Topic: CoinTerra announces its first ASIC - Hash-Rate greater than 500 GH/s - page 82. (Read 231002 times)

CoinTerra is dead or what ?

No news...

my bad, didn't read the whole thing 

I know. Earlier on this thread i stated that i saw somewhere that bitfury chips run at 2W/GH, but i couldn't remember where. Now i just posted proof that i wasn't talking out of my ass like most people here on the forum.

Exactly.  The burnin boards are 50 GH/s the "standard" H-boards are ~33 GH/s.  Higher clock, higher voltage, higher power consumption. 

Zefir is talking about the boards from burnin, mine are the c-scape/bitfury/bfsb. Burnin clocks the chips higher.

So i knew something...

My dear White Knight, the world is not limited to what you happen to be "aware" (or in this case, unaware) of.

I can't educate you if you don't pay attention and assimilate all the relevant facts:

-CryptX's principals were indeed seed investors in CT, hence their desire to double-dip by gathering public funds for their private pockets

-CryptX tried to copy Icedrill with their Petamine project

-Petamine failed to sell even 5% of its initial shares, then unilaterally changed the terms of their IPO (no voting rights, funny numbers, etc.) like typical scammers

-Icedrill sold out their first two rounds of funding but the market was too exhausted for the third round (much less CryptX's Petamine shenanigans)

you have a very active imagination.  most of what you say is not close to the reality and/or complete fabrication.

I'm trying not to knock any particular team because i prefer to stay pragmatic as much as possible.  your warped one-sided view has completely distorted your ability to comprehend the reality, and your trolling doesn't work here.

btw, cryptx were i believe the first investors in cointerra... and I'm not aware if they ever tried to do an iceDrill style group buy with them.. and i understand cointerra has raised private investment in more than one round and has been stepping up the valuation at each round.  I'm sure they've continued to raise private investment and are well north of $1.5m by now.

And what makes you think that delaying their tapeout by 4 weeks had a $1.5m cost to them?  why would that have any direct cost to them (beyond paying expedite fees, which are thousands not millions!)

Also, you mistakenly think the IceDrill ipo was a huge success.  It went ok.  no one's complaining... but in reality i think all parties thought it was going to be more popular and raise a lot more cash than it did.. and the order was scaled back accordingly.   i think had it been an investment in the company, rather than a group buy, it might've done better (and be more similar to AsicMiner in concept)

this will always be a dilemma...  no one seems to want ultra low power designs right now.

people want max performance..   so if you DID design a lower power asic, people would end up over clocking it more, and making it less efficient.. just so it would run faster.

the irony is that absolutely any of the 28nm asics could also be 0.4 W/GH/s, if they were undervolted, but no one actually wants that (in the short term).  they prefer to over-volt them and run them faster!

"Around the same time?"  Didn't I just tell you I saw, and participated, in the market for the entire duration of that time?

Listen Brave Sir Jez, your attempt to gloss over ("exactly the same...applies equally") the crucial distinctions I've drawn isn't going to fly.

IceDrill's first two IPO rounds completely drained what little leftover capital was available after months of ASICminer, Avalon, BitFury, KnC, and ACTM pre-orders/group-buys/stock bubbles.

IceDrill's third round never sold a single share, just as CryptX's initial round was a colossal failure.

Since you insist on being thick, allow me to succinctly pierce your shell of ignorance:

IceDrill (helped pay HashFast NRE as early bulk investo-omer): TWO successful IPO rounds, raising over TWO million dollars

CryptX (attempted to help pay Cointerra NRE as early bulk investo-mer): ZERO successful IPO rounds, raising about ZERO million dollars

Clear enough?

In the exorbitantly costly 28nm ASIC sector, $1.5 million is a medium-sized investment.  Large-ish begins at $2M.

If Cointerra hadn't botched their first attempted tape-out, it might have been enough.  But they blew it, and wasted that $1.5 million fixing their unforced error.

"Botched tape-out" is the last thing one may truthfully accuse HashFast of.

HashFast has several large investors thanks to their timeliness, while Cointerra has far fewer because of their tardiness.

Well your points (which I agree) indicate it won't be much of a drag.  The losses at PSU and DC regulator are perfectly linear, if efficiency is doubled then at the chip then (excluding controller and cooling) is going to be doubled at the wall as well. Also 55nm has 1/4 not 1/2 the transistor density of 28nm (55/28)^2 ~= 4.   While a 400% improvement in efficiency is unlikely 200% maybe 250% wouldn't be that unusual.  0.9/2.5 = 0.36 J/GH.   The controller is negligible (5W maybe 10W) so A lot will depend on what wattage is needed for cooling.  Still lets look at a Cointerra rig you are looking at two 8W pumps and six likely 6W fans plus maybe 10W for the host so that is ~62W which is only a 4% overhead.  So yeah that 60W or so isn't going to be reduced but it is a minor drag.

Of course who knows if Bitfury can pull it off, or be able to take the time needed given the more competitive environment, to do a hand placed 28nm chip.  Still it does make me lower my own upper bound on what efficiency I think is plausible for 28nm.   I think 0.4 J/GH (at the wall) might be possible compared to ~0.8 J/GH for all existing specs and rigs.

The 'at the wall' power includes not only the chips'  power consumption...  but also the Controller's power (which isn't going to change much between now and next 6 months)... and includes power supply & dc/dc converter inefficiencies... which isn't going to change at all and hasn't changed for years... so even if the chips consume HALF the power of their previous generation... the rest of the system is still going to be a drag to halving the 'at the wall' power.

Well 0.6 J/GH at the chip isn't <0.5 J/GH at the wall.  I honestly could give two craps about the at chip efficiency unless the company selling them is willing to a) pay for all the non-chip power use or b) I was designing my own miner and buying raw chips by the reel.  "Full custom" has no standard definition in the industry anymore.  So nobody can say their chip isn't "full custom" and it also means absolutely nothing.  

D&T... since both hashfast, cointerra, and a bunch of others are claiming 0.6W/GH/s for non-custom 28nm.. i see no reason why a full custom bitfury can't do better than that... but when will it come?  and will any one else supersede it?  bitmine also claims full custom (but i doubt it as they did that really quickly!)

As long as I never have to use chainminer, I would be on board.

Icebreaker.. you keep trying to throw shit at cointerra without looking closer to home.   both hashfast AND cointerra raised private investment at around the same time.  Any accusation you can level at cointerra you have exactly the same situation with hashfast.  the shit you're trying to pull doesn't work because it applies equally to your strange bedfellow.

And as for the IceDrill deal, that wasn't an investment... that was a bulk purchase.  they're customers of hashfast, like the rest of us.  Do you think they're happy customers... why not ask them?

theres no nonsense going on... and theres no tears.   both hashfast AND cointerra are doing pretty well.   Not as well as KnC or BFL, but still all of those guys are doing pretty well.  Ironically the one who isn't, right now, is AsicMiner, who had a brilliant start and has failed to follow through.

And don't start saying that $1.5m isn't a large investment.  it was enough... and it was more than some and less than others.   Do you know, by any chance, how much your team raised?   I'm curious to hear your view.

Impressive.  It makes me think a 28nm die shrink of Bitfury (Bitfury28) could be <0.5 J/GH at the wall (with high efficiency PSU).

BFSB sold them as 400 GH/s full kits, but nearly all of them make 500+ GH/s. My boards are unmodified, I will play around with that when they brake even equated the loss I made from the other actual ASIC all-stars.

All Hail Sir Jez of the Holy Order of White Knights!   

I am fully aware Cointerra sold out their first retail batch (the Coindesk article isn't exactly Top Secret  ).

I am also fully aware of the ASIC business investment climate at the time CT launched.

I was first in line to buy a massive stake of IceDrill shares, and almost bought an equal amount of CryptX.

Luckily, I fell asleep before trading began and thought better of it the next day.

I'm certain I dodged a bullet, because Cointerra has been a comedy of errors since the days of their megalomaniacal scheme to control 20% or whatever of total hashrate.

That nonsense ended in tears, as will their latest hubristic 'Whatever HashFast Does, Plus One' play for cash.

If you know of other (nearly) large(-ish) Cointerra investors besides CryptX, please name them and I'll admit being less than completely right.

Otherwise my assertions stand unchallenged, and I'm being completely fair.


PS  Sir pankkake, $1.5 million is hardly a "large" investment.  Anything under $2 million is small or medium when 28nm NRE costs ~$5 million.

i think thats a little unfair (and not surprising really, coming from you).  both hashfast and cointerra have sold out of their first two batches, so thats actually a pretty high amount of pre-orders in both cases.  we're talking millions of dollars here for each batch...  these aren't fly by night companies with zero sales.  and i think the sales of cointerra are probably more diversified (read decentralised) as their sales are mostly retail sales.

-- jez