my machines are sufficiently faster than yours so I earn the bitcoin.
If I'm not earning *any* bitcoin because my machines are too slow, it
doesn't really matter whether I have an energy efficient machine or
not.  I'm not receiving any coinage.  All my machines are a loss.

So, AIUI, speed matters very much.


Rod Pemberton

[toc] | [prev] | [next] | [standalone]

#27986

On Monday, January 20, 2014 10:34:33 PM UTC+1, Rod Pemberton wrote:
> On Mon, 20 Jan 2014 15:55:42 -0500, <mszepien@gmail.com> wrote:
> 
> > On Sunday, January 19, 2014 11:50:18 PM UTC+1, Rod Pemberton wrote:
> 
> >> On Sun, 19 Jan 2014 16:40:07 -0500, <mszepien@gmail.com> wrote:
> 
> 
> 
> >> > Well actually, as you alluded in your first post, it's not so much a
> 
> >> > matter of speed as of energy consumption.
> 
> >>
> 
> >> Bitcoins are first come.  I.e., whomever solves a "proof of work"  
> 
> >> problem first, gets the coins for it.  So, I'd say speed is very
> 
> >> important...
> 
> >
> 
> > I think that's actually not true, since the proof of work problem  
> 
> > involves brute-forcing a hash function and not some sequential  
> 
> > algorithm. I think speed of a single hash computation doesn't matter, as  
> 
> > long as it stays considerably shorter than the average block  
> 
> > confirmation time on the bitcoin network (calibrated to be 10 minutes).
> 
> >
> 
> > Suppose I have two machines at my disposal. The first consumes an energy  
> 
> > E1 to produce 10^6 hashes in time T1; the second takes energy E2 < E1 to  
> 
> > produce the same amount of hashes in time T2 > T1.
> 
> >
> 
> 
> 
> AIUI, you don't get to choose who your competition is.  Multiple computers,
> 
> or groups of them, attempt to solve the exact same proof-of-work at the  
> 
> same
> 
> time.  I.e., if this was a foot race, it's a sprint with many losers
> 
> and one winner.
> 
> 
> 
> If my machines are attempting to solve the same proof-of-work as your
> 
> machines, and my machines solve it first, you don't get any bitcoins.
> 
> I get the bitcoins.  Once a proof-of-work has been solved, any computers
> 
> that were attempting to solve that proof of work but failed stop working
> 
> that proof and start another.  Now, if your is working a different proof
> 
> than my machine, you might have a chance at earning some coin.  But, if
> 
> my faster machine is competing against your slower machine, you lose again,
> 
> and again, ...
> 
> 
> 
> Does it matter to me how much energy I consume?  Yes, but only once
> 
> my machines are sufficiently faster than yours so I earn the bitcoin.
> 
> If I'm not earning *any* bitcoin because my machines are too slow, it
> 
> doesn't really matter whether I have an energy efficient machine or
> 
> not.  I'm not receiving any coinage.  All my machines are a loss.
> 
> 
> 
> So, AIUI, speed matters very much.
> 
> 
> 
> 
> 
> Rod Pemberton

The proof of work problem consists of hashing a block containing recent transactions, so that the hash has a certain number of leading zeros (it's a small bit more complicated than that, see https://en.bitcoin.it/wiki/Mining, but this suffices for the sake of argument). In a sense it's true that all machines on the network are trying to solve the same problem, but they are all doing it independently, by altering a nonce appended to the block, in order to try to get the hash value below the target. Thus, the machine which has the highest hash rate will, over time, validate the most blocks and thus mine the most coins, but for any given block, it's a matter of chance which machine happens to produce the first hash that meets the target. Consider the extreme example of an extremely slow machine which can only attempt hash per block, it has a small, but nonzero probability of meeting the target and obtaining new coins (this probability fluctuates and is currently given by http://blockexplorer.com/q/probability).

Consider again my two hypothetical machines. Suppose I run them each for the same amount of time (averaged over a sufficiently large number of blocks), then their relative mining revenue would follow the ratio T1/T2, the faster machine having mined more coins. However, this tells us nothing about the profitability of either machine,  since we haven't factored in the cost of operating them. It is possible that the slower machine is the only profitable one of the two, despite having mined fewer coins.

[toc] | [prev] | [next] | [standalone]

#27989

On Tue, 21 Jan 2014 07:49:12 -0500, <mszepien@gmail.com> wrote:

>> [discussion over importance of hashing speed for Bitcoin]
>
> The proof of work problem consists of hashing a block containing recent
> transactions, so that the hash has a certain number of leading zeros
> (it's a small bit more complicated than that, see
> [link], but this suffices for the sake of
> argument). In a sense it's true that all machines on the network are
> trying to solve the same problem, but they are all doing it
> independently, by altering a nonce appended to the block, in order to
> try to get the hash value below the target. Thus, the machine which has
> the highest hash rate will, over time, validate the most blocks and thus
> mine the most coins, [...]

For that, we're in agreement.

> [...] but for any given block, it's a matter of chance
> which machine happens to produce the first hash that meets the target.

I think that's a faulty assumption to think it's "a matter of chance
as to which machine happens to produce the first hash that meets the
target."

AIUI, the vast majority of machines using to solve this are using
the exact same code and therefore algorithm.  This means they're
solving the same altered nonces in the same manner.  So, even if
they're using a random number generation to "randomly" attempt
different alterations of the nonce, their PRNG starts from and
runs through the exact same psuedo-random sequence.  I.e., first
come, first wins.

For it to be a matter of chance as to which machine validates a block,
each machine must be using a different algorithm or different PRNG in
order to solve a different altered nonce.  I.e., the probability that
multiple machines are attempting the solve the exact same altered nonce
is seems exceptionally large - IMO - due to the use of the same code
or same algorithm.  Therefore, the fastest machine wins.

> Consider again my two hypothetical machines. Suppose I run them each for
> the same amount of time (averaged over a sufficiently large number of
> blocks), then their relative mining revenue would follow the ratio
> T1/T2, the faster machine having mined more coins. However, this tells
> us nothing about the profitability of either machine,  since we haven't
> factored in the cost of operating them. It is possible that the slower
> machine is the only profitable one of the two, despite having mined
> fewer coins.

True, it's possible, although it's unlikely.  Your assumptions are based
on both machines being successful at mining, and having completely
independent chances of successful mining, and it's unknown as to what
the fixed and operating costs are.

Since the profitability of two machines is relative and on-going, you could
have a $50,000 ASIC that's not profitable for 10 years, and a $3  
microcontroller
that's profitable with just one successful mining.  That's assuming the
micro _ever_ has a successful mining.  It's clear that other machines are
out mining $3 machine so it machine may *never* pay off.  Of course, it's
a more likely scenario the two machines are similarly priced or at least
much closer together in prices.  If one is $4,000 and the other is $5,000
and faster, their profitability depends on their rate of successful mining
versus their cost to operate plus their fixed costs ($4k and $5k).  If the
slower machine has lower electricity operating costs, either due to less
consumption or lower prices, it's on-going costs are increasing at a slower
rate than the $5k machine, but it's costs start from a lower base, i.e.,  
$1K
less, so it has the cost advantage.  However, the rate of successful mining
of the $4,000 machine will be lower since it's slower, i.e., it'll be  
recovering
it's fixed cost plus operating cost at a lower rate too.  It's therefore  
more
likely the more expensive $5k machine recovers it's investment and  
operating
first, but determining exactly where, or if so, depends on a graph (or  
algebra)
which accurately drawn costs.  This of course assumes that both machines'
chances for each and every altered nonce, or hash attempt, or mining  
attempt,
is equal and independent of the other machine, which it isn't.  Machines  
use
the same code or same algorithm or same PRNG and attempt to solve the same  
block.


Rod Pemberton

[toc] | [prev] | [next] | [standalone]

#27997

On Tuesday, January 21, 2014 4:48:14 PM UTC+1, Rod Pemberton wrote:
> On Tue, 21 Jan 2014 07:49:12 -0500, <mszepien@gmail.com> wrote:
> 
> 
> 
> >> [discussion over importance of hashing speed for Bitcoin]
> 
> >
> 
> > The proof of work problem consists of hashing a block containing recent
> 
> > transactions, so that the hash has a certain number of leading zeros
> 
> > (it's a small bit more complicated than that, see
> 
> > [link], but this suffices for the sake of
> 
> > argument). In a sense it's true that all machines on the network are
> 
> > trying to solve the same problem, but they are all doing it
> 
> > independently, by altering a nonce appended to the block, in order to
> 
> > try to get the hash value below the target. Thus, the machine which has
> 
> > the highest hash rate will, over time, validate the most blocks and thus
> 
> > mine the most coins, [...]
> 
> 
> 
> For that, we're in agreement.
> 
> 
> 
> > [...] but for any given block, it's a matter of chance
> 
> > which machine happens to produce the first hash that meets the target.
> 
> 
> 
> I think that's a faulty assumption to think it's "a matter of chance
> 
> as to which machine happens to produce the first hash that meets the
> 
> target."
> 
> 
> 
> AIUI, the vast majority of machines using to solve this are using
> 
> the exact same code and therefore algorithm.  This means they're
> 
> solving the same altered nonces in the same manner.  So, even if
> 
> they're using a random number generation to "randomly" attempt
> 
> different alterations of the nonce, their PRNG starts from and
> 
> runs through the exact same psuedo-random sequence.  I.e., first
> 
> come, first wins.
> 
> 
> 
> For it to be a matter of chance as to which machine validates a block,
> 
> each machine must be using a different algorithm or different PRNG in
> 
> order to solve a different altered nonce.  I.e., the probability that
> 
> multiple machines are attempting the solve the exact same altered nonce
> 
> is seems exceptionally large - IMO - due to the use of the same code
> 
> or same algorithm.  Therefore, the fastest machine wins.
> 
> 
> 
> > Consider again my two hypothetical machines. Suppose I run them each for
> 
> > the same amount of time (averaged over a sufficiently large number of
> 
> > blocks), then their relative mining revenue would follow the ratio
> 
> > T1/T2, the faster machine having mined more coins. However, this tells
> 
> > us nothing about the profitability of either machine,  since we haven't
> 
> > factored in the cost of operating them. It is possible that the slower
> 
> > machine is the only profitable one of the two, despite having mined
> 
> > fewer coins.
> 
> 
> 
> True, it's possible, although it's unlikely.  Your assumptions are based
> 
> on both machines being successful at mining, and having completely
> 
> independent chances of successful mining, and it's unknown as to what
> 
> the fixed and operating costs are.
> 
> 
> 
> Since the profitability of two machines is relative and on-going, you could
> 
> have a $50,000 ASIC that's not profitable for 10 years, and a $3  
> 
> microcontroller
> 
> that's profitable with just one successful mining.  That's assuming the
> 
> micro _ever_ has a successful mining.  It's clear that other machines are
> 
> out mining $3 machine so it machine may *never* pay off.  Of course, it's
> 
> a more likely scenario the two machines are similarly priced or at least
> 
> much closer together in prices.  If one is $4,000 and the other is $5,000
> 
> and faster, their profitability depends on their rate of successful mining
> 
> versus their cost to operate plus their fixed costs ($4k and $5k).  If the
> 
> slower machine has lower electricity operating costs, either due to less
> 
> consumption or lower prices, it's on-going costs are increasing at a slower
> 
> rate than the $5k machine, but it's costs start from a lower base, i.e.,  
> 
> $1K
> 
> less, so it has the cost advantage.  However, the rate of successful mining
> 
> of the $4,000 machine will be lower since it's slower, i.e., it'll be  
> 
> recovering
> 
> it's fixed cost plus operating cost at a lower rate too.  It's therefore  
> 
> more
> 
> likely the more expensive $5k machine recovers it's investment and  
> 
> operating
> 
> first, but determining exactly where, or if so, depends on a graph (or  
> 
> algebra)
> 
> which accurately drawn costs.  This of course assumes that both machines'
> 
> chances for each and every altered nonce, or hash attempt, or mining  
> 
> attempt,
> 
> is equal and independent of the other machine, which it isn't.  Machines  
> 
> use
> 
> the same code or same algorithm or same PRNG and attempt to solve the same  
> 
> block.
> 
> 
> 
> 
> 
> Rod Pemberton

I agree with your analysis of the implications if all nodes on the network indeed calculate the same list of hashes in sequence, racing to find the first good one. However, that does not appear to be how it works (I'm far from an expert, but the code is freely available to inspect at https://github.com/bitcoin/bitcoin/blob/master/src/miner.cpp). What happens is that the node's unique public key gets included in the header of the block, giving a unique set of hashes for each node even though they all increment the nonce in the same way. Therefore, I believe that my assumption that each calculated hash (across all nodes) has an equal and independent probability of meeting the target, is correct.

As for profitability, I should have been a bit more clear in that I was talking about expected profitability and not the profit/loss of a specific run. As you point out, it is perfectly possible for a machine that one would expect to turn a profit on the long run, to hit a dry patch and make a loss in any given time interval.

I deliberately left out of consideration the acquisition cost of the machines, since making profitability assessments then requires assumptions about the value of the generated bitcoins over time, which is a tricky business to say the least. However, I agree that it should factor into any serious consideration.

That said, it's plausible that the GreenArray chips would have a far lower cost of acquisition than custom ASICs, so perhaps, in spite of lesser energy efficiency, they might win out over ASICs for long enough that the uncertainty in bitcoin price becomes the dominant factor and all bets, essentially, are off.

[toc] | [prev] | [next] | [standalone]

Page 1 of 2  [1] 2  Next page →

Back to top | Article view | comp.lang.forth

csiph-web