HashFast launches sales of the Baby Jet

[Puppet]

IIRC the power converters on Bitfury board are about $12 that is in bulk (i.e. 1000+ units).  Just the power regulators are $12 for 40 GH/s. = $30 per TH/s just for one non ASIC component.  Avalon board level BOM is open source (very similar to ASCIMiner who is very hush hush about components) and runs ~$10 per board (excluding ASICs, PCB and assembly).  That's $50 per TH/s for just minor components (crystal, resistors, capacitors, connectors, etc).

Yeah, and those are fair and honest comparisons per TH when asicminer needs 1200x more chips per TH than Hashfast. Im sure that has no impact on cost per TH whatsoever.

SO instead of pointless comparisons per TH with incomparable hardware, lets compare a 300GH Monarch to high end GPU's. High end GPU's have same form factor, similar powerdraw, similar cooling requirements, much more complex PCBs (14+ layers), loads of connectors, IO and non core functionality (DVI, HDMI, DP, PCIe, audio, Crossfire, etc..). This table is from Mercury research:


Take a look at the 6970. Eliminate the GPU and GDDR, and whats left is a BOM of $47 excl the asic (but including a load of stuff you dont need). But you think production cost of a monarch would be, how much exactly?

You say quality power isn't important

Please dont twist my words. Obviously efficiency is crucial. WHats far less important is reliability. Downtime of a KnC miner now in its first weeks is a huge deal. Not so much of a deal next year when your miners earn only marginally more than they cost in electricity, on a bad day it might even save you money . But I never said PSU's are not important, I said they wouldnt be included. THere is no PSU with a monarch, nor is there one with KnC. SHould miners calculate those costs? Of course. WIll they ? Very dubious, particularly people like me and assume you who still have piles of unused PSU's. And even if the cost has to be included, its shouldnt be written off in 1 or even 2 years. Those PSU's last much longer and have decent resale value, thats why in the cost calculation, they dont matter so much.

One last point about alibaba being so inaccurate; here you can buy a full ATX case, retail, 1 unit for 12 pound:
http://www.dabs.com/products/best-value-oem-717-midi-tower-with-hd-audio--no-psu--black-7XT4.html?src=3
Its not even the cheapest, I picked one that even looks decent. But you think $6-8 in large volumes is laughable and your $60 closer to the truth?

[RoadStress]

you think that mining vendors will live to see BIGGER volumes than ever? WOW...

In terms of hashrate (and thus wafer starts)



Of course unit volumes will skyrocket, almost no one is even shipping 28nm asics yet,  and current prices are at least an order of magnitude above marginal cost.
Dont tell me you thought the network was about to plateau at 5PH ?

Expect KnC maybe who already shipped ~3000 units.

Please make me understand your way of thinking. How do you expect someone to pay for NRE costs for 28nm asics and also sell miners very very cheap? I think a 28nm mask is around the sum of 1 mil $ (no ideea how much is it actually but i know it's expensive) and let's say you don't count any other cost then what company would make such an investment? How do you see such company's business plan?

Edit:

Take a look at the 6970. Eliminate the GPU and GDDR, and whats left is a BOM of $47 excl the asic (but including a load of stuff you dont need). But you think production cost of a monarch would be, how much exactly?

Do you think R&D is for free? Why don't you take those into your assumption? Apple increase R&D to 3% from 2% previous years...

[aerobatic]

you think that mining vendors will live to see BIGGER volumes than ever? WOW...

In terms of hashrate (and thus wafer starts)



Of course unit volumes will skyrocket, almost no one is even shipping 28nm asics yet,  and current prices are at least an order of magnitude above marginal cost.
Dont tell me you thought the network was about to plateau at 5PH ?

Expect KnC maybe who already shipped ~3000 units.

Please make me understand your way of thinking. How do you expect someone to pay for NRE costs for 28nm asics and also sell miners very very cheap? I think a 28nm mask is around the sum of 1 mil $ (no ideea how much is it actually but i know it's expensive) and let's say you don't count any other cost then what company would make such an investment? How do you see such company's business plan?

a 28nm NRE is several million, plus the first wafer orders (and possibly the second order too if you want more in a reasonable time) plus the first batch of all the parts to make the box.. so you're talking minimum $5m to start production!

-- Jez

[RoadStress]

you think that mining vendors will live to see BIGGER volumes than ever? WOW...

In terms of hashrate (and thus wafer starts)



Of course unit volumes will skyrocket, almost no one is even shipping 28nm asics yet,  and current prices are at least an order of magnitude above marginal cost.
Dont tell me you thought the network was about to plateau at 5PH ?

Expect KnC maybe who already shipped ~3000 units.

Please make me understand your way of thinking. How do you expect someone to pay for NRE costs for 28nm asics and also sell miners very very cheap? I think a 28nm mask is around the sum of 1 mil $ (no ideea how much is it actually but i know it's expensive) and let's say you don't count any other cost then what company would make such an investment? How do you see such company's business plan?

a 28nm NRE is several million, plus the first wafer orders (and possibly the second order too if you want more in a reasonable time) plus the first batch of all the parts to make the box.. so you're talking minimum $5m to start production!

-- Jez

Thank you. I would really like to see Puppet do some kind of simplified business plan or actually any plan how would a company invest this minimum amount of money and how would they sell their hardware in order to cover expenses and make a healthy profit in order to have those absurd prices per TH.

[Puppet]

a 28nm NRE is several million, plus the first wafer orders (and possibly the second order too if you want more in a reasonable time) plus the first batch of all the parts to make the box.. so you're talking minimum $5m to start production!

$5M is at the high end of my estimates, keep in mind a bitcoin asic in all likelihood is much simpler than typical soc's/gpu's/etc and requires only a minimum number of layers/masks,  but generally I agree and you are in the right ballpark. But so what? We were talking about marginal production cost, NRE by definition isnt part of that. NRE is a sunk cost, and therefore doesnt matter to pricing during the 'end game", when miners margins have dropped so low that vendors will have a choice between not selling anything, or selling something above marginal cost. If there is no longer a reasonable operational profit margin to be had, vendors may call it quits, but as long there is a profit margin to be had, they will sell (or self mine),  NRE be damned.  IOW  NRE will be a key factor determining if vendors end up making a profit or not (and IM pretty sure they all will), but its not a factor in pricing.

[DPoS]

next phase aint even pre-orders anymore..  that will be window dressing.. you will see the VC money roll in like that crazy blurb about BFL getting $million dollar payments from 'someone'
 

[aerobatic]

$5M is at the high end of my estimates, keep in mind a bitcoin asic in all likelihood is much simpler than typical soc's/gpu's/etc and requires only a minimum number of layers/masks,  but generally I agree and you are in the right ballpark. But so what? We were talking about marginal production cost, NRE by definition isnt part of that. NRE is a sunk cost, and therefore doesnt matter to pricing during the 'end game", when miners margins have dropped so low that vendors will have a choice between not selling anything, or selling something above marginal cost. If there is no longer a reasonable operational profit margin to be had, vendors may call it quits, but as long there is a profit margin to be had, they will sell (or self mine),  NRE be damned.  IOW  NRE will be a key factor determining if vendors end up making a profit or not (and IM pretty sure they all will), but its not a factor in pricing.

you talk of $5m like its nothing.  you say its sunk cost like you can forget about it.   The important thing is that its cash thats required to be paid, BEFORE, the chip goes into production... thus its cash that must be raised, somehow... which in the bitcoin world is via pre-orders since there is no other way of raising cash... as no investor is going to invest $5m+ in a risky venture like a bitcoin mining venture.

actually, one company did successfully raise the $5m needed to make their own asic without taking pre-orders - 21e6 LLC.   http://www.whogotfunded.com/deals/182363-21e6-llc

but no one else that i know of has successfully raised vc funding to make an asic chip, which means they must go through the pain and heartache for all concerned of taking pre-orders, til they hit the $5m level, and then they can tape-out.  actually, lets use the number of $4m (nre of $3m and $1m design cost)

And even after they've taped out.... the nre costs are not 'sunk costs' - because they haven't been spent yet.  Sunk costs are costs in the past that have been written off - but in the bitcoin world, no company has sunk costs.  The cost of the NRE must be raised in order to go into production - and because it is extremely low volume production the entire nre and other up front costs must be amortised or attributed to a very small number of chips manufactured.  we're not talking millions of chips like nvidia/intel/amd make...  we're not even talking hundreds of thousands of chips.  we're just talking of low numbers of thousands of chips... i.e., just a few tens of wafers.   And thats why the $4m up front costs will significantly and directly affect the price of each asic.

lets say, for arguments' sake that a 28nm wafer might cost $15,000 (in low volume, remember).  A Wafer is approx 70,000 mm² and a hashfast chip is 324mm² thus the max yield is 216 asics... but in practice, the yield will be less so lets say you can reliably get 150 good asics on each wafer.  Then that means that although the hard cost of each asic is approx $100 each, plus packaging/substrate costs... so lets round that up to $120-140 for a decent package that can handle a lot of power and is good for heat dissipation.   Now, since you may only make, say 10,000 asics in its short lifetime... (that would be approx 5 PH!) you have to divide the NRE and all other up front costs, and attribute them into the cost of each asic that you make - since none of the NRE payment applies to any future asic so its attribution is limited only to the asics you actually make... and $4m divided by 10,000 is another contribution of $400 for each and every asic.   YES, that *is* the way it works... you don't ignore your up front costs.  They are very real and they have to be amortised into each and every asic you make (since the volumes are so low, it becomes a significant part of the cost of each asic)

If you get to make more.. lets say you make 10 PH.. then you halve the amount, per asic, that your NRE costs will be attributed to each asic.

Thus the actual cost of each 400/500 GH asic is actually just over $500 (assuming 5 PH in total are made) and $300 (assuming 10 PH in total are made).

of course, the numbers get lower, the more asics you make...   but at 5-10 PH, per asic company, thats already a lot of PH's and there's a lot of asic companies so i think its unrealistic that any one company is going to sell more than 5-10 PH's on their own, in the lifetime of any one asic.

-- Jez

[Puppet]

Really, the big cost is the mask set. Once you have taped out and gotten your maskset, that is about the best example of a sunk cost I can think off. If that isnt one, I dont know what is.

Now of course these companies want to make a profit on their investments, which is why they will not get started if they dont think they can recoup the NRE, but once they decide they can, they will do what every company does: try to maximize profits. Right now your "$100 asic" can be sold for >$1000 because difficulty is so low and miners so delusional, so that is exactly what they do (and what made them decide to plunge down millions on the tape out). Not because the maskset was so expensive, but because they can.

But fast forward a few months, when difficulty is several billion and sales at the $1000 point dry up. What do you think is best, selling almost nothing or selling $100 chips for $500? You know, those kinds of margins would still even make intel green with envy.  And what when sales dry up at $500? Because they will sooner or later, you cant keep selling and shipping $500 asics without pushing up difficulty and thereby reducing mining profitability. So would you stop producing when you can still sell $100 chips for $250? Or $150? Maybe at $110 or $105 its not worth bothering with anymore, and some vendors may instead quit the business, but thats only when you reached near marginal cost and NRE is long forgotten. What you describe is one of the reasons you cant buy chips anywhere near marginal costs today (except for that batch of avalons, which may have been auctioned well below marginal cost, at $2 per asic for a complete miner), but it says nothing about where the market is headed when difficulty keeps rising.

Other example; lets take Bitfury. The have got their maskset. Lets assume for argument sake they also spent $5M on it and they have no intentions of doing a 28nm shrink. Will they keep ordering new wafers? As long as they can sell these chips, or mine with them for (substantially) more than they cost to produce (lets say $5), why wouldnt they?Why not sell these $5 chips on reels at $20 or even $10 when $20 becomes too much ?  Because the NRE was $5M they would rather not make a profit anymore? That makes no sense. THey may even end up selling below $5 if they ordered too much, just to get rid of the overstock.  $3 per chip is still better than $0. What does NRE have to do with that? What would it change if their maskset had only cost $500K or $50M ? Nothing, except when calculating their net profits. It simply doesnt factor in pricing or production decisions anymore, as its a sunk cost.

of course, the numbers get lower, the more asics you make...   but at 5-10 PH, per asic company, thats already a lot of PH's and there's a lot of asic companies so i think its unrealistic that any one company is going to sell more than 5-10 PH's on their own, in the lifetime of any one asic.

These upcoming 28nm chips are most likely the last ones we will see for many years. No one is likely to invest $3-5M on a new maskset 12 months from here, because you would be competing against chips that sold barely above cost and recouping the NRE would be almost impossible. And even if someone would, would they  be charging $1000 for their asic "because their NRE was x million", or will they charge as much as they can, which will only be  a fraction by then?  This is the end game and vendors will keep selling these chips (or deploy them) as long its worth the trouble of doing so. And that can only result in far more than 5-10PH for most of them.

[aerobatic]

Other example; lets take Bitfury. The have got their maskset. Lets assume for argument sake they also spent $5M on it and they have no intentions of doing a 28nm shrink. Will they keep ordering new wafers? As long as they can sell these chips, or mine with them for (substantially) more than they cost to produce (lets say $5), why wouldnt they?Why not sell these $5 chips on reels at $20 or even $10 when $20 becomes too much ?  Because the NRE was $5M they would rather not make a profit anymore? That makes no sense. THey may even end up selling below $5 if they ordered too much, just to get rid of the overstock.  $3 per chip is still better than $0. What does NRE have to do with that? What would it change if their maskset had only cost $500K or $50M ? Nothing, except when calculating their net profits. It simply doesnt factor in pricing or production decisions anymore, as its a sunk cost.

These upcoming 28nm chips are most likely the last ones we will see for many years. No one is likely to invest $3-5M on a new maskset 12 months from here, because you would be competing against chips that sold barely above cost and recouping the NRE would be almost impossible. And even if someone would, would they  be charging $1000 for their asic "because their NRE was x million", or will they charge as much as they can, which will only be  a fraction by then?  This is the end game and vendors will keep selling these chips (or deploy them) as long its worth the trouble of doing so. And that can only result in far more than 5-10PH for most of them.

Actually, in BitFury's case the numbers are way better since thats not a 28nm chip and the NRE's are much lower for 55nm than they are for 28nm... just like, when 14nm is available the NREs will be much higher than they are for 28nm... etc

anyway, maybe you're right.,. maybe after knc, hashfast and cointerra have made a bunch of their boxed systems, maybe their business will transition into a chip selling business and they will just churn them out for other board makers to sell on to customers...

selling chips sure is more profitable than selling boxes... so these asic companies should definitely get into selling chips as soon as they can.  however, these newer, bigger chips (knc, hf and ct) are more sophisticated than the previous little chips, and the circuits and cooling systems for the 'big hot chips' are not as strait forward as the systems for the little cool chips.... which is why these companies have to make their own boxes.. at least for the initial production runs... and then they can hand over their reference designs to the other board companies to start competing on cost and make cheaper boxes.

-- Jez

[Puppet]

Actually, in BitFury's case the numbers are way better since thats not a 28nm chip and the NRE's are much lower for 55nm than they are for 28nm...

i know, but the point is that it doesnt matter. Bitfury sells its chips at whatever price the market can bare and will decide on ordering more wafers solely on their marginal cost vs expected sales or mining revenue. Whatever their NRE was, at this point its irrelevant (unless you are bitfury and you are calculating your overall profits). And what most people dont grasp, assuming constant BTC price, every chip you sell or deploy, directly and proportionally reduces the market value of your next chip (and all already installed chips). So prices *must* come down Thats unlike any other market Im aware off. And also why preordering is so profitable for vendors, since it hides this value reduction.

selling chips sure is more profitable than selling boxes... so these asic companies should definitely get into selling chips as soon as they can.  however, these newer, bigger chips (knc, hf and ct) are more sophisticated than the previous little chips, and the circuits and cooling systems for the 'big hot chips' are not as strait forward as the systems for the little cool chips.... which is why these companies have to make their own boxes.. at least for the initial production runs... and then they can hand over their reference designs to the other board companies to start competing on cost and make cheaper boxes.

I dont know if they will all go down that route, but the market dynamics dont change fundamentally whether these companies sell only chips, or if they do everything in house. ITs a trade off between efficiency of specialization/competition and efficiency of scale. Its just that its easier to understand what will happen if you break it down in to pieces and consider asic vendors only selling asics.

[crumbs]

Speed (TH/s)   4081.00  Ding!

[DPoS]

Speed (TH/s)   4081.00  Ding!

groooowing

Network total 4272.397 Thash/s

[GigaWave]

Speed (TH/s)   4081.00  Ding!

groooowing

Network total 4272.397 Thash/s

Hit 4500 for a few minutes earlier.  Looks like it may have only been good luck though. Luckily.

[cedivad]

Hit 4500 for a few minutes earlier.  Looks like it may have only been good luck though. Luckily.

It looks like that someone had their chips at the end of this month and was able to bring everything online in 24 hours from start to finish.
Let me guess...

[iCEBREAKER]

Hit 4500 for a few minutes earlier.  Looks like it may have only been good luck though. Luckily.

It looks like that someone had their chips at the end of this month and was able to bring everything online in 24 hours from start to finish.
Let me guess...

If HashFast has our chips and is mining with them, that would be a serious problem worth of a Scam Accusation.

But there is no evidence of that being true (your irrational paranoia doesn't count as evidence).

It's funny to watch you become so desperate to bash HF that you're willing to stoop to such laughably bizarre tactics.

We can't even have a bit of perfectly normal variance without you eagerly mistaking it for an opportunity to spread lies about your nemesis.

Such fixation isn't healthy. 

You should trim the neckbeard and get out more.  Give Mom a chance to clean the basement, feed the cats, and do your laundry.   

[cedivad]

But there is no evidence of that being true (your irrational paranoia doesn't count as evidence).

Sure. Did i say anything about HF?
What's funny is you bashing every other competitor but HF. I've yet to get how much you are paid. Could you please repeat me that? You know, my english is bad...

[Easy2Mine]

If you can't stand the heat, stay out of the kitchen.
Cheer up and have another fucking pizza.

[cedivad]

If you can't stand the heat, stay out of the kitchen.
Cheer up and have another fucking pizza.

Who can't stand the heat? I'm having some fun here. It's the positive side of losing money like a fool, to troll the trollers.

[iCEBREAKER]

If you can't stand the heat, stay out of the kitchen.
Cheer up and have another fucking pizza.

Who can't stand the heat? I'm having some fun here. It's the positive side of losing money like a fool, to troll the trollers.

You can't counter-troll me!  I counter-trolled you first!   

I have no reason to bash HashFast because, unlike every other ASIC company, they are serious about building the best mining hardware on the planet.

How much are you getting paid by HashFast's competition go troll their threads with your silly, increasingly moronic and desperate FUD?

[Hoofprint]

If you can't stand the heat, stay out of the kitchen.
Cheer up and have another fucking pizza.

Who can't stand the heat? I'm having some fun here. It's the positive side of losing money like a fool, to troll the trollers.

You can't counter-troll me!  I counter-trolled you first!   

I have no reason to bash HashFast because, unlike every other ASIC company, they are serious about building the best mining hardware on the planet.

How much are you getting paid by HashFast's competition go troll their threads with your silly, increasingly moronic and desperate FUD?

hash fast could build and sell solid gold asic's
It wont matter how good the product is, if it wont ROI.

i wish they would change the MPP so its transferable, then i could sell my order an get some of the 45 btc i paid back