Should Bitmain make a Q7?

[QuintLeo]

This disparity is what enabled the violation of Moore's law.  The closing of this gap occurred in 2014.

 Agree entirely with the first part - Bitcoin ASIC initial tech was a few generations behind the "state of the art" when the first ASICs started being designed.
 The gap has NOT entirely closed though - ASIC tech is still behind the state of the art, just not as MUCH behind it now.

  The relative depression of BTC exchange rate in 2015 is why you are only now seeing new 28 nm chips being sold.

 Several designs from late-2014/early 2015 were 28 nm, BM1384 being probably the best known "home miner" example.
 I'm pretty sure I remember the KnC Neptune/Titan and I believe the Spondoolies "Rockerbox" based machines like the SP20E and SP35 were 28nm as well.
 Bitfury also announced they were shipping 28nm around that timeframe, but since they went "no end user sales, just BIG corporate sales" it's hard to verify details on them.

 I do think that the drop in BTC price did kill off some of the weaker/less competant manufacture folks. Too bad it didn't kill off KnC. 8-(

  I don't believe Bitmain can double efficiency again with a 28 nm chip ( hats off to them if they can ).

 I don't think they'll bother trying, they probably have tapped 28nm out pretty close to it's limit.

  Unless the BTC exchange rate goes up significantly I don't believe it is cost effective to produce a 14 or 16 nm asic.  

 Then explain why KnC has announced it's already deploying 16nm (Solar).
 Explain why both Bitmain and Bitfury have announced they're working on 14/16nm "next gen" designs, and I believe all 3 have announced plans to deploy "full custom" designs on one of those tech nodes (Bitmain definitely has, Bitfure and KnC's claims about their NEXT gen imply it).
 SOMEBODY thinks it's worth working towards the current state-of-the-art, even where they won't deploy it for likely another year or more.



 BTW - my CPU and GPU on my primary machine happen to use the same identical technology. AMD A10 based custom build.

 There weren't as many Bitcoin ASIC makers "back then" as you seem to think - don't forget to ignore the scams.
 I think it's fair to count the folks that actually did the work to deliver, even if they were like BFL or KnC and ended up ripping folks off through VERY VERY late deliveries and/or failing to meet performace spec announcements.

 Bitfury and KnC are more likely to RoI with their own 'cause they don't have "distribution markup", than because the designs are not cost effective. That's the one BIG disadvantage home miners face, we pay a LOT more for the machines than the manufacture does for the parts and labor to build them.

[sidehack]

Who has made their own chips in the past? Bitmain, Avalon (both as Avalon proper and Canaan-Creative), Spondoolies, BitFury, ASICMiner, KNC, Hashfast, Cointerra, Innosilicon, BTCGarden, BFL, 21e6 and I know I'm missing a lot more that are now defunct. There were a lot of vaporware scams, but there were also a lot of folks that did actually produce a chip - between 2013 and 2014 probably three times as many as are still doing it.

I believe the Neptune used a 20nm ASIC? KNC started at 28 when everyone else was doing 55 and then went down from there.

From a practical standpoint, the "state of the art" itself is still behind the state of the art. 14nm is about the best thing going but it's still so immature that even the guys that invented it still have trouble making things work right half the time.

[fullzero]

This disparity is what enabled the violation of Moore's law.  The closing of this gap occurred in 2014.

 Agree entirely with the first part - Bitcoin ASIC initial tech was a few generations behind the "state of the art" when the first ASICs started being designed.
 The gap has NOT entirely closed though - ASIC tech is still behind the state of the art, just not as MUCH behind it now.

  The relative depression of BTC exchange rate in 2015 is why you are only now seeing new 28 nm chips being sold.

 Several designs from late-2014/early 2015 were 28 nm, BM1384 being probably the best known "home miner" example.
 I'm pretty sure I remember the KnC Neptune/Titan and I believe the Spondoolies "Rockerbox" based machines like the SP20E and SP35 were 28nm as well.
 Bitfury also announced they were shipping 28nm around that timeframe, but since they went "no end user sales, just BIG corporate sales" it's hard to verify details on them.

 I do think that the drop in BTC price did kill off some of the weaker/less competant manufacture folks. Too bad it didn't kill off KnC. 8-(

  I don't believe Bitmain can double efficiency again with a 28 nm chip ( hats off to them if they can ).

 I don't think they'll bother trying, they probably have tapped 28nm out pretty close to it's limit.

  Unless the BTC exchange rate goes up significantly I don't believe it is cost effective to produce a 14 or 16 nm asic.  

 Then explain why KnC has announced it's already deploying 16nm (Solar).
 Explain why both Bitmain and Bitfury have announced they're working on 14/16nm "next gen" designs, and I believe all 3 have announced plans to deploy "full custom" designs on one of those tech nodes (Bitmain definitely has, Bitfure and KnC's claims about their NEXT gen imply it).
 SOMEBODY thinks it's worth working towards the current state-of-the-art, even where they won't deploy it for likely another year or more.

 BTW - my CPU and GPU on my primary machine happen to use the same identical technology. AMD A10 based custom build.

 There weren't as many Bitcoin ASIC makers "back then" as you seem to think - don't forget to ignore the scams.
 I think it's fair to count the folks that actually did the work to deliver, even if they were like BFL or KnC and ended up ripping folks off through VERY VERY late deliveries and/or failing to meet performace spec announcements.

 Bitfury and KnC are more likely to RoI with their own 'cause they don't have "distribution markup", than because the designs are not cost effective. That's the one BIG disadvantage home miners face, we pay a LOT more for the machines than the manufacture does for the parts and labor to build them.

First:  State of the Art is not mass produced.  Why?  In general for two reasons; first, reliability is a concern with new untested unproven designs.  Second, because it is not cost effective to do so.


You seem to think I said 28 nm chips are new; look again with my assisted emphasis.

"The relative depression of BTC exchange rate in 2015 is why you are only now seeing NEW 28 nm chips being sold."


Again you seem to not notice key words; look again with my assisted emphasis.

"Unless the BTC exchange rate goes up significantly I don't believe it is cost effective to PRODUCE a 14 or 16 nm asic."


Designing a chip requires paying engineers to design it; this is a very limited cost which can be done at any time.  Actually producing that chip is where things get expensive.


I never claimed KnC or Bitfury didn't have next gen chips.  You seem to have missed this.  I was suggesting they have investment concerns beyond mining.  They support XT strongly; why do you think that is?


Your A10 APU uses 28 or 32 nm tech depending on the model.  That should tell you something about the relationship between cost effectiveness and state of the art.


Obviously a company is going to markup their prices; what you seem to miss here is that they won't bother to make miners to sell if it is not profitable to do so.

[QuintLeo]

State of the Art is not mass produced.

 Not when it first appears, but it does get there - and is often STILL the state of the art when it DOES achieve mass production for 1-2 years in semiconductor fab.
 I don't count "experimental in the lab only" as "state of the art", BTW.

The relative depression of BTC exchange rate in 2015 is why you are only now seeing NEW 28 nm chips being sold.

 The BM1384 (among other examples) wasn't new in late 2014 when it was first released?
 You seem to have a very narrow definition of "new".
 Perhaps you meant "second generation on the 28nm process node" instead?

 Unless the BTC exchange rate goes up significantly I don't believe it is cost effective to PRODUCE a 14 or 16 nm asic.

what you seem to miss here is that they won't bother to make miners to sell if it is not profitable to do so.

 It appears a FEW companies disagree with your belief, and you just shot your OWN "belief" statement down with the "profitable" statement given the facts.

 I'll remind you that at least one company claims to have had 16nm chips IN PRODUCTION for a while now (KnC), though I concede they aren't selling to end users.

 On the other hand, another (Innosilicon) has already announced tape-out of 14nm implying "in actual production" soon (December timeframe given Lktec's announced miner etc. etc.) and given Innosilicon's track record I'm CERTAIN they will be making miners shortly after the chips hit mass production.

[TheRealSteve]

Who has made their own chips in the past? Bitmain, Avalon (both as Avalon proper and Canaan-Creative), Spondoolies, BitFury, ASICMiner, KNC, Hashfast, Cointerra, Innosilicon, BTCGarden, BFL, 21e6 and I know I'm missing a lot more that are now defunct. There were a lot of vaporware scams, but there were also a lot of folks that did actually produce a chip - between 2013 and 2014 probably three times as many as are still doing it.

https://en.bitcoin.it/wiki/List_of_Bitcoin_mining_ASICs
Discarding HASHRA (questionable at best by now), not counting the recent vaporware / scam announcements, and counting GridSeed/SFARDS as one: 25. That's missing others still for which I couldn't verify enough to warrant listing as of yet.  MBP may have (had) their own chip (claimed anyway), a subsidiary of a larger Chinese company that made some disappeared (along with most of the Chinese offerings - throwing spaghetti at the wall and seeing if it'd stick type ventures), and then there's the ones I don't even know that I don't know them.
(should anybody be reading this and thinking "that list is missing X" - please do contribute; if you can't edit the wiki, shoot me a PM with details)

In terms of reasonably successful worldwide deployment, it's really just a handful.

[sidehack]

While typing out that list I was saying to myself "TheRealSteve is gonna drop in here at any minute with a big fat list".

I knew it was many, but not 25. Daggum.


--------


The BM1384 was quite new when it was released in 2014, but he's talking about 2015. Avalon hasn't done anything this year, Spondoolies hasn't done anything this year, ASICMiner went out of business before they could do anything this year. Right now it looks like BitFury['s private mines], SFARDS (well, sorta - sha performance is mediocre for the price) and Bitmain are the only ones with a new 28nm so far this year, as opposed to how many companies with new chips last year? If coin price was better, demand for miners would be higher (and efficiency requirements a bit more lax) and more folks would be pushing out gear - just like we saw in 2013 and 2014. It's not a complete argument (as in, not all variables and effects are accounted for) but the point is valid. We have seen basically no new chips in any node size this year, compared to the previous two years.

I'm looking forward to seeing what Innosilicon does for a new chip.

[fullzero]

State of the Art is not mass produced.

 Not when it first appears, but it does get there - and is often STILL the state of the art when it DOES achieve mass production for 1-2 years in semiconductor fab.
 I don't count "experimental in the lab only" as "state of the art", BTW.

The relative depression of BTC exchange rate in 2015 is why you are only now seeing NEW 28 nm chips being sold.

 The BM1384 (among other examples) wasn't new in late 2014 when it was first released?
 You seem to have a very narrow definition of "new".
 Perhaps you meant "second generation on the 28nm process node" instead?

 Unless the BTC exchange rate goes up significantly I don't believe it is cost effective to PRODUCE a 14 or 16 nm asic.

what you seem to miss here is that they won't bother to make miners to sell if it is not profitable to do so.

 It appears a FEW companies disagree with your belief, and you just shot your OWN "belief" statement down with the "profitable" statement given the facts.

 I'll remind you that at least one company claims to have had 16nm chips IN PRODUCTION for a while now (KnC), though I concede they aren't selling to end users.

 On the other hand, another (Innosilicon) has already announced tape-out of 14nm implying "in actual production" soon (December timeframe given Lktec's announced miner etc. etc.) and given Innosilicon's track record I'm CERTAIN they will be making miners shortly after the chips hit mass production.

Context is important. 

"The normal profit model for Bitcoin ASIC has been dependent on the ability to use older tech to make comparatively cheap chips for their mining value.  This was possible because of both a high BTC exchange rate and the disparity between chips a company can cost effectively make and their relative efficiency to those already in existence.  This disparity is what enabled the violation of Moore's law.  The closing of this gap occurred in 2014.  The relative depression of BTC exchange rate in 2015 is why you are only now seeing new 28 nm chips being sold.  I don't believe Bitmain can double efficiency again with a 28 nm chip ( hats off to them if they can ).  Unless the BTC exchange rate goes up significantly I don't believe it is cost effective to produce a 14 or 16 nm asic."

I was not making a statement about the existence or non existence of any particular 28 nm chip.  The crux of what I was saying is that you are only now seeing New Chips; regardless of their type.  Why has Bitmain decided to Produce their BM1385? maybe in response to competition from Innosilicon; but I guarantee you they wouldn't sell miners with new chips unless they felt they would be losing market share by not doing so.  Do you think Bitmain just finished designing the BM1385?


"they won't bother to make miners to sell if it is not profitable to do so."

Here I should have said "a company will only produce a miner if they believe it will be profitable to do so."
I believe Innosilicon believed making a 14 nm chip would be profitable when they began production.  This does not mean they still believe this is the case; nor does it mean companies don't make epic mistakes.  Windows Vista; epic fail.

[QuintLeo]

You keep SAYING that "only now are new 28nm chips being sold".
This is NOT accurate as stated, as I've already pointed out.

 I think what you mean to say is "only now are NEXT GEN 28nm chips being sold" or some such, which has a TOTALLY DIFFERENT meaning from what you are actualy STATING.


 Personally, I think your entire argument is based on false assumptions. SHA256 ASIC has seen "gaps" in new tech introduction before, the current one is more about the low bitcoin price forcing out some of the marginal performers than about the cost making new tech "not effective".

 The other factor clouding things is that many of the remaining SHA256 ASIC have gone out of retail, making it harder to verify when they had "new tech" in production - KnC and Bitfury in particular seem to have introduced and started production on "new tech" this year, but non-BIG-commercial-miners don't see their product in the wild so they end up under the radar and not registering as having appeared.


 Bitmain has been producing their BM1385 for a while now, we just don't know EXACTLY how long as the initial production went into replacing THEIR OWN miners.



 Windows Vista wasn't an epic fail, though it was at least arguably a fail.
 It says a lot that Vista support was ended by M$ about the same time (same date, or a few months LATER than) they ended XP support.

 Windows 8.x, Windows ME, and ESPECIALLY MS-DOS 4.0 were the EPIC fails of M$ operating systems.

 Says a lot about MS-DOS 4.0 that is led to M$ having it's first WIDE BETA of an OS (MS-DOS 5.0) - and they didn't do that again 'till the current Windows 10 beta which speaks to how epic the Windows 8.x fiasco has been.

[notlist3d]

One thing I thought of is they managed to sell 1/3 more chips per S7.  So were likely talking 1/3 bigger profits per unit sold.   So in that view it was definitely a business decision on what S7 looks like, and it's power.

I still don't think they will take time to make a "Q7".  They will likely make a new U model but it will be much less power I would guess.

So our chance of a Q7 is if they sell chips.   Which I hope they do.  But I don't see it happening for a while if they do at all.  But I think it will take a 3rd party to get the 2 blade unit we expected.

[mavericklm]

i think they will leave quiet miners only for hobbyist, they will not invest the time and money for this!

[notlist3d]

i think they will leave quiet miners only for hobbyist, they will not invest the time and money for this!

I think were in agreement.  They showed this by launching with 3 blades, everyone thought it would be 2.

I think there will be a rise in miners in US with 240 and putting them somewhere other then sharing a room. Before a lot of us heated rooms with miners and it was nice.  I likely will use old gear to heat rooms, I don't know how much new gear will heat up the houses without being to loud.  But I could be off on this.

[mavericklm]

we need to get inventive with the fans and airflow
think 2 or more fans ducted thru one s7!

got another crazy idea, put the 3 blades to make the mercedes sign, but upside down! from the front, the miner to be seen as an Y with controller somewhere at the top.
and from here to get creative with the fans!
forgot about the 6pins...

[Biodom]

Q7 is a worthwhile idea, but i think that they will go to Antrouter instead for a quiet miner.
It will have S5 chips, and as such it will not be anything special as far as productivity is concerned.
All manufacturers had pretty much abandoned the $400-600 market that home miners favor.
Maybe it is more difficult than we all think to make something that will sell in this price area.
Why not "make something twice as loud for three times as much" pseudo logic seems to be winning.

[dog1965]

I have an S5+.  It is badass, the S7 will be even more badass.  

However; the S5+ is loud and the S7 will be loud.  

This is fine for prosumer miners ( those who have basements and garages, or co-location, or who make custom mining rooms / sheds / buildings ).

But there is another huge market in the USA and many other countries; supplemental heat.  Winter is coming, and the S7 will be just too damn fucking loud.  

A miner that uses 180 watt max per heat sink (150 optimal) in an S1/S3/S5 design ( no need for the S3 metal casing; good job with the S5) would be amazing.  Space heaters are sold in the USA.  A Q7 could most likely compete with the cost of oil or propane heat in New England, USA.  If you made this, I would push it 100% in New England, USA.

I am willing to bet that most USA home miners would be willing to pay a 15% premium to have a quiet miner.  

I think this is worth your attention Bitmain.

Please, please make a Q7.

Fellow miners please petition your support.

-fullzero

I am willing to bet that most USA home miners would be willing to pay a 15% premium to have a quiet miner. 



You pay 15 percent pay 100 percent while your at it. This is why bitmain jacks the price !

[Blawpaw]

Sure, yeah! We really need miners that can be a lot more silent so that we can use them to heat our rooms this winter!
that would be really great! 

[VirosaGITS]

But the AntRouter isint really a worthy miner. Its like a single chip running on 1-2W. Maybe it would run as a plug and play node, which in itself could be of a certain use but that seem pretty much like it. 8-15ghs isint worth fawning over.

A quiet S5/S7 on the other, now that sound nice.

[sergio]

A quiet S7 would be a great thing.

the problem is now that there is no competition, the S7 is sold at a price of $1823 USD or 7.5 BTC, the real price if BTC goes up is in BTC, and if BTC goes down it is in dollars.

If BTC goes over $300 dollars, the $1823 will go up, most likely over $2000, and if BTC goes down then the $1823 stays, unless if it goes down a lot, the miner price will not proportionally goes down.

The pricing scheme seems to be calculated at an amount you will never ROI.

https://bitcoinwisdom.com/bitcoin/calculator
This calculator takes into consideration difficulty increases.

at a 10% difficulty increase ( 20% is the default) and free electricity you will not ROI in over 530 days, and will end up 1.7 BTC short or ROI.

at a 5% difficulty increase and free electricity, it takes 296 days to break even.

Let compute the fair price of an S7, using difficulty increase of 10%, and 15cents per kilowatt, and a ROI of 3 months:

you will make 1.2 BTC, loss of 6.3 BTC.

lets use difficulty of 5%, and 10cents the killowat, for 6 monts:
3.5 BTC made, 7.5 expended, loss of 4 BTC.

Using this calculations it can be seen that the S7 will never ROI, and it is extremely overpriced, more than double what it should cost.

from a hardware point of view it is nice, but it does not help if it will never ROI, I would personally buy a few if it was priced at a price in which I could ROI, or even break even in 6 month, but the fact is that I will not even break even at 2 years.

Do not believe me? use the calculator yourself and plug in the numbers, even with free electricity you will not ROI.

Most likely price will go down, since few are buying, the item has not sold out in many days, but Bitmain has to seriously stop being greedy and lower the price by at least 60% if we are expected to have the opportunity of at least reaching ROI.

I am hoping that BTC doubles in price, but that will not help, since if BTC doubles in price, Bitmain will double the price of the miner.

The real price drop will be soon, and that is when the block halves, there will be a huge market of used bitcoin mining equipment, that will make all prices new or used hardware  to drop, I am personally waiting for that day, or the day in which I can get hardware that will ROI in about 4 to 6 months at most.

[alh]

Besides the power/noise profile of the S7, the price is a real deal breaker for me. Unless and until Bitmain has a reason to reduce the prices (e.g. serious competition), they will get a premium price for the S7. I think any BTC price machination are strictly cover for having Bitmain adjust the price the way they want. We saw this during the early days of the S5, and never made sense to me, It always seemed more related to what Spondoolies was charging.

[notlist3d]

Besides the power/noise profile of the S7, the price is a real deal breaker for me. Unless and until Bitmain has a reason to reduce the prices (e.g. serious competition), they will get a premium price for the S7. I think any BTC price machination are strictly cover for having Bitmain adjust the price the way they want. We saw this during the early days of the S5, and never made sense to me, It always seemed more related to what Spondoolies was charging.

I was surprised with price.  I bought a S5+ just as I was bored and wanted a new miner.   I am happy with it but I do think I overpaid most likely.

On S7 1800 dollars.... looking back I think I got my S4's new for around 1200 dollars.  Dont remember batch.  But seems miners are going up in price.  Unless someone else puts out a miner we are likely to keep seeing high prices. 

And to the topic I don't see them making a Q7.  Noise is not something they worry about I don't think.  It's become a mod if you want quiet.

[Biodom]

Besides the power/noise profile of the S7, the price is a real deal breaker for me. Unless and until Bitmain has a reason to reduce the prices (e.g. serious competition), they will get a premium price for the S7. I think any BTC price machination are strictly cover for having Bitmain adjust the price the way they want. We saw this during the early days of the S5, and never made sense to me, It always seemed more related to what Spondoolies was charging.

I am close to 100% sure that they can charge $1000 for S7 and still make a hefty profit.
After all, there are only 2.7 times more chips in there. $350X2.7=$945
the rest is cable and plastic plus $15 for a extra fan=less than $1000 with similar profit margin as S5.