Deep Explanation of how Ethereum ASICs are possible?

[michnern]

I am hoping someone can provide an in depth explanation of how an ASIC miner for Ethereum is even possible (in the traditional sense of what an ASIC is defined to be). When I think ASIC, I think of a custom chip that is designed to do ONE algorithm very quickly and efficiently. For example, SHA256 used for Bitcoin. After skimming the Ethereum wiki, I came across this line:

""Random circuit" - a proof of work function developed largely by Vlad Zamfir that involves generating a new program every 1000 nonces - essentially, choosing a new hash function each time, faster than even FPGAs can reconfigure. The approach was temporarily put aside because it was difficult to see what mechanism one can use to generate random programs that would be general enough so that specialization gains would be low; however, we see no fundamental reasons why the concept cannot be made to work"

This excerpt came from: https://github.com/ethereum/wiki/wiki/Dagger-Hashimoto

I think I answered my own question, but I will post my process here so that others can hopefully benefit. The above link was an predecessor to the current Ethash. Unfortunately, the current Ethash wiki makes no mention of changing the hashing algorithm every 1000 nonces. I wonder if they can figure out a way to add this in based on the current ASIC situation? It is unfortunate that this was ever set aside in the first place.

This Ethereum whitepaper talks about "Poisoning the well": https://github.com/ethereum/wiki/wiki/White-Paper

Which seems like a viable option they thought of in the beginning. Lots of people have been talking about this option for the last couple of months. With the strong possibility of an Ethereum ASIC, why have the developers not become vocal about "poisoning the well".

It is obvious from their papers that they acknowledged that ASICs could exist but their intent was to ensure the advantage of an ASIC was so marginally small that it would not significantly impact computer based mining.

What are  your thoughts?



Thanks,

[2stout]

Nothing is technically ASIC "proof" as of yet, but rather ASIC resistant.  Where there is money and opportunity, it's possible you may see one less ASIC resistant coin or a coin requiring a fork to re-establish ASIC resistance.

[philipma1957]

Eth can simply fork every 4 to 6 months and asics would not keep up.

The developer of eth is the issue not asics.

Any coin can have the will to beat asics with a fork every 4 to 6 months.

They can simply add an al-gore-rhythm   to the original coin.

right now raven x16r is hot

so bitmain says let us crush it.

Raven has x20r ready to fork today  but does it in 4-6 months

all of bitmain research is wasted  building the x16r chip

Raven then goes to x24r  in another 4-6 months  and on and on.

basically all al-gore-rhythms could do this and every asic could be turned to shit over night. won't happen with BTC or LTC  but the other coins could do it.

Asic builder fucked themselves doing this or not.  We will find out very soon.  next 4-8 months.  should be fun to watch it unfold.

[ikicha]

With this Difficulty, this is very possible ASIC eth and others Ethash algo coin,
For example, last day ago, Bitmain and others asic company announce Cryptonight ASIC.
I'm very sure, Cryptonight ASIC is using couple month ago before launch

[retard75]

Asic builder fucked themselves doing this or not.  We will find out very soon.  next 4-8 months.  should be fun to watch it unfold.

Asic builders just can't fuck themselves b/c there will always be a demand for shapely and beautiful doorstops.
Show the neighbours and your guests what sophisticated lifestyle is about! It's a status symbol.
"Look at my $13,000 doorstop you poor bastard. I made it! Despair and Begone!"

[pickleburglar]

I am hoping someone can provide an in depth explanation of how an ASIC miner for Ethereum is even possible

The computational part is no different from any other ASIC, just set down the exact transistors you need to do the same algorithm over and over.

What makes ETH asic resistant is its dependency on random memory access. There is no way to speed up this part, an ETH ASIC would have to use the same memory chips as computers.

Turns out this wasn't as big of an obstacle as people thought though, GDDR5 chips are expensive but there are mountains of DDR3 chips laying around for a cheap price. Since the algorithm scales perfectly in parallel, there is no reason why these slow DDR3 chips can't be used en masse to provide the required memory for ASIC's.

[maxell2hd]

Turns out this wasn't as big of an obstacle as people thought though, GDDR5 chips are expensive but there are mountains of DDR3 chips laying around for a cheap price. Since the algorithm scales perfectly in parallel, there is no reason why these slow DDR3 chips can't be used en masse to provide the required memory for ASIC's.

But putting chips in parallel is complicated and expensive to manufacture. It's a major reason why video card manufacturers have almost always gone fast but narrow vs wide but slow when given the choice. Even the most integrated wide and slow method, HBM, remains expensive and rare.

[honestinvestor]

as far as I know, the asiki do not work with ether yet, it can only be done on maps, but it's very, very dangerous to go into mining, it's worth thinking about whether you need it, it's easier to go to the factory, the nerves will be calm

[adaseb]

From what I understand is, ETH ASICs were always possible but the problem was that the GDDR5 was expensive and it would actually be cheaper or the same price as buying a GPU.

However DDR3 system ram is getting phased out due to DDR4 becoming more mainstream and there are millions of chips which need to be sold.

So Bitmain made deals with the chip manufacturers and bought these DDR3 ram chips very cheap and was actually able to design an ASIC which is cost efficient.

Until July when the E3 arrives in people's hands we will actually see if it's an actual ASIC, FGPA or just come mobile GPUs rammed in together like the Pandaminer.