Topic: Do you think Bitmain make better than 16nm in future? I think no. (Read 2889 times)

50/50, given the LOW volume, that those "12nm" chips are going into Bitmain's new "deep learning AI" related project.

For reference, all of the "12nm" processes that has been announced to date are NOT on a "new node", they are 14/16nm node process refinements and the "12nm" is a marketing tag.

Similarly, the announced 7nm node stuff out of TSMC and GF *appear* to be about the same level on actual gate feature sizes as Intel's 10nm node (not sure where Samsung's 10nm node sits yet, but I suspect it's close to Intel's).

Given that every 14/16nm node production line is running at full capacity, I have no idea why TSMC and GF are worrying about "getting the most business" via markating exaggerations.


Anyone else remember "Web 2.0"?
I am starting to sense a similar stench of "Microsoft Marketing" getting even more contageous....

I think of course they have a roadmap and plans for upgrading their process. If they are smart they already have developed technology we don't know about yet. They can use internally for better profits in their farm, and most important be ready in case a competitor comes up with an "S9 killer". They have no rush pushing it to the market though, the S9 is still king, but it would be stupid to just take a nap and wait for a competitor to push them out.

Also: What is Bitmain doing with the new 12nm chips they're buying in 2018?

i suspect they are, Ebang is working with Samsung for their 10nm device that is supposed to be shipped at the end of Jan. Though, i am no expert in the semi-conductor profession and there are plenty of members on this site that do have the experience, so id love to hear their feedback as an update. I was reading for a while that NO company would put out 10nm miners for a while because 10nm process is only used in low power devices but i guess things have changed since ebang is doing it.

It is a year I've been thinking about this, so far has something changed? Do you think they working on 10nm or so?

What about Bitfury?

they have zero need to do that.

they make a fortune with hashnest.
how many of the s7s below  are really s9's?

they earn money on them if all are s7s.  so when an s7 dies they stick in an s9  and they earn more.
as long as hash nest does well they do not need to make a better s9

oh coins are way up over 930

 BTC on the rise = HIGHER profitability on all gear and even the S5 might be profitable again if your electric is cheap enough. The S9 will be profitable LONGER if BTC price keeps climbing (which doesn't look likely for the short run, it's already dropped back into the mid-$800s).

 The limit on S9 production appears to be foundry space availability, and I don't see it dropping in price much for a while given it's competiton has the same "can't get lots of production of the CHIPS" issue.

Now that btc is on the rise and almost at $900 I don't think the S9 will last long. We will see the S9 pumped out to the max soon until its at $400 a unit new like the S7 was ground down to. There will be unlimited sales at this price.

I think we will see something new in 2017 that improves on the S9 by about 30% and sells for more than it should. This will finally be the generation of gear that will stick around for about 2 years. But I think the diff will keep going up regardless as the big miners try to squeeze each other out. In the end, in about a year or two, I don't think anyone anywhere will be making a profit.

And yes the miners will mine at a slight loss for quite a while because of the original capital invested in "data centers" and such. They won't all just walk away the first year of a negative return.

 I can see the possibility of somewhat better than the current BM1387 being possible, but I doubt it's going to be a major improvement like Bitmain pulled off on the 28nm generation.
 30-50% better efficiency should be possible, but given that the 14/16 nm node chips aren't running at much lower voltages than 28nm was capable of, there's some serious limit on how much more efficiency is available to achieve.

 We've already had at least one early 14/16nm design that proved to be little or no better than the best 28nm designs (LK-1401, the apparently-abandoned Innosilicon A3 might be another given the final efficiency ratings on the A4) and BW already claimed some major improvements going to the LK-1402.

 Under 0.10 watts/GH? Possibly. Under 0.06? Looking MIGHTY iffy at best, even the Bitfury demo chip didn't manage that at it's lowest power setting.

 I'm a bit supprised we haven't seen SOI move down to the newer node range before now - did I miss something, or did SoI completely skip the 22nm genration?

There are some marketing tricks that you don't know and that's why you speak so. Every hardware gives users something new and sayes that there is something new and innovative, behind this they have already better version but they aren't going to sold it because they need to sell less better product at first. Same situation is here with bitmain and what will you say about smartphone facts? huawei, apple and lg created smartphone with two camera at the same time?

So I think there is already released what you think that isn't.

but a better 14-16nm is possible.

I think maybe an improved s9+  next spring or an s11  but either more of the same chips on a lower clock.  or a better chip that is still 14-16nm


to see a 10nm  maybe 2020 at best.

What's sure is that they won't go further than GPU manufacturers, so if those only passed the 14nm mark, we'll wait a long time before we see 10nm or so.

GLOBALFOUNDRIES Fab 1 in Dresden, Germany is currently putting the conditions in place to enable the site's 12FDX development activities and subsequent manufacturing. Customer product tape-outs are expected to begin in the first half of 2019.

 We're still looking at probably 2020 timeframe for someone to start building miners on that node, is anyone goes that route even if GF manages to hold to that schedual.

 Also have to wonder how much of that production is going to be tied up by AMD's aggrement with Global Founderies reguarding that fab (which used to be an AMD fab).

It's secondhand, but I was told BW would start receiving the first batches of LK1403 in October. That was right about the same time I found out they wouldn't be selling to the public, which was right after I had a working prototype for a BW-based stickminer, including cgminer drivers, ready to roll.

http://www.globalfoundries.com/newsroom/press-releases/2016/09/07/globalfoundries-extends-fdx-roadmap-with-12nm-fd-soi-technology

From what I know, the GTX 1060 or the 1050 and 1050ti were manufactured on a 14nm process, and AMD uses all 16nm iirc.

I wouldn't be surprised if we end up in a mining deadlock, where efficiency can't go much higher and electrical costs will need to be insanely low to even make a small profit. We know for sure that future advances will be slow, but every little advancement counts.


7 and 10nm won't come right after mainstream processors on those manufacturing processes come out, as ASIC devs always need time to develop chips- this delay could be from a half year or less, which will certainly increase time before miners are introduced.

14nm and 16nm are on a practical basis interchangeable - I forget which of AMD and NVidia is set up for production of their cards on BOTH nodes offhand, but that's just one example.


 10nm and 7nm however should be significant moves forward - but I think Intel and IBM (respectively) and TMSC are all bit optimistic in their projections of when they'll be bringing these new nodes to market.
 7nm in particular requires completely different "ALLOYED" wafers of a type that have never been done before, which is going to take time to get production ramped up.


 BW should have their LK-1402 second-gen part and miners based on it in production, but not a 3'd gen and haven't even announced plans for such a part.


 A721 was definitely rushed, as Caanan didn't even announce they were STARTING on a miner design 'till very close to the same time BirFury announced their new part and demo of sample, and shortly before the S9 started shipping. I'd guess the chip in the 721 is closer to the LK-1401 and isn't a "ful custom" design, as opposed to the LK-1402, BM1387 (S9), and the Bitfury part.

 As I recall, the Bitfury part is 14nm, I *think* the LK-1401 and LK-1402 are both 14nm but might be misremembering on those.

 
 Based on all of my reading, I'd guess we won't see 10nm or 7nm in miners before early 2020 - and I won't be shocked it it's more like 2021 sometime.

 I remember some of the early predictions of 14nm and 16nm process node "would be available" almost 2 YEARS before they finally were, due to the teething pains - then TSMC got hit by the earthquake and their shipping scheduals slipped a couple more months....

BW has 14nm ASICs; from what I understand, they're already using their 3rd revision of 14nm (LK-1403) design.

Bitmain will most likely continue working with TSMC, so I guess a 10nm chip is to be expected from them next.

http://www.eetimes.com/document.asp?doc_id=1329217

I could see 14nm coming in a year or two, but it's likely that performance jumps won't be very big at all and that 14nm might just be skipped in the bitcoin industry.
I was pretty shocked by the efficiency of that miner; I expected something at least close to the S9 in terms of base hashrate and erficiency. Maybe it was rushed? It's a nice low cost miner and relatively low power, but I just don't see it being adopted as much as some others.

Yeah, the bottom-clock efficiency of the BM1385 at 28nm is only a little bit worse than stock setpoint of Avalon's new 16nm; I think it's spec'd down to about 0.18J/GH chip-level and I've seen an S7 get 0.22J/GH wall. It ain't easy but it's doable.

Ja. The '132 and '134 were laid out using standard automated chip design tools for the 28nm node - nicely functional, faster/cheaper to use but far from the best way to do it. Usually the biggest thing that suffers is timing of data to/from the cores. In other words - speed.

The '135 circuits layout inside them was 'hand-optimized' for pure performance. In other words people actually looked at the layout tools output and ran it through simulations, go back to the tools and tweak, re-sim, etc. several times to get highest throughput. Hell, given that by then 28nm tech was pretty much mature (read: lowish-cost) Bitmain no doubt had several runs of test wafers ran to verify results each time.

As a result he '135 was still 28nm node but performed much better than its 'quick and dirty' precursor. OF course also helped that Bitmain started using the 2-sided heat sinks with the s7 and were able to push more power into them but still able to keep a good GHs/J spec...

For a current (last October) write up of node-size What's To Come, http://spectrum.ieee.org/semiconductors/devices/leading-chipmakers-eye-euv-lithography-to-save-moores-law

They say 'maybe as soon as 2018' (emphasis mine). Not specifically said is that those chips will first be destined for SOC's integrating high-end CPU, GPU and FPGA-style processing with all attendant support circuits in one chip. Eventually will trickle down to boutique chips but don't hold your breath.

BM1385 is the S7 chip and is 28nm; BM1387 is the S9 chip and is 16nm.

To toss some more wood on the fire.

The 16nm chip from bitmaintech is meh.

now if I recall

BM1382 is 28nm

 https://enshop.bitmain.com/files/download/BM1382_Datasheet_v3.0.pdf

and BM1384 is 28nm


so if BM1385 is 16nm

BM1387 could be 16nm  and it would work better

as for a 10nm  3 or more  years from now 

and for a 7nm 5 to 8 years.

And just pray tell who Bitmain would have making 14nm chips for them?
TSMC does not and will not have anything at that node, their next target is around 10nm. Intel and IBM Foundries are for-internal-use only and Samsung mostly is as well which leaves GloFo and they would be the first to announce someone is using them... There are NO other Foundries on this planet able to produce 14nm chips much less anything smaller.

Just look for the next big hashrate jump on the blockchain. That will be the newest gear being switched on. Most companies in any field of research and development have things in the pipeline years before they get released. Bitmain is no different they would already be working on more efficient miners before the S9 was even bought out. The chip development is one thing but the programming and communication of the chip into a pcb and eventually miner is another. They probably already have 14nm chips but just making money out of the 16nm chips for as long as possible (look at the prices of the miners, they don't go down when they release new batches) then once they have figured out how best to incorporate the 14nm asic into a miner they will build them for themselves and earn from them inhouse. Only once someone else builds something competitive they will release it to the public and earn even more.

Lets just hope when that day comes that its better than the last few batches of S9's.

Like every technology makers...

I think they already have some better equipments made..Just waiting that everything is sold, making money, and when competitors will try to make their better asics available, they will release another better miner for customers.

I think they already 1 year or 2 in advance into miners, but not releasing those at the moment. ( marketing )

14nm to 10nm is a huge, huge gain again so eventually the benefits will outweigh the costs.

I think bitmain now going to produce, mine and not for invest for development new chips for maybe years...
Because 14nm is not such move forward to invest much money what it need.

What you think about? Because if I am right, we can start nice investment in miners today. S9 will be here for long time.