Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/didnt-see-that-coming-amd-rallies-on-cryptocurrency-mining-demand.9366/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021770
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Didn't see that coming - AMD rallies on cryptocurrency mining demand

(Reuters)
Shares of Advanced Micro Devices surged nearly 9 percent on Tuesday boosted by strong demand for its chips from cryptocurrency miners, leaving short sellers at a loss for the year. A rally in cryptocurrency Ethereum has boosted demand for graphics chips used by people to "mine" it and other digital currencies, with some of AMD's processors sold out on Amazon.com and other retail websites.

AMD rallies as cryptocurrency miners snap up graphics chips
| Reuters
 
Unfortunately, they don't benefit from this unless they pump up production to meet demand.
 
I think that it was also because of Apple (announcing massive use of AMD Radeon GPUs in the coming products)...
 
The Ethereum mining situation is very interesting from the view point
of computer architecture. The idea of bitcoin variants mining is to
encourage people to provide computing power to keep the public ledger
honest by preventing people from erasing transactions and spending coins
more than one time. It is a solution to the multi-party secret sharing
problem with the amazing observation that fairness can be provided
by making sure the majority of computing power is honest. Before
Bitcoin, the idea was to make sure the majority of parties involved in
the multi-party sharing were honest.

It turns out that bitcoins are am improved way to stop cheaters
(code breakers) but is not fool proof either. Here is the
original design document by the originator of the Ethereum coin with
a discussion of preventing ASIC use and fixing problems with Bitcoin.

On Mining - Ethereum Blog

The problem with bitcoin mining is that it is too easy to design
an ASIC that can perform the SHA256 hashes. Mining works by hashing
and then collecting a coin if the hash is small enough. Currently
less than 2*191 (75 leading 0 bits). There is no better algorithm.
The way Ethereum eliminated the advantage of cheap ASICs (the Vitalik
Buterin article above has interesting pictures of the 2014 Bitcoin
ASIC factory in China) is to require the entire Ethereum coin ledger
be stored (gigabytes of data) and making the mining operation more like
a normal computer program that requires a real computer.

Here is an article from 2013 describing why the AMD graphics
processors architecture made it more economical (faster gpus for
this application) to use AMD GPUs in graphics cards than Nvidia.

AMD destroys Nvidia at Bitcoin mining, can the gap ever be bridged? - ExtremeTech

Basically, the AMD cards were much closer to normal computers while
the Nvidia cards were more highly optimized for graphics display.
One main advantage of the AMD card was that it contained a rotate instruction
that took 3 instructions on the Nvidia card of the time. It turned out
that back then too, AMD graphics cards sold out, but it was bad business
for AMD because when they ramped up production, the Bitcoin mining
algorithms had moved on. I think AMD and NVDA will run into the same
problem this time, i.e. bad business as technology moves on.

Here is a recent article explaining why AMD cards are faster (more
cost effective) because their architecture is more like that of a
normal CPU.

Ethereum Mining GPU Hashrate Performance Roundup

I do not understand the attraction of Bitcoin type public ledgers because
they can be beaten with various attacks and when such a hack happens,
there will be no legal system to turn to (attacks are discussed in
Buterin's blog post above).
 
Most interesting is that Ethereum is even working on removing the need for heavy computation to secure their blockchain by switching to a prove-of-stake system from the prove-of-work system they are currently using. So the use of GPUs for ethereum mining will only be temporarily.
 
The Ethereum mining situation is very interesting from the view point
of computer architecture. The idea of bitcoin variants mining is to
encourage people to provide computing power to keep the public ledger
honest by preventing people from erasing transactions and spending coins
more than one time. It is a solution to the multi-party secret sharing
problem with the amazing observation that fairness can be provided
by making sure the majority of computing power is honest. Before
Bitcoin, the idea was to make sure the majority of parties involved in
the multi-party sharing were honest.
....
Excellent post - lots of info!
 
I took a quick look at proof of stake. It looks to me like it
takes a step backward to multi-party secret sharing that
has the problem that a group can collude at any time
(say Christmas Eve) to change or erase transactions. The idea
of hash code mining is that it is much harder for cheaters to
aggregate huge computing power. Mining computing power is a
large user of energy though.
 
The idea
of hash code mining is that it is much harder for cheaters to
aggregate huge computing power. Mining computing power is a
large user of energy though.

The idea is to replace huge computing power need which is expensive to the stake in the total amount of ethereum which is also expensive. Indeed a lot of research is happening on keeping the system fair. As in a proof of stake system the next block generator is determined at pseuso-random according to one's stake it can in theory be made more resistent to 51% attacks as you have in the proof of work system.
 
I do not think you are seeing public ledger algorithms
correctly. There is a long history of block chain algorithms
called multi-party secret sharing that were all seen as too easy
to hack using approaches where many of the participants colluded
to cheat. The Bitcoin break through was to see that by encouraging
people to spend computer resources on mining to make sure a majority
of the computing power is honest makes it much harder to hack public
ledgers (multiple spends of coins and alteration of transaction
history - ownership). It is very unlikely that some formal
algorithm is going to replace brute force coin mining (called
perebor).

I think I come from a different more skeptical world of
cryptography (I am not a cryptographer. I keep up with cryptography
because the combinatorics in Verilog simulators is similar to that
used in cryptography). For example algorithm security proofs do
not take into account the attack of obtaining (stealing) the
encryption IC net list.

There is a paper by Adi Shamir's (the S in RSA) student Dorit Ron
that I think shows that no algorithm can replace coin mining
(https://eprint.iacr.org/2012/584.pdf). Paper also has an argument
that controlling only 29% of the computing power is enough to
compromise Bitcoin security. Shamir is even more skeptical of block
chain security than I am. Here is a list of his 2016 predictions for
the future of computer security:

https://www.linkedin.com/pulse/adi-shamir-makes-15-predictions-next-years-andreas-sfakianakis
 
Back
Top