Topic: Cointerra Mining ASIC coming soon - page 7. (Read 35558 times)

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

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.

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.

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.

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.  


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.

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

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.

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

+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. 

Just thought it was funny that you were implying they weren't going to comment, when they commented just a few hours later.  No ill will, my friend .

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

I hear what you're saying, but I think very few have that opportunity available to them.  It's gonna really suck having to watch this from the "outside" once my current Asics cost more to run then they earn.

I hope come Summer of 2014 things start to look alittle more clear for us miners and the whole network doesn't consist just 6-12 Large Corporate Miners mining using the ASIC design that our pre-orders paid for to milk to Network for all it's worth.     Keeping my fingers crossed that they understand Bitcoin enough to realize a Centralized Network will ruin Bitcoin so no more $200,000+ a Day for 25% of the network.

problem is you are still in the old gpu miner mentality ...  with the stakes raised people can play 'cost basis' tricks on the old ledger.. etc

lol how>?>  by them saying a big good luck?

Exactly, the miners buying equipment will always be the ones losing in this scenario.  Just look at Bitfury, https://ghash.io/ they just added another 100TH/s Today so up to 625TH/s now!!  Meanwhile October orders for the V3 Bitfury Hash boards are still asking Dave where there gear is since it's running later than Promised...it looks like the answer to that question is Right here https://ghash.io/ once again.

We're gonna keep losing big until we make our stand and demand Gear that is priced much cheaper so as to be able to ROI before the Miner no longer pays for Electricity.

I just don't understand how anyone is still getting any Pre-order business anymore either.  Some must just have alot of money to blow...

bitchslapped. 

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.

Have yall thought of a hosting option? I am a native of Dallas, travel to austin once in a while, and run a hosting company out of Missouri (Somehow power here is cheaper than deregulated Texas... strange). I would of course be happy to meet in person etc, An excuse to fly home and see family in Dallas and SA.

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

Hello maursader,
we would love to have you visit our office and meet the team. I will PM you and we can set up a date for you to visit our offices and the beautiful city of Austin.