Samsung Announces 3nm prototype chip! What does this mean for ASICs?

[Searing]

LOL. Not sure what this means with 5nm coming out for Bitmain and Innsilicon for Q1. So we will have 3nm in Q4? WTF!

https://news.bitcoin.com/chipmaking-giant-samsung-reveals-3nm-semiconductor-prototype/
 
For that matter will even 3nm matter much after the halving of blocks to 6.25 BTC per block after say around May 7th, 2020? (currently)

https://www.bitcoinclock.com/

Will the madness continue? Take poll?

Brad

[1714715834]

1714715834

[1714715834]

1714715834

[1714715834]

1714715834

[1714715834]

1714715834

[Searing]

reserved

[Philipma1957cellphone]

Well first it means don't buy any gear for the time being.
We have three factors here 1/2 ing

5nm due this quarter
3nm due 4th quarter.

I won't order anything for now. My guess is 5nm comes in july and 3nm comes in Feb.

[wndsnb]

I'd be pretty surprised if they are able to go from prototype to production by the end of the year, and even if they did, bigger players than bitcoin ASIC producers will probably suck up all the production capacity.

If the news slows down investment in new mining operations between now and the halving it would be a big win for current miners.

It'll be interesting to see who jumps into the 3nm first. Another article says Samsung offered the design software for the 3nm process to some of its customers last May, so it's possible some BTC ASIC makers have been working on it for a while. I'd love to see someone (Canaan?) offer a miner that beats Bitmain's efficiency #s. It could happen if Samsung can best TSMC since I think Bitmain is pretty tied into TSMC, but Canaan has used both TSMC and Samsung in the past.

[revolution82]

I saw an article, that I can't find at the moment.

Basically it was saying that the 3 nm would also change the physical properties of how the gateways worked instead of being on one surface it would be designed like a square tube and all four sides would act as pathways.

I am slaughtering this description, can anyone explain these design upgrades to help the rest understand or if I am out in my own world?

[philipma1957]

notfuzzywarm may chime in with the answer to this question.

@ wndsnb

we have about 140 kwatts burins as I type

we could add 30-40 kwatts of gear.

we may expand to that and then add 200 kwatts

and do a new location with 200 more kwatts

These announcements  simply make me have to wait to decide.

Out power cost is very good.  The only way to really fuck things up is to tap out on gear that becomes out of date.

If 5nm comes out before may 1  it will drive cost down on all current gear.
If ½ ing does not push price high enough it will drive down the cost of all current gear.

if 3nm comes out in december gear prices drop again.

if a s17 pro is doing 42 th on low and burning 35 watts it is 1400 watts.

5nm is 15% - 20% better   the s17 pro will drop to under 1000

½ will drop the s17 pro more maybe to 800

so I wait for now.

[kano]

The only madness I see is FUD

Firstly, the halving: well we've done it before ... twice ... the 3rd halving is just ... the next one in a long line of them.

As for "OMG 3nm coming - everyone run for your lives"
Yeah, if you can't make any profit mining for the next year, then what are you doing here in the first place?
You'd actually have to know how dramatic the performance change is for these lower nm chips.
The decrease in J/H performance each step will be getting smaller and smaller with each generation, until some other NON-nm related change happens.

[Searing]

I agree I have no plans to mine. I'm just watching from the bleachers! But I doubt 10c kWh USA mining will come back in data hall form in the USA anytime in the future, so just buying as i go BTC now. But, still got to watch in envy those that can still mine any crypto, even if eventually it is only overseas or some such.

Brad

[wndsnb]

[...]

It's Gate-all-around FET (GAAFET), not a new idea but the industry just went in a different direction. Just allows more surface area for the gate, which improves various properties of the device. There's a description of it on Wikipedia, https://en.wikipedia.org/wiki/Multigate_device#Gate-all-around_FET_(GAAFET).

[mikeywith]

I wouldn't be too concerned about any upcoming improvements in the mining industry as far as efficiency is concerned. The Antminer S9 was built on 16nm and was released in 2016, four years later what do we have?

The S17 pro runs on low-power does 36w/th, the S9 now does 75w/th and that is 16nm to 7nm. the cost per terahash for the S17 pro is close to 50$, way too expensive for the amount of efficiency it gives.

5nm should be 15-20% improvement over 7nm, so what will 3nm bring us? 10% improvement over the 5nm specs? at what cost and when?

The way I see it, if you now own gears that do 40-45w/th you should be safe for years to come, the new 5nm gears will cost a lot to make and a fortune to buy, with rewards being cut in half, bitcoin price not going to the moon and adoption isn't skyrocketing, I see no economic incentives for mining gears manufacturers to actually try and rush into adopting 5nm based chips, let alone 3nm. It's not about just making the most efficient mining equipment, things are not that simple, you need to make gears that you can actually sell, gears that will ROI in a short period of time because making mining gear is like holding a hand grenade if you don't get rid of them fast enough, you are **** up.

[philipma1957]

The Antminer S9 was built on 16nm and was released in 2016, four years later what do we have?

I used to think that the bold type was true  and the s17 pro would be good for at least two -three years.

I am not sure 🤔 what will happen with the s19 and 5nm or the s21 and 3nm.

3nm and 5nm are announced and maybe they will push the new gear ⚙️ out.

I can wait for now.  I will only add a few units of gear and not push hard from now til the 1/2 ing.

[wndsnb]

The industry has gone much more "industrial" since the S9 was released. This means that the big mining farms are competing primarily against each other, and not against "home miners". So the margins now are much less than they were and the time for that margin to go to 0 as more efficient gear becomes available will be much less.

The good news is that the pace of efficiency gains will probably slow down, and the upfront cost for manufacturers to build on new nodes is increasing a lot for each new node.

[philipma1957]

So the margins now are much less than they were and the time for that margin to go to 0 as more efficient gear becomes available will be much less.

This claim is interesting 🧐.

If true mining will be more and more 2-5 cent power or you are dead.

Even if 5nm gets us to 20 watts a th and then 3nm gets us to 12 watts a th even if the best chip design of 3nm drops it to 8 watts a th there is a dead end in site.

the 1/2 ing is 6.25 in 2020. a 37 watt th gear ⚙️ is merely okay at 10k a coin
and then is 3.125 in 2024.   so a 19 watt th gear is merely okay at 10k a coin
lastly it is.  1.5625 in 2028.  lastly 9 watt th gear is just making it at 10k a coin.

we are going to see shifts in prices of Btc,bsv,bch should be very interesting times ahead.

[mikeywith]

the 1/2 ing is 6.25 in 2020. a 37 watt th gear ⚙️ is merely okay at 10k a coin
and then is 3.125 in 2024.   so a 19 watt th gear is merely okay at 10k a coin
lastly it is.  1.5625 in 2028.  lastly 9 watt th gear is just making it at 10k a coin.

There is another important factor that should be considered, people with free or close to free power sources, those will always "ruin" the game and make the most efficient gears way less profitable than they "should" be, Thus the economic incentives for making such gears will decline.

Imagine a network where having 100EH of 10w/TH and bitcoin price at 100k, the mining profitability is high enough to justify paying say 50$ per terahash. Take the same exact network but add 20EH at 50w/TH owned by those who have free or close enough to free power access, it now makes no sense to pay 50$ a TH for 10w/TH because of the "glitch" or the "unfair" distribution of electricity.

I am not sure I have made my self clear enough but I hope you get the point, let me know if more explanation is needed.

[philipma1957]

yeah when you have dirt cheap power say 1-2 cent.  no such thing as 0 cost .

0 cost power once you add in labor and maintenace adds 1 to 2 cent cost.

i have flat rate power  always 1/2 the coins mined goes to host. he maintains building and supplies power.

I supply gear fix gear monitor gear. the main building is a long ride makes it a bit hard to do everything.

just driving to and from the site costs 10 usd in gas ⛽️ 5 in tolls. plus 2.5 to 3 hours time. a good month is one trip a bad month is three trips.  we average two trips a month so two guys x 2 days x 12 months is

48 days of work maybe only six hours a day still comes to 300 hours labor and 360 usd in gas and tolls.

add in parts for repair and i check gear six times a day every day which is an hour work so maybe add 300 more hours labor.   this comes to 600 hours a year spent on mining. here in usa min wage is around 12 and this work would be more skilled so say 20 and hour x 600 = 12000 usd in labor.

so my earnings are not very high doing all this. but i do like it.  i am just not a hobbyist anymore.

[mikeywith]

0 cost power once you add in labor and maintenace adds 1 to 2 cent cost.

That is true but the cost of maintenance and labor applies to everyone regardless of how much they pay for power.

0 cents power cost + labor = 2 cents
6 cents power cost + labor = 8 cents

so it's a matter of ratio rather than exact numbers, in fact now that you mentioned labor and maintenance it's actually even worse for the guys who pay more for electricity, it is safer to assume that the higher the power rate, the higher the wages are.

in other words, if you were to hire someone to do your job in the U.S, you need to pay him 20$*600Hrs = 12,000$ a year, the guy with cheaper power in China can hire someone to do the exact job you do for 3$ an hour and end up paying 1,800$ instead of 12,000$, and they won't pay as much for gas and tolls.

think about this as a "glitch" in the game where a group of people uses exploits that can't be fixed and the rest have to compete against them that way.

[philipma1957]

Yes  which back to 3nm  efficient gear vs 28nm not efficient gear.

3nm for me is going to be slowly added in  while I drop out my s9 28nm gear.

This is due to my  power deal structure and available power.  ie 50-50 coin split  and 200kwatts power right now

with a possible 200 + 200 to be added over time.

I juggle a lot of gear

s7ln
s9 s9i s9k
s15
s17
s17pro
t15
t17
t17e
t17+
avalon a721
avalon 1047
avalon 1066
inno  t2t24
inno  t3t39
inno  t3t50

and L3+

From the hosters view point:

he would rather it all be s17pro his cost to buy is zero and he gets 50-50 on more efficient gear

from my view point I would rather it all be s9 as my cost to buy is cheap  and cost to run is 50-50 split

I have to think long and hard  about adding gear.

I really only want s17 pros at the moment so I am waiting to expand. By waiting now I have heard about both  5nm and 3nm gear.  which certainly will make me wait more.

[mikeywith]

I really only want s17 pros at the moment so I am waiting to expand.

Exactly my thought, while the S17+ has the same efficiency as the S17 pro running in normal mode, the S17 pro can do 36 watts per terhash running on low-power mode which in the future might make the difference between winning and losing when and if 5nm and 3nm hit the market, they go for about 1700$ in the Chinese market now, that is 34$ per terhash, the S17+ 70th goes for about 2100$ and that is 30$ per terahash, the quality of the chips and the low power mode is what makes these gears the most expensive at the moment.