Let’s talk about ASIC mining

Do you guys know the story of 2 wolves ?
An old Cherokee is teaching his grandson about life. “A fight is going on inside me,” he said to the boy.

“It is a terrible fight and it is between two wolves. One is evil – he is anger, envy, sorrow, regret, greed, arrogance, self-pity, guilt, resentment, inferiority, lies, false pride, superiority, and ego.” He continued, “The other is good – he is joy, peace, love, hope, serenity, humility, kindness, benevolence, empathy, generosity, truth, compassion, and faith. The same fight is going on inside you – and inside every other person, too.”

The grandson thought about it for a minute and then asked his grandfather, “Which wolf will win?”

The old Cherokee simply replied, “The one you feed.”

Well Zcash has been feeding the evil wolf (ASICS) atm.

1 Like

It is more than likly dead for ZCash. I just dont see Zcash switching back to GPUs anytime soon. Even if they do it right now, its probly to late already. Why on earth would they drop almost ALL the hash power they have right now, in HOPES that GPU miners will come back and secure the network.

If forking was too much of a risk a few months ago, the risk has increased exponentially now that all the GPUs are gone.

As a GPU miner, the bridge has been burned for me. Unless Zcash becomes MUCH more profitable than other coins, I doubt I will ever mine it again. The team let down all the GPU miners.

How long before a new ASIC comes out on the new Zcash algo, and the devs are to busy with other upgrades again. ASIC resistance is evidently a much lower priority to them than it is to me, im better off with coins who backup what they say, aka ASIC RESISTANT

4 Likes

I still stand by what I said when this topic first came up, ASICs are not good for decentralization. Let’s talk about ASIC mining - #2 by Shawn

My hope is that as big releases like Sapling come out (and free up the focus on the protocol) that the Developers will have time to re-visit the subject. I would love Ethereum to adopt ProgPOW just to show that it works and to show the Zcash team it’s worth looking at.

3 Likes

me too hoping that sapling was the main argument for not taking any action.
asics not only difficult to get, but imho problem is that producer aka seller of this devices always trying to squeeze maximum from maximum markets. That include price\hashrate pumps, dumps, decreasing\increasing hardware price up to 3x depending on the situation in crypto… Im not saying that its not natural for business, im saying that gpu producers and dealers tend not to involve so much, which in my very humble opinion makes the whole thing more stable, fair and predictable compared to asics.

we have to discuss an elephant in the room.
how exactly to modify algo for GPU’s?
what are the possibility of bitmain learning a lesson, and developing asic for 144_5, for secret use?

As i understand it could be discussed if the Constantinople update is further delayed than February. My English understanding isn’t the best but right now it does not see to be on the daily discussion on ETH at all.

Prior to the decision of postponing Constantinople, Martin Holste Swende, the lead security of Ethereum, said:
I’m going to go out on a limb here and say that if we do decide that Constantinople isn’t until January or February, then I would probably try to push for including ProgPoW into Constantinople.

Even not sure what sense it would make to switch an algo only due some 2-6 months delay. But as well i have no idea how a change to ProgPoW would influence the current state of Constantinople at all. For me personally a switch to another algo is very unlikely having in mind they are aiming for POS anyway. (Hope this POS remark is ok and doesn’t get moved!).

I think 144_5 is weak, we should push for higher resistant algo like 192_7 or progPoW

I wouldn’t concentrate only on Bitmain. Some other producers are way better on “harder” algos than Bitmain. Baikal and Innosilicon and some smaller producers would develope better hardware on such algos than Bitmain, obviously.

In my opinion such algos like 144_5 will eventually force Asic producers to make better hardware with better upgradeable components and software. I anyway wonder why there isn’t such asic hardware on the market yet. I mean is there really a burden (other than cost) to make an asic design with for example Memory slots that allow an upgrade? What about hybrid asics/fpga’s like the Baikal X10 Giant for example or such like designs?

Again my opinion, the more coins are forking away, the better and more upgradeable asics will be released, at least that is my logic without being a software/hardware expert of course, just following common logic.

in my understanding the more flexible hardware is, the less advantage in efficiency it gets.

1 Like

I don’t think it’s that easy and simple. While this may apply to current asic designs (cheap designs) i doubt it would be the same with really good designs.

Just as a theory example as i’am indeed not an asic/fpga hardware expert:

  • An Asic design is made with easy software update function.
  • The Asic has free unused slots for memory upgrades.
  • Memory upgrades are produced as plug&play.
  • Eventually other critical components are upgradeable as well.

This would only rise a given cost factor, not? Which for companies with big pockets woulddn’t be a real problem. Instead of having a doorstop asic you invest another USD 200-500 for example to upgrade it and are in the race/game again?

As said, just a more or less logical thought of mine and i would like to hear a comment from someone that has good hardware knowledge to see if this would be possible at all.

Edit: What for example would happen if there was an easy way to make the Z9/Z9 mini upgradeable with 10x the memory it currently has? What would be the result? Would it be able with some software/firmware tweaks to mine on 144_5 or 192_7?

Maybe questions for @mistfpga

Expandable memory slots are easy (and cheap!) to do but off-chip RAM is horribly slow & there’s not much you can do to speed things up… not really a viable product.

The reason ASICs are so fast is because the memory is on the chip, if you moved it to expandable slots, it would most likely have to travel down a BUS lane to get to the processor, this greatly slows down the process.

Imagine a store where everything was stored at the checkout lane(small storage, but extreamly fast), You walk in, walk to the checkout lane and make your purchase, and your done. on chip memory

Compared to a store, where you have to walk all over the store collecting the items(huge storage, but extremely slow), sometimes you have to drive to another store location to find what you are looking for, after taking time to find everything you needed, you then bring them to the checkout lane and purchase them. off chip memory

3 Likes

Also when memory is off-chip, you really want a high-speed interconnect.

An interesting example of this is AMD’s Threadripper 2 (2990wx):

2 Likes

I understand and i’am fully aware that current asics are like that. But i was more talking about possible future next generation asics and if it would be possible. The current asic design is here of no interest at all, hence i wrote a better asic design.

I understand that one of the current factors they are fast is the onchip memory. But than again, what would be IF:

  • the onchip memory is 10x or 20x times bigger on the Z9(mini)? How much more would that cost to produce on Bitmains scale for example?
  • Would it be possible to use off-chip memory and really high speed interconnect? Eventually cheaper?
  • What are the current limits for high speed interconnect?
  • How much off chip memory would be needed in case there is a solution for a very effective high speed interconnect to make an asic again superior on an algo like the different new equihash algos?
  • How much onchip memory would be needed to make them again superior on the new equihash algos?

I ask these questions as i think it’s pretty important while thinking about possible new algos someone should make the calculations not only just on current asics but as well on improved or next generation asic designs, not?

Edit: May sound stupid but just trying to read through the memory issue, so some more questions:

IT would cost much more…Z9 mini on 144/5 would need 70 gb off DDR5 Ram…for 192/7 more then 100 gb off DDR5 …and would slow his performance against GPU drastically… so for ETH Bitmain E3 has 18 chips with 4 Gb each mining a 10 mh…And that is costly … all algorithms that ASIC are much better then GPUs are one with low memory requirements …like curent Zcash…even Monero has very very low memory requirements…just 2 mb per thread ( so it can fit in CPU cash) but they doing fork every 6 months to not have ASIC.

Zcoin switching to MTP on December 10th Home | Firo - Privacy-preserving cryptocurrency

I was talking about next generation/new asic designs, not current ones. I’am aware what the possibillities are of the current cheap asic designs. But that’s not of interest at all in my opinion.

I mean when we think and discuss a new algo it should be talking into account what the asics eventually could do with better designs.

Reading now the 2nd day about HBM2 i still think this would make new asics pretty good and way more flexible. (in case it could be used in an asic at all!).

high bandwidth, but not low latency for random access, which most mining is

If ASIC resistance is actually possible, I’m all for it. The problem is that we aren’t sure if ASIC resistance is actually possible. If we’re not sure then we shouldn’t act to give a false sense of security. That would allow secret, hidden ASICs to pose a greater threat. It’s better to have them out in the open, in front of us, instead of silently taking a greater percentage of rewards, or the potential of a hidden 51% attack.

2 Likes

i personally would love to read from a professional asic dev, about which algorithm would create conditions when even theoretically asic would be almost as GPU in terms of productivity.