Topic: Swedish ASIC miner company kncminer.com - page 1794. (Read 3049501 times)

They are only inventing problems, as the answers would take reading. It comes down to effort, and unfortunately 'money printing machines' draw the crowd that dislike work...

It's not like there aren't enough troubles in the world, people have to invent a whole slew of hypothetical problems to get upset about, too....

I'm not sure how to feel after reading this^^^.  You have [what seems to be] a clear understanding of what's happening, you're not intentionally deluding yourself, so its hard for me to understand the appeal of paying someone to bring about centralized mining.  ASIC is the crack of bitcoin mining, ASIC rigs are no more a hobby than a $500/day dope habit -- sure, we've chipped for a while and we were bound to get strung, but must we *pay* somebody to bring that wretchedness about *sooner*?

Leichenträger zu wache?
(Awful close to breaking Godwin's Law here, it's a quote from Vonnegut's Mother Night, German for "corpse carriers to the guard house."  
Apparently concentration camp inmates were allowed to work carrying corpses to the cremation ovens.  The lulzy part is that there was no shortage of prisoners willing to do it -- it was seen as a privilege.)

That would make sense if the script didn't read "Shipping on Production day X"
 "Production Day"    not      "Shipping Day";  which means to me, they will be ready to ship on the same day they are produced.
(not taking into account the UPS man may not get it until the next day, although the COULD be driven to a UPS location same day)  
But I also see your point, which is comparable to his(btc_usr) scenario 2

Bitsyncom.

lmfao

As for the debate about ASICs streaming online and BTC prices dropping because of it....the people who are buying now are doing so with an eye to their purchases breaking even or in the case of some older orders from other companies...making a pretty good profit...this will change and indeed has changed when you consider all the people refused refunds (elsewhere) who no longer want to be in the game now there's nothing in it anymore.
Less people with better kit, that seems likely to me. Not thousands more diving into mining for a possible loss.

I remember KnC mentioning litecoins, maybe that will one of their future projects? Once this game gets too much for private buyers.

The only things I remember them saying are:

that shipping would begin in September and
units ordered now, today, would be still be shipping some time in October.  

And when the shipping dates start stretching out into November, they'll be dropping the price.   That means they estimate they can make or ship X units per day.  Where in the world does the idea that they plan to ship day1 and day 2 orders, then take a break until October come from?

Isn't there a simpler explanation?  Say KNC can finish final assembly, testing and shipping of 200 units/cores per day (where Jupiter = 4 units).  They won't ever stop just ship out 200 units a day starting from order 1 until the last order.  So 200 units on shipping day 1, 200 units on shipping day 2 (400 total), .... 200 units on shipping day 10 (2000 total).   Now for the sake of argument lets say they had 1,200 units ordered on day 1 and 800 units on day.  If they started shipping on Sept 20th then it would take them 6 days to ship out all the units ordered on day 1 and 4 more days to ship out the units ordered on day 2.  That takes you to the end of September and orders from day 3 onward ship out (200 at a time) starting on Oct 1.  If you ordered on day 1 (in this hypothetical excercise) your order wouldn't necessarily ship on Sept 20th it would ship on a day between 20th and 26th depending on where you were in line.

The numbers are just illustrative and obviously no company is going to be able to ship exactly x units a day everyday.

I believe the fellow may have a reading dysfunction of some sort.

  If you were here, I'd have you tarred & feathered.  

I asked knc about this, if there was any delay between September and October shipments, this was there response,

Hi,

 

We are aiming to ship every order that has been paid for as soon as it is produced according to the shipping queue. Our first orders are expected to ship in September and we will then continue to ship out the orders working down the shipping queue list.

 

Let us know if you have further questions.

has anybody actually received and orders?

Lets not obsess about the efficiency. In the broad scheme of things it doesn't really matter. What matters is TTM (time-to-market or time-to-mine). Wise men know that a birdVrm in the hand is better than two birdsVrms in the bushcatalog. The best Vrm module is probably the one that is the easiest to obtain and has the highest number of alternate suppliers and pin-compatible replacements.

Layout and board fabrication aren't trivial. Layout is actually extremely critical for a repetitive circuit that is simultaneous-switching-noise-limited. If you want to read a recent story take a look at the Enterpoint's Cairnsmore1 thread. They also apparently thought that they know how to put 4 chips on a PCB and look how much time it took for the developers to actually distribute the clock signal properly.


I don't share your concern about the importance of the rise time. I imagine that after the power-on all hashing engines are disabled. Then the SO-DIMM Linux computer will run some tests on the hasher cores and selectively enable and tune the PLL clock synthesizers. Thus the ramp-up could be quite gentle.

Thank you very much, guys. The problem is that there isn't sufficient information to make a meaningfull comment. I don't want to be like Amy Woodward from HashFast and start making some half-assed and quarter-brained assumptions.

If you want a productive subject for the speculation about the PCB, here's a one.

The Vrm requires some way of setting the output voltage. Thus far most of the dedicated mining boards had that voltage set using a pair of surface-mount resistors. That's the cheapest way. The most flexible way would be to have some software-programmable Vrm that can change the output voltage on command. But that is a risky feature to offer when KnC is offering a warranty and everyone will try to overvolt and overclock.

There's also am old-fashioned way: use a trim-pot to set the voltage. During the factory testing tune the voltage with a screwdriver and seal the trim-pot using lacquer, glue or epoxy before shipping to the customers.  To protect against tampering, use some lacquer imported from Chernobyl and test it with a Geiger counter when somebody returns a burned-out miner under the warranty. 

So what would you guys want to see in the shipping configuration:

A) surface-mounted pair of resistors
B) tamper-evident trim-pot
C) software-programmable voltage divider.

There is yet another possible scenario...  They start shipping a week or two before the end of September, like you say...15th or thereabouts. (My order says "Shipping on second day of production", not "Shipping on shipping day 2" With that in mind, (production not stopping because they said not batch shipping either) the next units should flow out in the days immediately following.
Granted, If production was to begin Friday, the 27th, scenario 1 is perfect.
But I'm expecting a sooner "Day 1 of production" as many of us are hoping...
there's no way "Production day", OR "Shipping day 1,2,3,4,5,etc." are going to be separated by weeks imho, which leaves scenario 1 or 3(this one), so I'm thinking/hoping that there will be around 10 days of September that they may ship, just to surprise everybody, or at the very least Monday, the 23rd, for 5 or 6 days of production in Sept. (All speculation, of course)
Someone in the forum earlier said they heard a very specific date...  curious about that.
I'm going to throw a KNC Production day 1 party if I'm even close...   None of my friends will even know what that is, but they'll sure eat the cake & drink the beer! I'm still extremely optimistic in case nobody noticed... 

Potentially, there is a wave of hysteria Bitcoin mining is riding on culminating from both the attention sub ECB theft in Cyprus and the subsequent media furore and with that of the tech transition to ASICs which has become something of a race. Whenever profits are to be made, especially from something as ludicrous as plugging in a machine, people will dive in, and as you've seen on this forum, many diving head first without looking. You have fools wanting to turn a quick buck at one end, and private equity where marginal profits spread over greater investment and cheap electricity still give greater gains than traditional investments at the other. Somewhere in between you have clued up geeks who get the tech and know what they are looking for. For now I see KnC potentially riding in pole position for their customers. Technically ASICminer, Bitsyncom and Butterfly Labs have poled 1, 2, and 3, for the manufacturers trophy, but proportionately very few of their customer base have benefitted. ASICminer to an extent, more so if you we're fortunate to hold shares from there initial offering, but not from their stock, Bitsyncom and BFL have both kept to bare minimum happy, but in reality, one way or another they've all scalped the lions share of profits for themselves with Bitsyncom and BFL knowingly shafting the majority of those that have supported them.

Amusingly when purchasing the coins, the greater interest clearly leads to a rise on their value, where as with mining the increase leads to a reduction in profit. I see mining inevitably going to areas with cheap renewable electricity and shared ownership.

For now though there is a window of opportunity to profit from owning the next gen tech. Asics haven't arrived yet in the grand scheme of things, BFL have fumbled the ball, Bitsyncom have some huge explaining to do, as their current excuses are total crap, Bitfury is on track, and the main 28nm competitor I don't think is being at all honest with respect to their 'anticipated' October timescale, I maybe wrong, but they definitely don't want people holding them to it.

So in answer I have some coins, they've done nothing exciting of recent. Short of major and favourable regulatory legislation from the Western world, or Far East, or for that matter another central bank theft, or a major international incident forcing savers to leap to Bitcoin as a vehicle for currency protection I don't see huge movement in the next six months. If KnC deliver on time, I'm confident in at least having an entertaining gamble work out. We'll (us plural, reading this thread) have to reconsider our options come January, but w.r.t. KnC they clearly have ROI covered and are very smart and competitive, so I expect to see them adapt and continue to offer those that support them a favourable deal. They aren't about to hang up their gloves and retire if this works out. ORSoC want to remain fully involved with crypto, they were very specific about that at the openday. Plus, clearly they are forward thinkers, they had the nouse to hit 28nm ahead of everyone else and break records for bleeding edge tech development, whilst limiting the risk for those supporting them, so expect plenty more innovation if they pull this off. I just hope they blatantly reward those that support them with advance purchase of next gen, preferred pricing, etc.

I'm not an "inductorologist".  Noone is.  It was a sarcastic statement someone else made.   I am a mathematician who works in the electronics field and to be honest, I typically don't design our power systems, I work on ICs.   However after > 10 years of working side by side with the people who do, you tend to pick up a few things.   Of course my circuit uses off the shelf components.  I wasn't indicating my design suited their purposed, was only mentioned to illustrate doing better doesn't cost much time.  As previously mentioned, I've got ~20 hours into my design, most of which was going through data sheets, setting up spice sims, and altering schematics.  80% is crap efficiency, especially at > $20 BOM cost for the module (modules are generally more expensive).  All designs start as schematics, excluding trivial ones.  Layout and board fabrication is trivial.  Factoring in extraneous variables ahead of time is the difference between winning and losing.

Solution:
Get any one of the dozens of modules that have > 90% efficiency at the voltage and currents desired and internalize the power supply.   Overall same heat inside the chassis, better performance per watt.  Both things that are key factors in the market they are trying to sell into.

IF Day 1 & Day 2 orders are shipping in September
  AND everything else is shipping in October THEN

there are 2 possible scenarios:
1) day1 is Friday 27th, day2 is Monday 30th, so everything beyond day2 is October
2) they ship day1 for ex. on Monday 16th, ship day2 on Monday 23th and rest in October

I remember that they mentioned continuous delivery, not batch based.
So scenario 2 would mean boxes are ready to ship, difficulty increases, but they are not shipping them... oh boy, ppl will freak out and travel to Stockholm knocking on KnC's door to get their pre-paid miners I guess.

This implies from my point of view that scenario 1 is the more likely one.
Then I'm curious to see how ppl will behave when difficulty increases during September, I'd not be surprised if many buyers will cancel orders.

Please point out in case I'm wrong or missing sth.

I figured you were an inductorologist. (you claim to be anyways)
I'm curious if your design uses off the shelf inductors.

I clearly stated that better designs were possible.

Your design is still on the drawing board.
Their supply is pictured mounted on a printed circuit board.

Why would you design a supply if there are such good choices available already?
Is it possible that the price and/or availability was a factor in their decision?

My point is that it is likely good enough for the application.
It comes down to price and availability.
They had to pull the trigger a while ago.
All designs involve compromise.

As an inductorologist you might naturally have different ideas in your field of study.
Do you also design ASICS, and have connections to get them to market quickly?
There's money to be made in them thar hills!

Why don't you contact KnC and see if they are interested in your design for future products?
You could also sell retrofit parts for rev 1 miners.
According to you everyone who purchases a miner will need one.

If you are right and the rise time prevents proper function of the miner I will admit I was wrong.
On the gripping hand,,,
Will you do the same if the roles are reversed?

Perhaps one of us should "die in a fire immediatly" if we are wrong.
You do advocate such things.

Time will tell

OK, so you are saying that even though the calculator says that the miner will not break even, it might end up having long-term profitability if the BTC exchange rate increase keeps compensating for the electricity costs. But if that is the case then wouldn't the difficulty also be increasing a lot with even more miners jumping in? And thus don't we get back to the situation were it is better to buy BTC upfront and forget about the miner?

 






 

Exactly what you said is the answer.  IC's are not typically designed to handle under current very well and typically things like enable signaling is left out of the design on mining hardware for the reason you indicated, they are always running.  The slow rise times, as I'm inferring that you know from your previous knowledgeable posts, have adverse effects, especially on larger dies.  I don't have any extra cards here, but one could test this by pulling the GPU BGA, attaching a sequence of increasingly sized load resistor, and measuring rise times.   Following this, one could add rise delays to another card, which should be easy given the required pour patterns of most h-bridges, and measure GPU errors at different delays (going from idle to load is analogous to the problem we're discussing).