Topic: Rack Style Mining (AntRack) (Read 726 times)

Wow! If I had only known this on my 1st BTC on my 1st day on October 18th, 2013! Price was $150.00 USD that day, at least at the time I looked. I was somewhat sad that day being 'delayed' by 6 or more months of getting an ASIC miner due to a BFL scam/etc. I tried GPU mining but had no skills. I think I made .6 BTC with GPU card..er...maybe.....I think I spent it...I was a clueless newbie and every week BFL said the miners were SOON! (tm BFL) So held off on GPU mining.*duh*

When I started on the forum here or 3 days after I started on the forum here the thread Wall Observer which was started on April 16th, 2013. I love looking at that first post on WO to this day that stated that Bitcoin was $67.43 USD https://bitcointalksearch.org/topic/wall-observer-btcusd-bitcoin-price-movement-tracking-discussion-178336.

At the heady price of $150 on the 1st day I mined above, so sure I'd missed the 'boat' and Bitcoin was gonna crash back down to $50.00 USD soon! Annoyed at my 'late start' to mining after lurking on Bitcointalk from April 13th, 2013....so much wasted time..so far behind! Also annoyed and scared self having to use a paper clip on the PSU to boot the KNC Jupiter BTC Miner, because KNC was too frigging cheap to add a frigging SWITCH on the unit. Yep was one generally freaked-out newbie that day wondering about this whole kool-aid drinking BTC idea! I guess my doubts have been cleared up since then, er maybe? Huh? May the next 10 years see the same price rise and adoption speed or better than when I flicked the KNC on.

But damn, really, I mean really, $4,350,000.00 USD without the containers and setup and tariff and shipping and import fees and 2nd pump and plumber and electrician and permits, etc, etc. We'd then be talking at least a cool $500,000.00 USD for 1 Bitcoin a day now! Holy crap!

Philipma1957! (tm "He does math well!")

edit: Philipma1957 you should start a new thread on this. Using the popular Bitcoin ASIC miner of the particular historical year from say CPU's on up till now and say above future or if equipment is announced like the AntRack within the year in question. It would be one shocking graph/chart/list or whatever indeed...and only will look more and more impossible as time goes on!

Brad

viabtc says 1 th gives 0.00000697 btc.

so :
14.5th gives 0.000101065 btc a day
145th gives  0.001010650 "
1.45ph gives 0.010106500 "
14.5ph gives 0.101065000 "
145ph. gives 1.010650000 "

so a set of 10 fully loaded containers does the trick

and you burn 3 megawatts an hour if the gear does 30 watts a th.

at 30 dollars a th the setup above = $$$$

30 x 145 x 1000 = 4,350,000.00 USD.

Just as an aside. When I started Bitcoin mining on Oct 18th, 2013. I made 'briefly' 1 Bitcoin a Day. Are my calculations correct, without electricity, that I would have to have like 10 of these at 15ph each to make ONE Bitcoin now? or approximately what is the cost of these again? I think you said someplace on here of $40,000.00  (will re-look)

So to make ONE Bitcoin per day, without figuring electricity and just the equipment, and at retail no tariff or import or shipping fee...I'd be looking at around $400,000.00 of equipment or likely double that for One Bitcoin of profit per day now vs 2013! Damn....anyway ball parky around 1/2 million dollars damn. Damn, the game is rigged since 2013 for this game! Ack!

Feel free to correct me and use more precise math, but you'd think I could find a link for such a thing or real costs in mining equipment to make ONE Bitcoin a day. Anyway, a bit off-topic...but Antrack is the latest and greatest coming out from Bitmain for my calculation purposes here and/or with the above $$$ nailed down with the idea of 1 Bitcoin per day profit...if you are a 'mega-whale' I just gave you a Bitcoin kool-aid inspired reason to blow a lot of money on these beasts!

Hey it worked for me chucking out what I thought was 'big money' for ONE BTC or so per day back in 2013.... let's see if the same logic applies...actually that would only probably work at $100,000 BTC in the next 5-10 years maybe. (All this math, all these $$$, all that math hash...my head hurts...Ack!)

Brad

The 19 series uses a single large heatsink that covers all chips, Microbt M20s and M30s have two heatsinks each covering about half of the chips, also, IIRC the heatsinks are not glued/soldered to the chips, they are attached using some screws, of course, it's more than likely that they put the thermal paste in between.

This one large heatsink instead of one heatsink per chip design looks like it's going to be the future for all miners to come.

I'd really like to get an up-close look at the guts of one of these. Looks like they have one heat exchanger per hashboard instead of individually mounted heatsinks on each asic. That should make reliability an order of magnitude higher than the air-cooled ones. I think repairing them gets way easier too. >50% of replacing an asic on a 17 series miner is messing with those damn heatsinks.

Different scenarios require different set-ups, at times more of less-efficient hashrate makes you more profit than more-efficient but less total hashrate, this is one good thing about custom firmware, using Vnish and a management tool like awoesomeminer you can automate this process, you enter certain variables such as price, difficulty, your power rate and etc, and then the software will handle the efficiency profile based on what's the most profitable at any given time.

If bitmain wants to stop the usage of custom firmware, they should implement this in their new Antrack, at least 3 to 5 power modes should be there, I don't think it's all that difficult for them to implement.

Not that it is needed with prices of coins high.

I believe the S19 pro will do a lot better than these numbers, the stock efficiency for S17 Pro is 39.5 W/T,  the stock values for the S19 pro is 29.5W/T, if Vnish can turn 39.5w to 26w then it should do a lot better in tuning 29.5w to below 20 range, anyway it won't be too long till we find out, I believe in a one to two months Vnish for S19pro should be ready.

If s17pro can do 24-26 watts a th and do 38-39th which it can on vanish

then the s19p may do 22-24 watts a th and do say 80-82th on the next vanish firmware

these racks claim 30-33 watts a th 10 per antbox 15ph

they may down clock to 12 ph with better firmware an pull the said same  22-24 was a th.

That's true phil, but it isn't always a question of efficiency, I don't know how many TH you are getting out of those S17pro when you underclock the life of it, but I am sure the cost per TH at that efficiency is very high, which also means these new S19 boards will do a lot better than the 17 pro once underclocked, but overall this isn't a fair comparison IMO because it isn't just 17 pro vs s19/19 pro, the rack style setup and water cooling is what makes this "special".

The real question now is: How much is it going to cost and whether it will be justifiable to buy the rack over the normal air-cooled S19?

Yeah since it is running 15ph and earning 57000 a month pre power cost built in redundancy for the pump is important.

I think we could run 2 x 9ph = 18 ph  and then run out of power.  Which would mean 2 boxes of 6 racks each.

I wonder if you could run your own pump to 1 rack of 1.5 ph

I still see an issue that issue is 33 watts to a th.

I have the one s17pro and on vanish aftermarket firmware I can do 25 watts a th.
So it is less efficient than aftermarket s17pros and stock s19pros.

Nope, not just a NA thing. I've designed equipment for use in Taiwan, Korea, Germany, Brazil, Mexico & France along with NA and looking at chillers/pumps literature from companies based in said countries, never once have I seen fluid flow spec'ed as weight, only volume over time.

The technical reason is simple: fluids are non-compressible so any specific volume of a given fluid will always be the same mass regardless of the system pressure. (edit: for complete accuracy, yes most fluids do expand/contract slightly depending on temperature but rarely enough to care about when it comes to flow measurements.) Gasses - yeah flow can either be based on volume (in which case because mass rapidly changes depending on pressure, flow will always be stated at a specified pressure) or mass flow rate which takes pressure out of the equation.

Back to main point, 33GPM isn't too bad, actually on par with what a 5kw laser sucks. As for pump reliability, as with many things that mainly depends on how good it is. A decent properly spec'd industrial rated pump should run 24x7 for decades with zero issues.

Personally, just to address the slight chance of pump failure, I'd design the system with redundant twin pumps each throttled down to 50% speed with VFD's (Variable Frequency Drives). 2 reasons: 1st is of course redundancy but 2nd is that power consumption is much lower- for any given amount of horsepower driving fans and pumps each 10% reduction in RPM drops power consumed by almost 20% (within limits of course). If you lose a pump a PLC or smart pressure sensor would speed up the remaining pump to take over until a replacement can be obtained & installed. For the same reasons I would do the same for fans on the radiators as well.

I wouldn't say so, in Serbia we use m³/h for pumps/fans/anything else moving air or water.

I am not sure what's the point of using kg/h but I'd assume it has to be some stupid small amount to be worth using that metric as 1 kg/h equals 0.001 m³/h but then again, why wouldn't you use liters per hour instead of kg/h. (1 kg/h = 1.35 l/h IIRC)

I've heard older people using kg/h while gas companies and everything I noted above is using m³/h nowadays, so maybe kg/h is an old metric that they used before our (read: "my") time.

So down sides are :

1) will need a back up pump
2) seems to be 30-33 watts a th a step back from the s19pros

15ph is earning 15000 x .1267 =1900 usd a day or  57000usd a month.

Perhaps that's just an American thing?, where I live, we use kg/time, water is mass and what you want here is the mass flow rate, so while ton/hour might seem strange to you, it sounds pretty much normal to me  , after all, 1kg of water mass per hour equals 1L per hour, anyway, that is a ton (pun intended) of water to be moved.



There is no back-up pump in the design, a huge SPOF in this, if one decides to go for these, a back-up pump must be installed, someone I know might be getting one of these in Feb, if that happens I will most likely be providing you guys with exclusive videos and images and more details.

9 x 224 = 2000 gallons an hour.

So do we have a pump for 10 racks which recycles at 33 gallons a minute

If the pump dies do all 10 racks go off line?

Or is there a backup pump?

WTF kind of flow spec is 6.9-9 tons per hour??
You talk to any facility services engineer/manager or any industrial cooling company and they will think you are a loon if you use weight to specify flow rate...

Flow is always spec'ed as gallons or liters per sec or min or hr. The only variant I've ever seen was in cu-ft or cu-meters which are still volumetric units. NEVER in my 40+ years in the systems design biz have I seen flow spec'd as weight.

FYI, if you don't mind doing math, 1 ton of (distilled) water = 224 imperial gallons (1.018 m3)

Sorry, just as an aside...I can't flip BTC for equipment anymore at that level of play..I'm out....big boy games (or guys with solar farms whose name I should not mention) can do such...but I mean EVEN if we had BTC/Crypto prices at the same level of risk/reward of say 2017...I just could not pull the trigger on such equipment..at current price of btc/crypto and the equipment investments...even if I had insider cheap kWh in the USA. (by the by what is the best rate in USA 4c kWh?)

There comes a time, when you not only find yourself shocked by the price of BTC/Crypto from when you were a newbie and started...to now..when I'm shocked at the investment of BTC/Crypto into mining/equipmetn units...even if I had a 'legit' return on investment...too old in crypto..I am afraid I can't pull the trigger on this kinda thing anymore....like my newbie days of like the ancient 4 years ago..alas, I miss my 'newbie' BTC/Crypto big brass ones...seems at $18.5K I've become a conservative HODL'er ....ie or no more 'big brass ones' like back in the day. I hope the newbies of today can see this post like 4 years from now and at 100k BTC and realize the same.

Sorry had to post this, in that Philipma1957 at least can 'casually' toss about infrastructure costs for equipment like this..me..on top of $18.5K mining equipment prices...It just has got to the point on both..that from back in the day...it boggles my mind...and can't get my head around such.

I'm out..just reading the above on other's actions and/or contemplations above on getting such.....no matter how good the hardware/return on with using these and mining with a cheap kWh rate...freaks me out and gives me the shudders. Bitcoin mining has now moved from my little Eskimo rowboat to commercial whale mining..so I'm out...Though it is good to reminded on every ATH or close to such or every pump/FOMO and equipment innovation or so...I get this slammed into my head with the above reality. Your OLD, crypto wise, no longer BOLD, crypto wise, so you can just only HODL now, crypto wise!

Time to hang up the mining spurs I'll be on the 'sidelines' in the bleachers waving at the playing field from now on.

Brad

I got you some updates, thanks to the Coindad.

Specs:



More Specs:



The layout:



Hashboard side view:



Swapping in action:

https://www.youtube.com/watch?v=EhuXO8Y2paM

More action:

https://www.youtube.com/watch?v=mzH3KMPak1w

---

They look dope, but I'l pass.  , what does everyone think about these Antracks now? of course, prices are still unknown but I will update the topic accordingly.

Well a lot depends on the rack price.

I am talking about gpus since it is an easy way to compare.

A good well cooled 8 gpu bare bones rack with good psu's and cooling  is 500-600 bucks.

you can fit 8 x 800 dollar gpu's in it

so 6400 in hashing to 500-600 in rig

a 10 or 12 to one ratio.

If a rack is 4-5k

and to fill with cards is 50k

I could buy a rack and fill it ¼ for 12.5k

then add to it as time goes.

I think we can do 2 racks  and spend 100k to do so.

Maybe it will come unloaded and they will start selling hash boards sepertly? makes more sense no?

I would also be interested to try it, just for the sake of it, my power rate and business model don't justify buying such efficient equipment but if the price isn't crazy high, why not? also, one thing to notice is that they will probably be shipped by sea as these are going to be heavy and large in size, so air-freight will unlikely be an option so expect a delay of 1-2 months.

I would buy one if

no trump tax

price is right.

Pretend it has all s19 pro boards

and that it can do 1.5 ph

that would be at least 40k

not sure I want to spend 40k in gear in one shot.

I am not against your theory, I find it perfectly thought of, the issue is the small details that actually matter, while this rack does have so many "money-saving" advantages, it's the initial cost is the "be all end all".

Let me point out the fact that at least 90% of the miner's value is the hash board and power supply, so by not having to buy a new miner you aren't really saving a lot, just a few dollars worth of steel.

Also, the idea of swapping isn't something exclusive to the rack design, some company made hash boards that you could insert in your S9 and it becomes incredibly more efficient, yes you have to power the miner off while doing that and the firmware is different but it's very cost-effective, and given that you will only need to do this once in a blue moon then spending an hour in replacing 100 hash boards in air-cooled gear rather than 15 mins in the water-cooled rack isn't that much of a difference.

Replacing whole miners also isn't an issue if the economic incentives are there, so it really as I said in the begging of this post, it all comes down to HOW MUCH are they going to sell the Antrack for and how many hash boards can it handle, after that we can do the math.

okay I need to talk about gpu mining ⛏ but not because of altcoins.

A mobo
a cpu
a psu
cooling

multiple slots for gpus.

I have old mobos running new cards.

mobos from 2015 running 2020 cards.

So you are correct about swap in newer gear.

mikeywith is correct that a 1-2 cent guy mines older gear.

I could see a 4-6 cent guy in the usa 🇺🇸 shipping his 5nm gear used to a guy in china once 4nm gear comes out.

some of my mobos had

amd 480
amd 580
amd vega
amd 5700 xt

and if and when I get them

amd 6800 xt

the key is  don’t max your clocks 🕰

I had

nvidia 1000 series
nvidia 1600 series
nvidia 2000 series
nvidia 3000 series

all run on the same mobo.

If we get a rack we have three phase

if they start with 7nm
then 5nm
then 3nm
then 1nm

we could get close to 10 years from a rack.

Time will tell.

but the module upgrade and price improvements of never having to buy a complete machine again..just cards or modules from 5nm now to 1nm and all hot swappable and less space and water-cooled and mix/match

etc etc etc...if people move at large data halls 1/2 way around the world for a 2c kWh electric improvement...that would correspond to that kinda savings IMHO at least for all the next equipment of a big data miner guy

going to this format for any expansion.. too much in the way of hidden profits and upgrade to not do so I parsed out from all the crypto algos available from 5nm to 1nm say with a calc of 3=5 years averaged out in the

future. Also probably only need water cooling and ant racks about 1/4 of the space of traditional miner setups and cooling issues. again, too many things that save money combined for this not to be how everyone will do

their pow machines in the future anyway, you can see if I'm right or not 2-3 years from now from this post but that is my view now...combined with all the above little bit by bit advantages this is 'huge' indeed even if

you are right for the time being but for 'future' ASIC pow design this is set in stone IMHO for all mnfg's IMHO. chump or champ we will see in a couple of years.

be aware, however, this pans out Bitmain will mine them 3 months to drive difficulty up before letting them out the door in their own data halls ..and probably with more than one algo..,tis how Bitmain rolls don't ya

know! Bitmain "Evil Never Rests" (tm Bitmain)

Brad

I highly doubt the Chinese miners will be buying this anytime soon, now based on the assumption (which is a fact as of now) old gears still make a fortune while being so cheap, if you have the power cost advantage you will have to fully utilize it, I am certain that the prices of these racks will be super expensive and very unattractive, there is no way that this can compete with the regular air-cooled miners.

With that being said, since many miners end up losing money by making uniformed decisions, we can't tell for sure how things are going to unfold, there is a chance that 2021 could very much be a crazy bull year taking BTC prices to levels that make manufacturing less efficient but cheap gears more profitable, I mean a gear like T17 now makes almost $6 a day which is much more than it made when it was the most efficient gear, when mining is "over profitable" miners look for cheap gears rather than efficient once.

As I said in my post above..from an economy of scale situation...big miners in China will just kill on this. No Tariff. 2c-4c kWh electric. Less space (water-cooled). Hot swapping would allow you to mix and match pow algo

modules by day..in other words you could mix-match 5mh (say they are older) BTC modules (or hell they could be cards within a module) and say newer 4nm ETH or someday etc modules with hot swapping by the

day/couple days/week vs profitability. Bitmain will of course mine the hell out of any POW (you pick BTC/ETH/LTC) or whatever to the max 3 months ahead of pre-order sales of such with a release of 3 months after that .

...tis, how they roll...

I suspect the price will be attractive (like plug and play big boy toys...antboxes) to huge big miners in china etc...will pretty much IMHO be the death bell for midsized or small data halls for the rest of the world, Too

much capital to invest to get these up...only maybe big data halls play toys. Again, I think if you want to toss around $50k for a fully built antrack of your configuration not counting the building and water-cooling

setups

you will kill in china on these..in say texas at 4-6c kWh..meh...not sure how you would migrate if you already got on the s19 bandwagon but sure offers a lot more profit options vs stand-alone units on top of 2-4c kWh

for the China Miners vs the rest of the world. Bitmain: We Do Evil Well! (tm Bitmain)

anyway, re-looked above, most of you have seen Bitmain antics 2x to 3x along these lines before. I also will bet my own $$$ again, that the antrack concept will be where ALL pow miner makers go now...including

water-cooling etc, it just makes to much sense to get the big dollar folk to play. Also, as Bitmain also states...works in nicely with there 'supposed' everyone cloud mining dream you can buy into in the future (not) but

that is the hopem there they are pushing also besides above antrack. The data hall that spent 16 million in Texas for s19s must be kicking themselves (unless they are getting them sooner than January 2021 pre-

orders) https://cointelegraph.com/news/bitmain-inks-18m-bitcoin-mining-rig-sale-with-riot-blockchain. Anyway, they are in North America, again I think Texas.

Brad

In the pics I see 4 units in a rack with 4 PSU's per unit and 12 boards.

Lets assume boards are comparable to the S19 Pro -> 37 TH/s at 1050 watts each.
444 TH/s and 12,6 kW
total rack about 1.78 PH/s at 50 kW per rack
50kW is already achieved in HPC clusters with watercooling, so that sounds plausible...

Sounds great, hopefully the cost is acceptable

Myself, in my fantasy land of rainbows/unicorn farts/faries...in which I 'imagine' I'm a Bitcoin Billionaire..and how I'd approach mining ..the concept of AntRack would give me a BTC 'woody' indeed. Not Bitmain, per say, I'm not a fan, but the AntRack 'concept'. Indeed, I can see the brakes being put on 'right now' on any 'expansion' or adding another room or building to folks with big data halls if they are paying attention to the Bitmain announcement. Thus, right now, if I had to expand my mega data hall at 2c kWh or whatever...I'd 'on the get-go' add the permits and connections outlined by the Bitmain protocol above ...and water-cooling....which I was told is a whole nother permit process and hoops to jump through to get in place. But IF I had not already started on the new room/building or expansion..being the 'fantasy' btc billionaire I am..you could add these changes to your 'expanded' vision to incorporate Antiminer type protocols, easy enough. (I expect ALL major ASIC makers to adopt this idea of plug and play). So, being 'smart' BTC Billionaire (fantasy) and all...I get the units...put in place...they are 'upgradeable' from 5nm to 1nm supposedly...you can hot-swap...I 'assume' you could hot-swap different modules for different ASIC pow algos like ETH, LTC, BTC all in the same rack...etc, etc.

Yeah, this makes too much sense, especially if we are looking at a time line of Spring 2021 at best for this stuff and always remember Bitmain will run AntRacks in mass on their own data halls for 3 months to blow up difficulty before they go out the door to anyone else. So plenty of time to 'tweak' any big player, expansion plans as a big miner on any buildings or whatever...even if the concept fails and I don't use such Antrack connections...a small 'tweak' and plenty of time, with current plans, before 2021, to setup just in caese or likely, IMHO, that this would be the route a big miner would take. How I see this anyway, if I had the means to do so! (Satoshi contact me ...do I have a 'deal' for you! )

Alas, I would have been a fine/beneovolent/happy-go lucky/bitcoin billionaire ruler of all you minions on bitcointalk...alas....not to be!

Brad

Looks to me like the end of a lot of small mining ops.

It is 3 phase power which is more industrial than 240.
I wish specs got released.

10 k-watts and 500th each?

or 18 k-watts and 1 ph each?

Well, from what I can tell...if this is a 'big boy' data hall toy...and you are in China...at 2c to 4c kWh depending on the season....(and say you have the added advantage of being Bitmain and all this hitting your data halls for 3-4 months before everyone else. (whew!) Add to the fact that you can 'hot swap' our modules from 5nm down to 1nm as available...and I'd assume costs are cut by just this module swap...add to the fact you can be first in line for any 5nm to 4nm improvments on whatever algo/flavor crypto you want to put in this beast and indeed....a variety of crypto modules..say BTC at 4nm...eth at 5nm etc etc. Always having the 'cutting edge' of latest to pop in and or move at the same time ...rotating them through ..by the time the 5nm of whatever is no longer making you money at  your big boy data hall 2c kWh or whatever..you sell them to smaller minions ..in say 3-4 box racks...used...as you have made your $$$ and moved on. there is so much eff and ways to tweak such a setup with cheap electric if you are bitmain to your own uses...and mix match and get stuff like from 4nm online or lower faster...with this all in place it is 'dizzying' how much money they are gonna make.

But as to mid/size small data halls...er that boat has probably sailed by the end of 2021 with above enhancements assuming my guesses above have any validity (dubious..but still). indeed could be looking at a diff between CPU mining to ASIC mining ..skipping the step like GPU mining stage in between as a comparison of how this could blow the doors off for the big guys..

but don't worry...I 'forsee' that Bitmain will be MORE than happy to give you many 'cloud mining' options instead of all this pesky hardware that only big guys can afford. man...I got to admidt it .... Bitmain: (We Do Evil Well!) (tm Bitmain) Really does apply in this case....the setup seems clean to me and robust enough and expensive enough..that everyone will want to do it..but few will be able to afford such ..this is a ground floor build from the bottom data hall solution. but i you had big BTC/Crypto bucks..it makes as much sense if not more so than my first ASIC in 2013....

terrified pow wise but impressed!

Brad

Well I have 3 phase.  I can do more power this could help my room size issue.

Price
Power per unit
Hash per unit

I need to know.

Must I buy an ant box?

Ja. Canaan's Moose was a nice drop-in block of hash but did nothing to address the PSU and cooling issues.

The Antrack promise PnP replacement/upgrades. The n+1 PSU redundancy is a damn fine move to both increase reliability/serviceability and reduce the need for maxed out single PSU's.

Welp compared to the moose this looks like a better way to move to maximize density and efficiency. Not that I'm worried bitmain would just fall off the map but they need to be transparent about their support life cycle for this.

IMHO liquid cooling is long overdue. Coming from an industrial power systems background I can firmly state that it provides numerous advantages:

  Allows for very high equipment density
  Far quieter
  Far fewer fans to replace and no dust bunnies to remove as they gather and breed on numerous finned heatsinks
  The larger fans on the PSU's normally translate to lower rpm's needed and that generally = longer lifetime
  Far easier to move the heat outside via plumbing vs massive fans/blowers & air ducts

When properly done liquid cooling requires very little maintenance, just a filter and biocide to keep the water clean, if in an area that gets below freezing then glycol is also needed. Outdoor dry cooling towers are extremely reliable with near-zero maint required. Wet towers - ja they need a bit more monitoring but still very very reliable. With BM now offering 1-year warranties to me it make sense that they start getting rid of potential failure points such as the known heatsink misalignment/touching issues and the fans.

As for 1nm tech ROFL! Perhaps try asking in a few more years. There are better approaches to improving performance that are starting to come online that do not require the so far un-manufacturable smaller gate sizes.

Main features:



Source: https://blog.bitmain.com/en/bitmain-releases-rack-style-new-miner-bringing-next-level-computing-power/

This will be water-cooled, plug and play style, I am not sure why they decided to go with the expensive water cooling style, is it because the next 5, 3, and 1nm can't be cooled with the normal air cooling? While this seems like a great idea, I am pretty sure the price will be unbelievably expensive, the maintenance of the water cooling system isn't going to be cheap as well.

The good thing about it is that when you upgrade you don't throw away a whole miner, imagine if you could use the same S9 or even S7 and just have to swap the hashboard for the new once and/or PSU? how much money could one have saved through all these years?

Another point would be how likely are we going to get to that 1nm technology,  how much improvement in terms of efficiency do we expect?

What are your thoughts?