Cointerra Mining ASIC coming soon

[ralree]

Going off topic a bit here... But in 2014, the bulk of the sales of bitcoin hardware will be those products that are low cost, high performance and low power.  Ie: Complete Systems (not chips) selling for <$3/GH in early 2014, <$2/GH in mid 2014, and probably even <$1/GH by end of 2014 (assuming a die shrink to hit that figure).

There will be no die shrink in 2014 and probably not in 2015. 

+1.  20nm doesn't offer much over 28nm (30% speed increase, more density or less power - read TSMC's page on this).  28nm will be the standard for most of next year I think, if not longer. 

[1713922362]

1713922362

[1713922362]

1713922362

[1713922362]

1713922362

[erk]

Going off topic a bit here... But in 2014, the bulk of the sales of bitcoin hardware will be those products that are low cost, high performance and low power.  Ie: Complete Systems (not chips) selling for <$3/GH in early 2014, <$2/GH in mid 2014, and probably even <$1/GH by end of 2014 (assuming a die shrink to hit that figure).

There will be no die shrink in 2014 and probably not in 2015.  

+1.  20nm doesn't offer much over 28nm (30% speed increase, more density or less power - read TSMC's page on this).  28nm will be the standard for most of next year I think, if not longer.  

Intel have been doing some 14nm, but lot's of yield problems. 28nm will be the best ASIC price/performance point for a couple of years I reckon.

http://www.xbitlabs.com/news/mobile/display/20131017232018_Intel_14nm_Atom_Airmont_Processors_Are_On_Track_for_2014.html

[aerobatic]

+1.  20nm doesn't offer much over 28nm (30% speed increase, more density or less power - read TSMC's page on this).  28nm will be the standard for most of next year I think, if not longer.  

Intel have been doing some 14nm, but lot's of yield problems. 28nm will be the best ASIC price/performance point for a couple of years I reckon.
http://www.xbitlabs.com/news/mobile/display/20131017232018_Intel_14nm_Atom_Airmont_Processors_Are_On_Track_for_2014.html

I think if a newer process is available, then someone will use it.  all the people who rushed into 28nm (KnC, HF, CT, CC) will probably also rush into the next one just as soon as they can afford it (the NRE's are even higher).  Im assuming thats H2 of 2014, but maybe its early 2015 if the next die shrink is late due to yield problems.

[aerobatic]

+1.  20nm doesn't offer much over 28nm (30% speed increase, more density or less power - read TSMC's page on this).  28nm will be the standard for most of next year I think, if not longer. 

I think youve misunderstood the benefits of 20nm....  maybe you didnt understand so i will repeat it on each line...

1.  If it gets you 30% speed improvement... (all your hash engines go faster)
2.  AND it gets you 1.9x the density (double the number of hash engines)
3.  AND its 25% lower power (probably means more hash engines can be used without hitting thermal limits)

So we're talking quite a bit of potential speed improvement (and benefitting from lower power) at the same cost per die as 28nm.

If people can double the number of hash engines, AND, get 30% speed increase, AND have lower power... all from a die shrink, then thats three reasons to consider using it... outweighed by the huge cost of another NRE...  but if they can amortise that nre cost over enough chips made...

Also, i think everyone who has made silicon will find optimisations they can make...  whether they be circuit improvements, or relaxing the timing... (theres a trend in some of the designs to introduce gate closure timing errors to eek out faster hashing rates with some error allowance)... im not sure this concept wins much, but we'll see...

And for 14nm, whenever thats available... the differences are even greater.

-- Jez

[DeathAndTaxes]

+1.  20nm doesn't offer much over 28nm (30% speed increase, more density or less power - read TSMC's page on this).  28nm will be the standard for most of next year I think, if not longer.  

I think youve misunderstood the benefits of 20nm....  maybe you didnt understand so i will repeat it on each line...

1.  If it gets you 30% speed improvement... (all your hash engines go faster)
2.  AND it gets you 1.9x the density (double the number of hash engines)
3.  AND its 25% lower power (probably means more hash engines can be used without hitting thermal limits)

So we're talking quite a bit of potential speed improvement (and benefitting from lower power) at the same cost per die as 28nm.

If people can double the number of hash engines, AND, get 30% speed increase, AND have lower power... all from a die shrink, then thats three reasons to consider using it... outweighed by the huge cost of another NRE...  but if they can amortise that nre cost over enough chips made...

Also, i think everyone who has made silicon will find optimisations they can make...  whether they be circuit improvements, or relaxing the timing... (theres a trend in some of the designs to introduce gate closure timing errors to eek out faster hashing rates with some error allowance)... im not sure this concept wins much, but we'll see...

The bolded part is the fatal flaw of your logic.  Fabrication price per wafer @ 20nm is about 5x the cost as 28nm right now and it will be years before it hits cost parity.  The industry rule of thumb has been 2 years but (see NVidia slide at bottom) that has been slipping and the time before a new node reaches parity is taking longer and longer.  

However today lets assume all the fab marketing brochure-ware and give you the theoretical max of 30% higher speed and 1.9x density.   1.3 * 1.9 = 2.47.  So each wafer has 2.5x the hashing capacity which sounds great until you realize the nominal cost per wafer is 5x higher.   As a result you end up with 2x the cost per GH.   In reality you likely are going to get less, at more cost and with more potential delays, complications and risk.  When you consider lower yields, and the new higher NRE costs which have to be ammortized, the lack of preorder apetite and the fact that your product will be unattractive from day 1 conservatively you are looking at triple the cost per GH/s when comparing your state of the art 20nm tech to your 28nm competitors.   They can just set prices below your marginal cost and force you to sell at a loss or end up with mountains of expensive chips that nobody wants.

There will be no die shrink below 28nm in 2014.  It will be 2016 (maybe late 2015 and that is optimistic) before 20nm is cost competitive with 28nm.   Heres a hint.  AMD and NVidia have no 20nm plans yet.  That is how uneconomical it is.  These are companies which move MILLIONS of wafer starts a year, command the lowest marginal price, and an in a never ending battle to eke out a performance gain over the other.   They likely will extend another half product cycle at 28nm.   Don't point out Intel.  Intel owns their own fabs and plays their cards close to their chest.   Intel isn't a good benchmark for the rest of the industry.   For example 2012 was the first year TSMC had high volume cost competitive 28nm wafers ... Intel went full scale production on 22nm in 2011.  

And for 14nm, whenever thats available... the differences are even greater.

Sure .... eventually.   14nm high cost product will likely ramp up in 2016.  If it takes 3 years to achieve cost parity well you are looking at 2019(ish).  Given the need to offset NRE, and the sluggishness TSMC and other have had in getting new tech up to speed lets just call it an even 2020.

[Ytterbium]

I still don't get it how come they didn't raised 5 mil $ while HashFast raised 20 mil $ (seems a bit much), Avalon had 8 mil $ for chips and KnC managed to raise more than 5 mil $.

Because of how late they were shipping. KnC had no real competition for units, other then Avalon clone makers, and they started shipping in September (supposedly).  HashFast was supposedly going to ship this month (What's going on with them?) And Cointerra isn't supposed to ship until December.  Look at a graph of the difficulty and it's not hard to guess why they're selling fewer units.

[DeathAndTaxes]

I still don't get it how come they didn't raised 5 mil $ while HashFast raised 20 mil $ (seems a bit much), Avalon had 8 mil $ for chips and KnC managed to raise more than 5 mil $.

Because of how late they were shipping. KnC had no real competition for units, other then Avalon clone makers, and they started shipping in September (supposedly).  HashFast was supposedly going to ship this month (What's going on with them?) And Cointerra isn't supposed to ship until December.  Look at a graph of the difficulty and it's not hard to guess why they're selling fewer units.

Well KNC said "end of Sept".  HashFast said "end of Oct"  Cointerra said "end of Dec"  that is just SHA-2 industry talk for the next month so:

KNC - Oct
HashFast - Nov
Cointerra - Jan

The rest of it is on point.  Cointerra was/is simply too far out.   Trying to predict hashrate 30-60 days out if tough, trying to do it when 4 competitors will deliver product after you buy and before you receive is essentially impossible.

[aerobatic]

Quote...
The bolded part is the fatal flaw of your logic.  Fabrication price per wafer @ 20nm is about 5x the cost as 28nm right now and it will be years before it hits cost parity.  The industry rule of thumb has been 2 years but (see NVidia slide at bottom) that has been slipping and the time before a new node reaches parity is taking longer and longer.  


No fatal flaw. The cost of the entire system versus the die. It's somewhat low impact if the wafers cost a bit more as the rest of the cost in the box are the same.
Ie. chip costs. Eg. bumping. Dicing. Packaging. .. are the same.

System costs. Eg. Power supplies. Pcbs. Case. Cooling system etc. All of those are the same cost based on the number of parts.
and are much more than the cost of the actual die. I haven't calculated it but I'd bet the cost of the dies is less than 10% of the cost of the system. Faster dies makes the system much faster without adding much if any cost to the system.

[cointerra]

I'd like to see cointerra's input on this.

Of course no one in the ASIC industry can make any guarantees when it comes to ROI times etc. However, looking at diff increases historically and the ROI of CPU/GPU and first gen ASIC we foresee that our miners will indeed have quite a long time of very healthy ROI on arrival.

The basis for this is not primarily diff increase speculation but the fact that diff as a function follows the general rule that people will mine until their miners simply are not profitable any longer. We can see this in the plateau in the technological shifts so far and there really is no reason to expect a different behavior with Gen 2 ASIC. Meaning that when 28nm in this case becomes the 'standard' for Bitcoin mining, both GPU (already happened to a large extent) and Gen 1 ASIC will be forced off the network as a result of no longer being profitable.

What we expect here is that the dominant technology will be the best Gen 2 ASIC from a hash-rate/power consumed standpoint, and until Gen 3 ASIC hits the market these miners will provide ROI.

As CoinTerra produces ASIC with what is currently slated to be the most profitable hash-rate/power consumption calculation, we are confident that our miners will provide ample ROI as the leading technology for mining Bitcoin quickly and with low power consumption.

Sorry for the wall of text.

The CoinTerra Team

The problem is.... you see.... your customers will be COMPETING with YOU/YOUR TEAM in mining as well (so being the most efficient hardware is pointless) Guess who will have the upper hand?

If anyone here dont see this, they deserve to lose their bet.

We are not planning on setting up any form of mining operation before we deliver hardware to our customers. We also plan on offer very competitive hosting solutions allowing our customers to push our miners ROI as far as possible.

The CoinTerra Team

[Nemesis]

I'd like to see cointerra's input on this.

Of course no one in the ASIC industry can make any guarantees when it comes to ROI times etc. However, looking at diff increases historically and the ROI of CPU/GPU and first gen ASIC we foresee that our miners will indeed have quite a long time of very healthy ROI on arrival.

The basis for this is not primarily diff increase speculation but the fact that diff as a function follows the general rule that people will mine until their miners simply are not profitable any longer. We can see this in the plateau in the technological shifts so far and there really is no reason to expect a different behavior with Gen 2 ASIC. Meaning that when 28nm in this case becomes the 'standard' for Bitcoin mining, both GPU (already happened to a large extent) and Gen 1 ASIC will be forced off the network as a result of no longer being profitable.

What we expect here is that the dominant technology will be the best Gen 2 ASIC from a hash-rate/power consumed standpoint, and until Gen 3 ASIC hits the market these miners will provide ROI.

As CoinTerra produces ASIC with what is currently slated to be the most profitable hash-rate/power consumption calculation, we are confident that our miners will provide ample ROI as the leading technology for mining Bitcoin quickly and with low power consumption.

Sorry for the wall of text.

The CoinTerra Team

The problem is.... you see.... your customers will be COMPETING with YOU/YOUR TEAM in mining as well (so being the most efficient hardware is pointless) Guess who will have the upper hand?

If anyone here dont see this, they deserve to lose their bet.

We are not planning on setting up any form of mining operation before we deliver hardware to our customers. We also plan on offer very competitive hosting solutions allowing our customers to push our miners ROI as far as possible.

The CoinTerra Team

I never said you will have mining operation b4 shipping. As long as you have plan for a mining operation during shipping or even afterward, you will be competing with your customer. So your estimated ROI for your customer is MOOT.

Its also a conflict of interest.

Either you're purely operating as a hardware vendor for general public or you're not....(cant be producing hardware for your own mining operation or any private mining). Its very tempting because you can:
1. Fund your mining operation thro sales from your customers (oversold your products)
2. Having private deal with mining corps (oversold again)

Your customer got shafted either way.

[Nemesis]

Quote...
The bolded part is the fatal flaw of your logic.  Fabrication price per wafer @ 20nm is about 5x the cost as 28nm right now and it will be years before it hits cost parity.  The industry rule of thumb has been 2 years but (see NVidia slide at bottom) that has been slipping and the time before a new node reaches parity is taking longer and longer.  


No fatal flaw. The cost of the entire system versus the die. It's somewhat low impact if the wafers cost a bit more as the rest of the cost in the box are the same.
Ie. chip costs. Eg. bumping. Dicing. Packaging. .. are the same.

System costs. Eg. Power supplies. Pcbs. Case. Cooling system etc. All of those are the same cost based on the number of parts.
and are much more than the cost of the actual die. I haven't calculated it but I'd bet the cost of the dies is less than 10% of the cost of the system. Faster dies makes the system much faster without adding much if any cost to the system.

Make sense, i dont know the exact number but based on i've read the chip cost of say ...a GPU card, is quite low percentage wise of the whole cost of the card. So even if the chip cost is double, the cost of the card is only marginal higher.

[cointerra]

I'd like to see cointerra's input on this.

Of course no one in the ASIC industry can make any guarantees when it comes to ROI times etc. However, looking at diff increases historically and the ROI of CPU/GPU and first gen ASIC we foresee that our miners will indeed have quite a long time of very healthy ROI on arrival.

The basis for this is not primarily diff increase speculation but the fact that diff as a function follows the general rule that people will mine until their miners simply are not profitable any longer. We can see this in the plateau in the technological shifts so far and there really is no reason to expect a different behavior with Gen 2 ASIC. Meaning that when 28nm in this case becomes the 'standard' for Bitcoin mining, both GPU (already happened to a large extent) and Gen 1 ASIC will be forced off the network as a result of no longer being profitable.

What we expect here is that the dominant technology will be the best Gen 2 ASIC from a hash-rate/power consumed standpoint, and until Gen 3 ASIC hits the market these miners will provide ROI.

As CoinTerra produces ASIC with what is currently slated to be the most profitable hash-rate/power consumption calculation, we are confident that our miners will provide ample ROI as the leading technology for mining Bitcoin quickly and with low power consumption.

Sorry for the wall of text.

The CoinTerra Team

The problem is.... you see.... your customers will be COMPETING with YOU/YOUR TEAM in mining as well (so being the most efficient hardware is pointless) Guess who will have the upper hand?

If anyone here dont see this, they deserve to lose their bet.

We are not planning on setting up any form of mining operation before we deliver hardware to our customers. We also plan on offer very competitive hosting solutions allowing our customers to push our miners ROI as far as possible.

The CoinTerra Team

I never said you will have mining operation b4 shipping. As long as you have plan for a mining operation during shipping or even afterward, you will be competing with your customer. So your estimated ROI for your customer is MOOT.

Its also a conflict of interest.

Either you're purely operating as a hardware vendor for general public or you're not....(cant be producing hardware for your own mining operation or any private mining). Its very tempting because you can:
1. Fund your mining operation thro sales from your customers (oversold your products)
2. Having private deal with mining corps (oversold again)

Your customer got shafted either way.

Thank you for your input, and to clarify:

We do not have any set plans on running an in-house mining operation.

Our priority is and have always been to deliver the fastest and most effective Bitcoin mining hardware to our customers. I absolutely understand the logic behind your argument as this has been a subject of contention in the ASIC industry ever since the beginning, even in regards to burn-in testing on the test-net or on the live Bitcoin network etc.

We believe that the best companies in the industry can do is be transparent to the customers and this is something we intend to be.

[DPoS]

We believe that the best companies in the industry can do is be transparent to the customers and this is something we intend to be.

https://picostocks.com/businessplan/31

 

[maursader]

@cointerra, I have sent you a PM.

[aerobatic]

We believe that the best companies in the industry can do is be transparent to the customers and this is something we intend to be.

https://picostocks.com/businessplan/31

 

DPos -

ive seen several bizplans from bitcoin asic mining companies while they were in their original finance raising mode... and several of them originally thought they would raise large amounts of private money to cover the cost of production ($6+m) and then build a private mine.   But almost none of them actually happened.

The only ones that seem to have done it are the publicly funded ones (eg: AsicMiner, or the newer ones like IceDrill and 100/200TH that did it using other people's hardware.)

The only one that I know of that actually raised a big phat chunk of private money and stayed 'secret' is the '21e6 llc'.

As for Cointerra.. they werent the only ones who originally thought it might be a good idea to design an asic and start their own mine... but they didnt do that.  That would be INSTEAD of selling to the pre-order market.

Asic companies have two ways of raising the finance they need to start production (NRE costs are millions).  1.  they can raise from private investors.. in which case they wont be selling their asics as its more profitable to mine with them.. or, 2.  they can offer them to pre-order customers who are willing to put cash down.   once the asic company has followed track 1 or track 2, their destiny is pretty much set and theyre unlikely and unable to switch.

If an asic company DIDNT raise pre-orders, and doesnt need them, you should assume they will (most likely) NOT be selling their hardware to the public and (most likely) WILL be mining with it themselves (in secret?)

Now instead of private secret mines, the trend is for asic companies to offer their hashing hardware either directly or indirectly via resellers into the hashcloud market, where their hashpower is hosted and re-packaged into smaller and more affordable chunks that are more easily sold and much more decentralised in ownership.  these are 'no fuss, no muss' hashing power... sold by people like cloudhashing.com, cex.io etc

I really wouldnt hold it against cointerra that they originally may have thought they would raise lots of private cash and have a private mine, and instead, changed track, and offered their hardware for pre-order sale to end users.   Its pretty much what most asic companies have done... since the main people willing to take a risk in investing in asic mining hardware, are existing miners and bitcoin enthusiasts who are familiar and comfortable with the risks and rewards of the bitcoin world.   Regular financial investors arent.

[crumbs]

... 1.  they can raise from private investors.. in which case they wont be selling their asics as its more profitable to mine with them.. ...

Sorry for nitpicking, but selling chips is more profitable than mining.  AsicMiner is case in point, with business model involving sales of demonstrably unprofitable gear. 

[aerobatic]

... 1.  they can raise from private investors.. in which case they wont be selling their asics as its more profitable to mine with them.. ...

Sorry for nitpicking, but selling chips is more profitable than mining.  AsicMiner is case in point, with business model involving sales of demonstrably unprofitable gear.  

thats ok to nitpick.  i really dont think its as clear cut as that.

if you are building a secret mine.. your boxes dont need to be user friendly.  they may not need their own controllers or power supplies cos you can build out shared facilities.  they might not even be air cooled.. as you can use facilities' water to cool them externally.

building a secret mine, you dont need a sales department nor a customer support department.  you need minimal skeleton staff to keep your mine running and you dont need any externally facing employees at all.

building a secret mine, you can overclock and overdrive your own hardware much harder than you would allow a customer to do it.. since if it blows up, its your own fault and you can take the risks you wouldnt let your customers take.   In fact, the hardware can be more bare bones... you may not even need a case for it.. you can simply slot bare boards in racks.

and ultimately, selling hardware requires you to have a markup to justify selling it.   whereas, if you are buying at cost price, you get to mine at a lower cost.

thus, i argue that operating a secret mine could well be more profitable and controllable, than taking pre-orders and selling boxes to customers.  It all depends on how easily you can sell and at what price you can sell to customers.   Since Cointerra is selling their TH's at a lower price than anyone else, i think its safe to say that they arent making the same profit per box that the other guys are selling, so you cant exactly accuse them of profiteering.. can you !?

(edit) Actually, to nitpick a bit further and take your words literally, you said Selling CHIPS was more profitable.. and i think thats the key.  Selling CHIPS is indeed much more profitable.  But most of the new guys dont sell chips... or, tried and didnt find traction.. maybe these new 28nm chips are too complex and require too sophisticated a board.. or maybe they need to try harder, as clearly, if they were to sell chips, they dont have the huge box costs associated with them and its a much better business to sell chips than to sell fully built working boxes.

[DPoS]

  Since Cointerra is selling their TH's at a lower price than anyone else

Until they ship that is an open statement.   

As we all know from ASICminer, they sold their hardware as high as they could and only dropped price when conditions merited it (competition, diff increase)

KNC could slash their 1st gen prices right when Cointerra starts to look like they are shipping Dec/Jan (as they said they will look into doing things if competitors actually deliver) and also they have a 2nd Gen chip that may come out whenever too

Hashfast might be in the same driver seat if they deliver soon


I'll stick to my belief that Cointerra's real business model is to try to make a splash and get a greater fool to take over on the venture capital level.  They have raised the area for the bagholding. Instead of just the usual late pre-order customers getting the shaft, it will also be a late VC that wants to get into this 'bitcoin thing' that is left with a bigger bag

Look how bitfury is screwing some customers with how they are providing more to the mining pool as priority..if a VC buys Cointerra out, the pre-order customers can consider themselves screwed as the VC can see a fixed cost of developed hardware into short term mining gain.  Right now it is too muddy to think of it..  but not when it gets closer to an actual hashing machine to run through some cost/return numbers

[cointerra]

We believe that the best companies in the industry can do is be transparent to the customers and this is something we intend to be.

https://picostocks.com/businessplan/31

 

We will be communicating more information on the Picostocks listing. In short I can say that the Picostocks listing is not CoinTerra as a corporate entity selling shares but rather Picostocks as a company investing in CoinTerra and selling a form of pass-through ownership. We have a conversation open with Picostocks and will be making a clarifying statement regarding this.

[fractal02]

... 1.  they can raise from private investors.. in which case they wont be selling their asics as its more profitable to mine with them.. ...

Sorry for nitpicking, but selling chips is more profitable than mining.  AsicMiner is case in point, with business model involving sales of demonstrably unprofitable gear.  

But this business model is dead now...

See my sig for the best business model right now