Let’s talk about ASIC mining

Mining
5054

Why do people think more hashrate = more secure?

The hashrate has gone up with ASICs, and we are LESS secure now vs GPU mining.

EDIT: I find it funny how Equihash coins were getting 51% attacked a month or 2 before the ASICs were shipped out. Almost like Bitmain was using them to attack coins before they shipped them out(ASICs were pre-ordered and already sold at this point). If it was because of NICEHASH like everyone says, why did it stop? Why did it start right when ASICs were announced? Why is it not still happening?

5055

This is quite informative @zooko @boxalex The Equihash ASIC Takeover is Complete | by Brian Stafford | Good Audience ?

5056

I hate to burst your bubble, but no, they will not be for gamers, they will be used to compete in cluster computing, AI (cars, etc). These things will suck at at games worse than vega. Think of what they are making to be similar to volta without cuda cores (depending on who or what you believe, it will be a fpga at its heart). or that 38 cluster thing nvidiia showed off,that is the market segment they are after.

Don’t quote me on this but im pretty sure they are going to be closer to fpga’s. it would make a lot of sense and work with their AI software development stuff.

The gaming stuff is just smoke and mirrors unfortunately. I mean it will play games, just it will be really bad at it.

This is very much inline with Intels previous attempts at GPUs. These things will cost 1000’s but they will be amazing at everything but games. That being said they have employed some amazing gpu people, but I think they will be building a core that runs on an fgpa rather than a full core.

edit: the fpga rumour has quite a bit of weight due to intel owning altera.

5057

now its my turn to burst your bubble: the rumours part i was talking about was about previewing them at CES 2019, not that they were working on gaming gpus

Intel is working on gpus mainly for gaming, so whatever personal opinion you’ve got about them doesn’t change their official statement about willing to relize gpu for gamers and willing to compete against nvidia and amd.

2020 is around the corner, we will see with our own eyes (:

1 Like
5058

No bubbles for anyone!

I was referring to the same gpu.

more info

Take this however you want. but the intel IoA teams have been 90% reassigned to 2020 gpu - no, sorry no code names, nothing. this could all be lies. Intel are not making a specific consumer GPU especially not before 2022 but I might have more access to intel internal intel than most.

This GPU will be a compute based GPU with an FPGA core. Ignore the hype. This will be a beast, just not at games.

no, this is hype. and wrong.

What? my personal opinion doesn’t make the slightest bit of difference. I will edit this post in a few hours with the links.

Edit 1 (of a few): Project Larrabee, Intels first “gpu”, fpga core. - look at its goals, sounds like an rtx right? (like real time ray tracing in 2007) - notice how it emulates CUDA cores rather than having them. I actually have a vid around here somewhere of someone taking one apart. maybe buildziod. cant remember.

Larrabee was intended to differ from older discrete GPUs such as the [GeForce 200 Series](https://en.wikipedia.org/wiki/GeForce_200_Series) and the [Radeon 4000 series](https://en.wikipedia.org/wiki/Radeon_R700) in three major ways:
  • It was to use the x86 instruction set with Larrabee-specific extensions.[11]
  • It was to feature cache coherency across all its cores.[11]
  • It was to include very little specialized graphics hardware, instead performing tasks like z-buffering, clipping, and blending in software, using a tile-based rendering approach.[11]

This had been expected to make Larrabee more flexible than current GPUs, allowing more differentiation in appearance between games or other 3D applications. Intel’s SIGGRAPH 2008 paper mentioned several rendering features that were difficult to achieve on current GPUs: render target read, order-independent transparency, irregular shadow mapping, and real-time raytracing.[11]

More recent GPUs such as ATI’s Radeon HD 5xxx and Nvidia’s GeForce 400 Series feature increasingly broad general-purpose computing capabilities via DirectX11 DirectCompute and OpenCL, as well as Nvidia’s proprietary CUDA technology, giving them many of the capabilities of Larrabee.

They marketed this at gamers even though compute and flexibility was their main focus.

raw speculation

The timelines for the poaching of staff from AMD would fit in a lot more with using compute to emulate cuda (all speculation and rubbish speculation at that) oh they also got the dude to do the integrated graphics too iirc Intel already have the staff, they need the specialists. their GPU will be nearly complete so an FPGA based solution would make more sense.

/raw speculation.

2 Likes
5059

As a manufacturer level, it is real…
e.g. Z9
The cost maybe USD100 only…
before they launch to market, if they mine in factory… ROI in 1 day is not a joke :slight_smile:

5060

Hi,

Are these rough numbers or do you have a link?

I understand where you are coming from, I would like to fix a common misconception about the cost of ASICs. A lot of the cost of ASIC production is in the 90% of the development that does not involve chips. So the final product might cost 100 in parts (but probably a lot less) but they still have a few million to recover on the development process + the more machines they sell the less they are worth…

the last HSM I worked on (think trezor mixed with enigma machine, but for the cash machine network, this is what the cash machines talk to to get auth for transactions.) cost 25million and 5 years to develop. the retail for about 30k and have about 2k’s worth of components inside.

Currently and pre ProgPOW ideas, ASIC resistance is meant to have come from the sunk R&D cost to produce the actual device, and then the devs could flip a switch and invalidate their R&D. This is a whack a mole strategy and threat. Seems like monero flipped the switch. not ideal, but it buys time. Like @OhGodAGirl said software engineers are not the best at hardware resistant algo’s.

She also recommend a couple of really good hardware engineers to review the algos. I strongly suggest they work with the gpu benchmarking companies, these people have all the info you need.

5061

Hi. I was summoned.

What device are we specifically talking about? I’ll give you a cost breakdown, based on the silicon and parts. Then you can do the rest of the work in evaluating markup.

By the by, RnD is dirt cheap on these kinds of devices. Depending on your team, you can do a new design in 30 days. It’s all about the C > VHDL > Synth > PoC on FPGA.

Monero is incredibly vulnerable to FPGAs, regardless of what algorithm they choose.

Before a coin is designed, it should think about what kind of user base it wants on the mining side, and what kind of user base it wants on the adoption side - that will dictate what kind of hardware they should model on, and will then dictate the PoW algorithm they should use.

5 Likes
5062

I personally doubt the cost is as low as USD 100 per unit, but than again i make a different calculation when it comes to the self cost of a given product. My feeling is more that 1 unit is more in the range of USD 200-300 after all involved costs are splitted and added.

However, when someone makes a product there are more costs involved as just the parts in the product itself and maybe that’s why i estimate a higher product cost. Have in mind that these factors as well should be included into the self cost of given product:

  • R&D, if i remember right, at Bitmain work 1.800 engineers, these are costs to cover.

  • Workers for assembly, packing, whatever, yes, in whatever product such paid work is involved.

  • Electricity, no idea how much but it’s for sure again a factor.

  • machinery, tools, buildings, maintance, whatever

  • taxes, insurances, amortization, interest for debts/investors, and mostly 50 more expenses that come with a factory, lol.

Means, if we talk about the cost of a product i think someone should be good to add at least double or triple the mount of the parts involved to get an idea of the whole cost involved in a given product, just taking only the price for the parts isn’t really showing the cost for the product. Just a generally thought.

5063

I’m in the same boat as @johnwisdom and have many GPUs. It would benefit me immensely if we switched to ProgPoW.

The problem is the hastiness of changing things before we’re sure that it’s a safe. Changing the PoW algorithm is a big deal because it signals developer-driven consensus (that a few people decide who wins or loses). A change would reject the idea that miners are competing to find the most efficient way to mine a given coin. It pulls the rug out from under users that are currently securing the network. Forking doesn’t hurt Bitmain, it hurts Zcash users, a lot of them being early adopters.

I also think we should respect the community governance panel’s decision to not run away from ASICs just for the sake of being unpredictable. The KYC/AML Bitmain issue is alarming, but there are other ASIC manufacturers that don’t yet have this requirement and distributors also don’t have this requirement.

Finally, just to give a semi-relevant comparison: Siacoin recently made a decision to fork away from ASICs manufactured by others in order to create their own ASIC monopoly. It’s completely self-serving of them to do so and it removes any legitimacy of Siacoin being decentralized or community-driven. This is causing a split in their community via Sia Classic which could hurt that project over the long-term.

Zcash needs to avoid any change even remotely close to what Siacoin has done. Any PoW change should be voted on by the community governance panel so that it is a decentralized-driven choice.

Just my two zooks.

3 Likes
5064

This is False. The devs can change the code all they want. If the miners do not accept the new rules, they dont have to upgrade. This would cause a fork of the coin, happens all the time. The miners still have the final say. So a few devs can not decide who wins and loses unless the network agrees on it.

This is completely false also. Think back to how GPU miners have spent about 2 years investing into the Zcash network. Then within a span of 2-3 months they have all been kicked off.

How would forking have hurt Zcash early adopters? This would of allowed them to stay on the network. it WOULD of hurt Bitmain because all the units they were trying to sell would be useless. The only people that would of been hurt were people that bought into the ASIC bullshit.

5065

Doing nothing is also a decision and “a big deal… resulting in winners and loses”.

Doing nothing also goes against the whole point of launching Zcash with Equihash which is meant to be ASIC resistant and still is if the parameters were to be updated.

6 Likes
5066

No, it’s not false at all. @root is 100% correct, forking to ProgPoW would only serve to hurt Zcash early adopters at this point in time. Many early adopters I spoke to at Zcon0 had made the decision to buy Z9s. Any GPU miner who had any interest in continuing to mine Zcash have also necessarily invested into Equihash ASICs by now. Advocating for a PoW algorithm switch is simply adovcating for your own selfish interest.

I don’t think anyone can claim to have mined ZEC for as long as I have, and nobody kicked me off the network. Rather, I made a choice to invest in hardware that would allow me to continue mining ZEC. Like any GPU miner, my GPUs go where the maximum profitability is, which is why they’re not ideal for securing the network. Decentralization with GPUs is a myth, as hobby miners can’t possibly compete with the massive GPU farms that exist now. These Equihash ASICs actually gave me the opportunity to get as much hashpower as I could with 1,500 RX 480’s.

4 Likes
5067

This decision to do nothing was made earlier this year, and the result of that decision is that a large percentage of Zcash early adopters invested money into these Equihash ASICs in good faith. Making a PoW algorithm change at this point would be detrimentally harmful to the Zcash community. Anyone who made the decision to invest in hardware solely because they wanted to continue to mine ZEC would suffer a huge loss.

Maybe I’m being a selfish hypocrite, and I definitely have a conflict of interest in arguing for ASICs, but so do people advocating for a pow change. They’re upset because their GPUs aren’t earning as much as they used to. But the big difference here is that early Zcash GPU miners have already paid off their hardware by now and can go mine something else, whereas ASIC buyers are nowhere near the point of ROI yet. Equihash ASIC buyers would have no recourse to recoup their investment, whereas GPU miners have many options.

1 Like
5068

This cant be farther from the truth. The reason most of us are upset, is one of the top ASIC RESISTANT coins flopped to ASICs months before they were even shipped out.

Im upset we all agreed to a horse race, and someone shows up at the horse track with a car, and the officals are like, ITS OK! Now we all move to cars or we have to quit. If we wanted to race cars, we would goto the 1000 other tracks that do it. We as a community agreed on a set of rules, and now the trust in them rules were destroyed.

You invest in something you know the community was against, and then act like you are the victims when there is talk about switching back to what the community agreed on and was the norm for 2 years. Yet the GPU miners are the ones that just want to make a buck and are greedy…

1 Like
5069

Please tell me how im wrong? He is talking about developer-driven consensus. Devs can change the code all they want, it does not mean the miners have to update to the new code.

You can watch the ASIC hashrate jump from pool to pool, how is this any different?

Once again, how is this any different for ASICs? People with money will always have more…Your telling me there is no chance people will make ASIC farms?

5070

By the time ZcashCo actually would fork to ProgPoW the current batch of ASIC’s will not be making any money if history tells us anything. I would be fine with a fork in April or later and I have both ASIC’s and GPU’s. We thought we would make money with all this extra hashpower, but I still haven’t broken even on my batch 1 miners.

Just because there are large GPU farms doesn’t mean there is less decentralization. Every person with a GPU can mine and that in itself means greater decentralization. How decentralized will Zcash be when ASIC profitability is in the red for miners with .10 electricity? Only large ASIC farms will be profitable at that point. How many end users will buy ASIC’s then? But there will still be gamers and other users with GPU’s that could mine. Some would do it in their spare time even if they didn’t make money since they didn’t invest a lot into the project. Zooko says he wants people to be invested in the coin, but the more people invest the more they have to make a profit. That is where the problems start. Zcash says it wants to empower everyone yet allowing large ASIC farms benefits the few.

3 Likes
5071

I was referring to your second point in which you were disagreeing with the following:

Forking hurts Zcash early adopters more than it hurts ASIC manufacturers. You can be sure Bitmain and Innosilicon already made their money back and then some.

The difference is GPUs can go mine CryptoNight or Ethash instead of Equihash, and they’ll always jump to the most profitable coin. Equihash ASICs are stuck mining Equihash, and even though they can switch between ZEC and ZEN, Zcash will always be the dominant coin and will attract the majority of the hashrate. The result is a more stable network hashrate over the long-term.

I guarantee some people have made Equihash ASIC farms, but at this point I would be surprised if any serious business would consider doing so. With coin prices falling, it doesn’t make much sense for anyone to invest in these machines unless they expect ZEC prices to rise. And in that case, it would make more sense to simply buy ZEC outright. I had plans for an ASIC farm back in July and decided against it because revenue projections didn’t justify the capital investment.

5072

The only way directly investing in the coin is POS, at least i’am not aware of any other way than directly buying a given coin that would allow naming it “investing in the coin”.

5073

You have already said ZCash early adopters have made their money back and then some from GPU mines. Now your saying if Zcash fork away from ASICs it would hurt early adopters. 90% of the community were GPU miners only and refused to buy ASICs or had no access to them. The other 10% that wanted to and were ABLE to get the limited supply are the ones that are still here talking about how helpful ASICs are to everyone.

Way to avoid the point. ASICs can bounce around on a algo just as much as a GPU can. Just because Equihash only has a few coins that are profitable right now does not change the fact it does not help secure the network. Right now if a new Equihash coin came out and was way more profitable, what do you think would happen to all the ASIC hashrate? It would jump to the new coin, there is no investment into a coin.

5 Likes