Topic: CoinTerra announces its first ASIC - Hash-Rate greater than 500 GH/s - page 84. (Read 231002 times)

I'm curious... I've heard people say that the KnC chips are just fpga copies.. but I've not heard it from any legitimate nor direct sources.  Do you have a source for that info?   Its often said by hashfast as one of the FUDs that they spread, but i really haven't seen it directly mentioned anywhere.

And anyway, if they've got them running at less than 1 Watt per GH, at the wall.. thats pretty acceptable to me.  Its not as low power as hf or ct, but then, they're nit shipping yet... so the difference is moot.

Yes, your failure to understand the difference between raw performance and underlying technology is hilarious.  Derp!

Who said anything about "lousy?"  Not I; my KnC group buys are mining me plenty of Bitcoins, so don't worry your silly little head about that! 

OTOH, the chips in those KnC machines are about 1/3 as efficient as HashFast's on a GH/sqmm basis.  Because KnC chips are FPGA copies.

I know math is hard for you and you don't understand what an "FPGA copy" is, much less why HashFast's chip is more advanced.

Just forget about all the confusing things going on under the hood and focus solely on the output, like a good consumer.  Leave the icky details to us hardware nerds and you'll do fine!

For those interested:

We are buying Batch 1 (December) and Batch 2 (January) Cointerra pre-orders.
https://bitcointalksearch.org/topic/wtb-terraminer-pre-orders-334606

could be true..KNC says it will mine up to I think no more then 5% of their units sold.....but with how nuts it all is in the asic world...if they go thru with what they say and don't do a new product till march ....that would be in line with that...(unless others continue to ramp up production then they will do the same) But on a side note as a legit business lets say trying to make the next gen ASIC miner...what would you do ...with difficulty flying up ...everyone and his brother gonna pop stuff out from dec thru march...?

imho cointerra's batch 3 unit for 2th at 6k is over priced considering that it is competing against batch 1 and batch 2 increase of bandwith from its own products alone not to mention others...so if I was in kNC's shoes and really did want to make another next gen asic unit in march say...what would you guess to shoot for in TH and price?  so say your guess then is to shoot for a 4th unit at the same price as cointerra cause you are a month behind? would that be enough? this would drive me nuts!

imho KNC would probably from a business sense be much better off just keeping the cash they have made probably what is that about 20 million....?  I mean even if they want to make equip the cost vs risk of guessing this far ahead and getting it out in the timeline needed is daunting imho

on the other hand this article says that KNC made $3 million in just 4 days with BTC rise etc http://www.wired.com/wiredenterprise/2013/11/400-bitcoins/

so anyway ....thats the dilemma  and as an example of forward thinking being wrong I was told by Josh at BFL in chat that no way KNC would get a unit out before Jan 2014 in July....well BFL imho really missed the boat on what they thought difficulty was gonna be with expected 600gh units selling at best in Jan 2014 (not counting the limited bullet run they may get out in dec)..... so they took that bet it looks like and KNC crushed it out in time ..thus imho killing the 600gh monarch card before it got out of the gate....you need a 6 month timeline at best to get something out the door but the price/speed has to work when they hit the street...man that is gonna be rough guessing for the next 1/2 year at least imho

man such guessing on what to put out and if it was technically possible to do so in timeline with difficulty as an asic mnfg would drive me to drink (again assuming they legitimately want to make product that returns ROI for customers to return and make more product ...which is what cointerra and KNC say they wish)

anyway above my pay grade...myself I understand ..amd probably gonna sit this out till march 2014 myself..got a KNC Jupiter 550gh unit (was supposed to be 400gh pleasant surprise) and unit that made ROI in 26days..but even with them shipping my unit on oct 15th on their timeline my ROI on the unit was due more to BTC rise and luck then anything else.......too razor thin on shipping these things out in time ..and when they ship ALL the units in say 3 weeks like KNC did and they all go kerplunk on the network at the same time from the company and a large splash in difficulty ensues..that further compresses the customers available timeline to make ROI at least

whatever will watch from the sidelines..but last point if I do buy from a asic mnfg in the future it will be by credit card so I have the option to 'bail' if they don't follow thru on their timeline ....not going thru bfl hell again...so far that is only KNC ...perhaps cointerra will change its mind again on paypal or cc orders

so it goes will be interesting to watch

Searing

my guess is those asic manufacturers that mine themselves will make sure not to oversupply so that difficulty won't hurt its themselves and its customers as well...

Yeah the sign said "Will Work For No One, I'm Rich, Bitch"

Didn't I see you standing at the red light, holding a sign?

ditto Jupiter Oct 18th received order.....now at 1300 bucks over cost of Jupiter at 550gh for $7131.80..not bad for 25 days of mining

no complaints..

the problem with all these ASIC mnfg types is you either are at the head of the line or don't make ROI

I've no idea if/when I'll buy another miner but it will probably be KNC in that they take CC and paypal w/cc so if I have issues with the ASIC mnfg company getting the item out the door and keeping my $$$ (ie we take all the risk as miners) I can call the credit card company and pull the plug...no more BFL games

wish more options but after being screwed by BFL ....not gonna pay up front on the promise of delivery

anyway my 2c worth on all ASIC mnfg ...they are playing it too close to the ROI point or not by barely getting these units out in time or not ....w/o recent price increase of BTC even some KNC units of the 1st batch may have been iffy on getting ROI (sometimes better to be dumb and lucky )

and admittedly w/o the recent BTC increases I could of been screwed from an ROI viewpoint...so yeah something has to change in all this then just tossing out cash to the asic mnfg of choice and crossing fingers..they will get the beast out to you on time and/or not getting trampled by the TH they are tossing out as a whole can be a fun nail biting time

anyway CC or no play is probably a good start if you are thinking of getting a miner .....always best to have an exit strategy if they don't deliver

my 2c worth

Searing

Hilarious!~

I already mined 55 BTC with those lousy KNC chips...  how many you mine with HashFast's superior chips?  How many will you EVER mine?

derp

Hi aerobatic,

Most of your above post is completely reasonable.

I disagree that KnC's chip is comparable to HashFast's.  KnC is an FPGA copy; HF's is not (hence the 3X GH/sqmm diff).

You need to explain how my sincere effort to answer VE's pertinent question constitutes "trolling."

If my answer was off-topic, content-free "trolling" why did you produce such voluminous, technically perspicuous responses?

I don't see how mere "trolling" could possibly engender such a wide ranging, thorough hashing (pardon the pun) out of various intricate ASIC production issues.

Why do you need to denigrate me for holding opinions contrary to your own?  Lack of confidence perhaps?  White Knight Syndrome?

I considered being a Cointerra customer/investor, then thought better of it.  Sharing the results of my due diligence isn't "trolling" no matter how much you object.

I'm far from being the only one who pegged them as a whiteboard/Powerpoint driven marketing company, promising great things before those things were actually designed.

Also, by virtue of being a HashFast customer, I can hardly fail to understand that POV!   

I'm not denying they did it real fast. they're a damn good team and they were extremely efficient and aggressive.  but to say record breaking implies they did it faster than knc, and i don't think thats the case.  i think knc is definitely the record holder and neither hashfast (nor cointerra) will break that record!


and you, equally don't know that they won't deliver in the same month.  but right now, their announced dates, are within a couple of weeks of each other!


hashfast has yet to say if and how they're going to 'make it right'.  what will they do?  refund everyone more than half of what they paid... to bring it into line with a december delivery $/gh price ?  e.g. to have charged the same as what cointerra charged per $/gh for december orders?  

and i agree that botching a tape-out is very annoying.  but if you can pay extra to everyone in the rest of the timeline to bring it back onto schedule and deliver when customers are expecting it, then no harm done!   Customers just care that they will get what they ordered, in a timely fashion.  they don't care what the company has to do to get there.


they're both engineering driven companies.  i don't know a single marketing person inside cointerra.  they don't even employ one, as far as i know.  they're just engineers, and cust support.  not a single marketing person.  they don't even employ someone of john (of hf)'s talents... though i think they should!


you have no evidence to the contrary either.  you don't KNOW cointerra's chip is more than 60 days away and you don't KNOW that its not 1 Ghz either!?   you're just trolling.   And yes, if you want to bet on it, i am confident enough on both of those facts that i would happily take your bet of both those statements exactly as you stated.   that cointerra WILL have chips within 60 days of tapeout, and that it WILL clock at 1 Ghz or over.

But i think you're missing a big point.  the customers don't care whether any chip is a particular clockspeed.  they care about delivery date, performance, price and power consumption...  hashfast's chip doesn't run at 1 ghz (i think it barely runs at half that speed... and it doesnt need to).   just as long as it delivers more than 400 GH, thats what everyone cares about.


So you're saying that hashfast's team for their 2nd chip will be identical to the 1st chip are you?  Do you think thats likely?  Is it even relevant?  i mean, any change in architecture between chip 1 and chip 2 will probably require a different team with different skills.   i think that in almost every asic company's case, the team makeup between different generations of chips will be slightly, and possibly even wildly different depending on what they want to do different next time around.  You can be pretty sure that whoever contributed to the delay in tape-out at cointerra probably won't be flavour of the month next time around.


no one cares whether gf is a better company than tsmc.  we all know tsmc is better.   its 4 times bigger, for a start.   but gf isn't a 'botch job' company either.  it has more than $4 billion in revenues and is the number 2 silicon company...  and they're both fully capable of churning out bucket loads of 28nm silicon, and if anything, making a deal with the number 2 company may get you more tlc than a deal with the number 1 company... as the number 2 is hungrier for the business and probably more willing to accommodate special requests (like expediting silicon production when its needed!)


Now you're just trolling again.  none of those companies are amateurs.  What a dumbass thing to say.   And do you think it makes hashfast and uniquify look like saints that they dropped the ball on the substrate thing?   no!  everyone makes mistakes.  If you measure companies just by their mistakes, they all look like amateurs.  The final proof... Will cointerras systems be on time, deliver the performance, and consume the power that they said?...   and we shall see.   And in January, when certainly both hashfast and cointerra will have systems in the wild, the 'customers' will be able to see whether they both delivered what they promised and whether they received 'good value'.  hopefully they both will!  everyone's a winner.  You don't need one to fail to make the other one good.  Both companies are good.. both companies are making good products and as far as I'm aware, both companies have sold out of the first two or more batches, so they both must have plenty of customers as well!   All they've got to do now is keep them happy!


Hashfast started with a two month lead alright.  And lost it.  Yes, cointerra is a month behind.  But hashfast is now more than a month behind.. nearly two (promised date was late oct).   But cointerra had the benefit of telling people a delivery date that had some slack in it, whereas hashfast was very optimistic.  They had a HUUUUGE lead, now disappeared.  bet cointerra are kicking themselves cos they could've come out ahead if they hadn't also had a delay!


yet again, you mis-understand the customer.   they don't care whose chips do what.  they care about systems and performance.. and price.. and delivery dates.  knc doesn't sell chips...  they sell systems.  and their system is a 550 GH today (and tomorrow, i believe its going to be faster.. 700?  750?).   hashfast's equivalently priced system is 400 GH today (and maybe 500 when oc'd).   price per GH - for the system - is what matters most, and availability.. and is regardless of how many dies or chips or whatever.  its a system to system comparison that customers most care about.   And there are other competitors too with completely different architectures like Bitmine, AMT, and Black Arrow... all of whom will now be in the marketplace in a similar-ish timeframe to hashfast and cointerra.


You, Icebreaker, are just a troll.  You'll say ANYTHING, absolutely ANYTHING... to try and knock a competitor.   Well, I'm trying to be pragmatic.  I will defend cointerra against your trolling (and i also defend hashfast against other people's trolling.. and knc too).  I try to stick up for whoever is being unfairly maligned.  just trying to keep it real.  you're just a troll.

HashFast went from logical design, to physical design, to complete tape-out in record-breaking time.  That speaks to their engineering and managerial prowess.


You don't know that cointerra will deliver "in the same month" as HashFast.  Stop assuming facts not in evidence!

Of course I know about hashfast's delay.  It's funny when you dress up uncontested, widely-reported facts as some kind of argument clincher.

Sure, hashfast dropped the ball on the substrates.  That sucks but they are going to make it right.  Botching a tape-out is a far worse unforced error.

I'd rather be a day-one customer of an engineering-focused company like hashfast than a marketing-driven one like cointerra.


Again you assume facts not in evidence.  Cointerra's chip is 60 days away, not running at 1 GHz.

We know from their PR Hashfast is very far along with their 2nd chip.  But unlike Cointerra, they didn't fire their incompetent original design team and bring in new guys to design it.


Ask AMD why they decided GF's 28nm process isn't good enough for them, but TSMC's is (GF invested in the wrong 28nm tech years ago and TSMC didn't).

Just like Cointerra and Open Silicon, Global Foundries are amateurs compared to the 28nm pros at TSMC.


The substrate delay is a one time thing, completely unrelated to TSMC's unparalleled ability to pump out 28nm wafers by the truckload.

Cointerra and HashFast are both a month behind, but HashFast started with a two month lead. 



You don't know WTF you're talking about.  HashFast's chips are three times more productive per sq. mm than KnC's.

VolcanicErupter asked a very good question on this thread, and I gave him the best answer I could.  Stop trying to control what other people say/do/think.

It's funny when you decry my post as unworthy, then write a giant wall of text strenuously attempting to debunk it.   

didn't everyone else do that too including your 'home team' ?  hashfast announced their 400 GH chip specs at least a month before they taped out.  How is that any different ?


What planet are you on?  you're such a fanboy of hashfast you didn't notice they're more than a month late, even after they've got wafers back?   cointerra may have a month delayed tapeout, but they had built in some margin into their production and they say they can still meet their promised delivery schedule.  Possibly they can do this because they didn't promise an overly aggressive date, and just said late december, which they are still targetting.  They didn't over promise an early date then disappoint everyone but instead they gave a date that they believed they had a good chance of hitting.   hashfast now is in the unfortunate position of having sold their systems at october prices and will now deliver in december and beyond, making all their customers unhappy and feeling like they overpaid, whereas cointerra sold their systems significantly cheaper at december and january prices, and they're planning to deliver in december and january.  And if they deliver to those dates, their customers will get what they were promised, at prices that represent the extra time.   Its ironic that hashfast and cointerra will now deliver in the same month, yet they charged very different prices to customers.

Really?  designing the 2nd generation chip tells you they're not happy with their first one?   A chip that outperforms, and is lower power than any other 28nm chip?  and you think that makes them unhappy?  Can you honestly, hand on heart, tell us that your own favoured hashfast team is not already thinking about their next generation chip?  Do you think any asic companies would be so dumb as to not have started designing the 2nd chip the second the first one has been put to bed.   And if they could afford two teams, they'd be doing them in parallel (like Intel does with Tick Tock architecture!)

Absolutely everyone who has taped out their chips, is already working on the next ones.   knc already announced theirs is in the works.   avalon's already announced the next three chips.  bitfury's on the next chip etc.  Good asic design takes time, and any asic company with any sense (and they are all very smart cookies) would most definitely be architecting their next chip the second they can so that in 6+ months' time, they've got a new product to offer their customers.  And I'm quite sure they have all learned a lot from their first chips so their second ones will be wayyy better.   And you know as well as i do that in the bitcoin world, hardware gets obsolete real fast... so you need to have a follow up in development ready for when you need it.


They've paid expedite fees to get their chips delivered as fast as possible.  from tape-out to end of dec is less than 60 days so we'll see if they get them before end of dec.  Only time will tell but i think you're wrong that it will take more than 60 days!


had hashfast been on time, i think that might've been true... but now that they're nearly two months late, in reality hashfast's multi petahash batches may well land in the same month as cointerras so i think the entire bitcoin mining community may well benefit from hashfast's delay (especially knc's customers who are probably extremely ecstatic right now... but also even cointerra's customers, who will now have much less network hash rate to contend with when their kit arrives, than it was originally going to be!)

Just having wafers back, without substrates is a pretty unfortunate position to be in as I'm sure you realise that wafers are useless without chips to put the dies in.  announcing an unexpected and significant customer delivery delay was pretty shocking for all concerned.. but do you think you can with a straight face accuse cointerra of a mere four week delay (and so far at least, unaffected delivery dates) when your own professed team has similar but much bigger problems (and very unhappy customers)

Delays during the tape-out process aren't unexpected... its an uncertain time and has to be taken into account during project planning... but delays AFTER your chips come back is a more surprising (and disastrous) outcome.  i expected to be mining with my hashfast systems by now.  instead, I'm literally not sure who will deliver sooner out of hashfast and cointerra, and knc is about to announce their new and improved systems in the next few hours!  and my existing knc gear has been hashing away with much less network hashrate than anyone could've expected this month.

you really do have rose-tinted glasses on don't you!?

All three of KNC, Hashfast and Cointerra are making relatively similar products with quite similar ish performance... differentiated only by price, delivery date and performance.  So far KnC has delivered on time.  Hashfast has announced substantial delays.  And Cointerra says they're still on time despite a four week tape-out slip.    Does it make you feel better about your own choice of miner's delay to try and bait another mining company  customers instead?  Perhaps you're better off placating the angry mob on your own forum before they try and storm the castle than spend your time worrying if cointerra will deliver on time.

Your greater concerns are

1. Cointerra advertised physical performance (GH/W/s) before completing physical simulation.  They sold you a dream, not an actual design.
2. Cointerra took a month longer than scheduled to complete tape-out, telling us the mock tape-out was a disaster.
3. Cointerra is hiring a new team for their next chip, telling us they aren't happy with the current one (ASIC All-Stars? LOL!).
4. Cointerra's chips will take at least 60 days to fabricate, assuming Global Foundries' inferior 28nm tech doesn't botch the job.
5. HashFast already has their first batch of wafers and will have lots 2/3 dumping multiple petahashes on the network by the time batch one Cointerras ship.

Those signatures are really ugly.
Who designed them, John?

John, you know that there is an hexdec color code that is usually used on websites, right?

please elaborate .. what are our greater concerns?

I dont want to trivialize water cooling too much (particularly having no experience with it), but at this point, I think the cooling solution is probably among the least of CT customers concerns.

Im worried with these single - four chip miners... when chip burns it´s gone.

It was 90 days at one time. Wonder why they have lowered it to 30?

I've worked with the guys over at Cool IT Systems. They are a good group of people and engineers and I've used their solutions in my personal computer for low acoustic CPU cooling. When I was working with them, we had put together a PICMG 1.3 backplane and SHB into a desktop chassis running 6 GPUs. We had run a custom cooling and manifold system from 3 radiators on the chassis to cool the 2 Xeon CPUs and 6 GeForce GPUs. All this we managed to put together in a week with some help doing custom fabrication on the chassis. This would have been back in 2007/2008.

At that time, they were doing an early prototype cooling system for rackmount computing. They had essentially done the redneck setup by mounting a radiator from a Honda Accord onto the rear panel of the cabinet with a couple large fans. Was cool to see. Actually looked pretty neat too.

There may be some different staff there now as one of the guys I know has moved on, but in general, I have found them to be a very good group of people to work with and they know their stuff. They have done a lot of OEM work for other PC companies, and I believe they currently license their pre-plumbed CPU coolers under the Corsair brand.