Topic: CoinTerra announces its first ASIC - Hash-Rate greater than 500 GH/s - page 92. (Read 231002 times)

That is very good to hear and I hope that you succeed.

However, this brings me to my first question.

I graduated with honors in mathematics, but there is one thing that I just cannot figure out. What am I missing?

Shipping/Delivery for late December (last day of 2013) = $14K per unit (2 TH/s)
Shipping/Delivery for early January (second day 2014) = $6K  per unit (2 TH/s)

But you can be up to 30 days late without penalty, which brings me to my second question.

Is your pricing for late December (2013) and early January (2014) now the same, and if not why?

We are very much on track for our shipping schedule, production time involves a lot of factors outside of just tape-out and we have our supply chain set up to minimize lead time.

Our partners are all lined up to produce miners at peak capacity and all components for the TerraMiner series will be ready for assembly at high capacity with minimal build time for our shipments. 

I don't think you understand how this whole mining thing works

Sorry, it's just funny to see someone who consistently flames on bitcointalk to infer this.  You aren't in your Mom's basement, then, ICEBREAKER?  That ignore is getting nearly too dark to read.

OMG you don't think difficulty will be 13,744,638,000,000 by next december?  OPTIMIST!

that makes absolutely no sense.. What do you work for them? Don't insult our intelligence

Depends on how broadly or narrowly you interpret the phrase "best team for designing Bitcoin mining hardware." 

Remember, we were talking about the "best team" not the "best design."

Broadly defined, the "best team" would include people responsible for financing the design process.

Narrowly defined, the "best team" would only refer to the chip architects.

The broader interpretation is more reasonable and thus superior, because it doesn't make sense to refer to three broke but brilliant RTL coders sitting in Mom's basement with no funding as the "best team."

Part of (actually the main constraint to, because SHA2 is trivial) being the "best team" is having people on the team who can get the NRE for tape-out and finance wafer production.

For what purpose?

Those two statements are mutually exclusive. You can have the best design in the world, but not have to necessary funding to complete the tape-out.

Nope, I am not joking. I am very used to high quality hardware and I see no reason to lower my standards, but pricing of the hardware will be a factor. I plan on running the hardware for years.

You're joking, right?  Would you also rather that KnC shipped a couple of months later, if you were a KnC customer now?

I never said SHA-2 algorithms are difficult to write. 

I did say Cointerra has ZERO EXPERIENCE designing a new chip from scratch and that will cause them problems.

SHA-2 is the easy part.  Managing the logic/power/heat of a giant chip with dozens of SHA2 cores is the hard part.


You have arrived at a contradiction:


Both of these statements cannot be true; one excludes the other.

You need to check your premises because one of them is demonstrably false.  Hint: it's not Number One.   

Now that Cointerra has botched their mock tape-out, it's time to stop assuming they are the ASIC Dream Team implied by their slick marketing and website.

Logic demands we judge Cointerra based on their demonstrated incompetence, not their fancy hagiographies/pedigrees/titles of nobility/etc.

Actually, SHA-2 algorithms are not that difficult to write for hardware, they are very FPGA/ASIC friendly. Any graduate student in Electrical/Electronics Engineering could whip one out in about 3 months or so, I did so back in 2009. The problem that you are going to run into is designing/optimizing for maximum efficiency, and doing so requires additional time and effort. And I believe that Cointerra is 10 times more qualified to do so than I am.

However, I do agree with you that tape-out was expected in September and that did not happen. Now, tape-out might happen towards the end of October, and if it does, there is no way the December delivery is going to happen. Cointerra is going to have about a 65 day window to complete everything and have the hardware at your door step. Not to mention that Quality Assurance (QA) by itself takes a long time for the type of  hardware quality that they are aiming for. I would rather have them to be late than rush and lower their standards such as KnCMiner did. No, no, no, I would rather they delay by a couple of months and deliver high quality products.

The problem I see with Cointerra is not with their Engineering skills, they have the best team for designing Bitcoin mining hardware - their problem is KnCMiner. Why, because a lot of people would rather buy a higher priced miner today, than a lower priced one 3 months from now. They also want the ability to use Credit Cards, PayPal, and have refund options. KnCMiner has all of that going for them because the got in the market first and there was no real competition.

Cointerra started late and the pre-order well is running dry, the other manufactures who got in first drained it. They have to balance funding for research and development, and potential customer refund requests. So how do you deal with that without outright refusing refunds such as what Butterfly Labs and HashFast are doing? I do not know, but I believe this is their major problem. Refuse refunds, and you alienate a big pool of potential customers.

Please Conterra, just come out and say that you are going to be late for December delivery because you cannot rush perfection. Your team will gain more respect and bigger wallets down the road.

problem is fiat vs. BTC. never pay a ASIC producer with BTC. pay only with fiat. if you obey this rule then you will have a large problem if he delivers in time and you did bet with BTC against this.

still waiting patiently for it.

Just do the math. Take hashfast specs; 324mm² 260W = 0.8W per mm².
If you look at that intel cpu, its a ~80mm² die for lets say 100W TDP. You may think that equals 1.25W per mm², but in the hotspots its probably closer to 100x that.

im not so sure.  a bitcoin mining chip is mostly hashing units, and hashing units have a very high toggle rate (i.e., most of the gates change state almost every cycle) ... so i think that almost ALL of a hashing chip is pretty much the maximum power consumptive thing possible.

Actually, the cooling problem is far greater with CPU's. Overall power consumption per mm² of a cpu may look lower than for a bitcoin chip, but its not, because in a cpu the power consumption (and thus heat) is highly concentrated in a few tiny tiny parts of the cpu. For reference, here is an infrared picture of an Intel cpu:



Basically all the power is consumed in a just a few % of the die, less than 1mm². A bitcoin miner will not have such hotspots and be much more homogeneous.  That makes cooling a lot easier because silicon is not the best heat conductor. Sure you still have to get rid of those 100s of watts, but thats not rocket science when they are spread over a far great area, at least with current designs.

thats simply not exactly true now is it?   making out that the only difference is the power requirements is making light of a huge issue.  with great power requirements comes great heat and thus great cooling requirements, and great power supply requirements.

 the components on a pc board are to support the intel processor and in general, it runs at a very low wattage most of the time, and is very peaky.  i.e.: it only draws LOTS of power when doing something very taxing.. which is not very often....   whereas the components on a mining rig have to run at full pelt, all of the time, 24/7.  thats a very different demand were making of them.

An Intel chip has a TDP of 125 watts (thats the big ones) and yes you can over clock them band make them run in the hundreds of watts... like 200, maybe even 250 watts max.  The Intel chips never run at 300 watts or 400 watts during normal use.   Sure, there are over clocking crews that make them do that, as one-offs, using liquid nitrogen cooling... but for normal pc use, the intel chip never gets that hot and doesn't need extreme cooling....   whereas, on a mining rig, we're expecting our 'extremely powerful hashing chips' to use 'extremely high amounts of watts' and run 'extremely hot' all of the time... which is why the cooling system for a bitcoin mining rig must have extremely efficient cooling.

Power supply.  so, cointerras 2 TH box... will consume more power than any pc you can buy.   on a pc, it barely uses any power in the pc itself.. like a few hundred watts.   plug in a few top graphics cards and some hard drives and maybe it will consume 750 watts or 1000 at most.  its very rare to find a pc that needs more than 1000 watts... but ok, some of the top pc power supplies are 1200 watts.  you can even buy some that are 1500 watts..  but these aren't the norm...    whereas, we want our mining rigs to run really really fast, and even with extremely efficient power (cointerra claims 0.6W/GH at the chip).. we're not talking a 50 GH butterflylabs thing here... we're talking 2,000 gigahashes!!  that means 1200 watts... then add in the inefficiencies of the power conversion circuits, and the power supply, and the fans and cooling system, and the on board controller board (it has a network connection so there's a pc or single board computer in there) and you're already at over 1500 watts...   and you're already at about the max power that a US household circuit can muster...but then what if you want to over clock or over volt it...?   which probably explains why they're using two 1,000 watt power supplies (and those aren't cheap.. did you look on newegg to see how much those cost!  heck, even the seasonic ones that hashfast is using retail at 250 each and they've got two as well.. I'm sure they're not paying retail prices, but still, these are expensive items)

of course, i expect most of the 2TH machines will be hosted and not in people's homes, unless they live in cold places and want 1500 watts of heat coming out of the fan from a box, 24/7.

Anyway, my point being that yes, you can build high end bitcoin mining rigs out of pc components, but they're the top top top components that are maxed out in performance and they're not the cheap low cost end of the pc component space.

The components on a PC motherboard do the same things as the components on an ASIC board.  In fact, the ASIC board does even less.  The only difference is the power requirements, but you can split that up into multiple smaller pieces. "500-1000 times faster at hashing" doesn't mean 500-1000 times more actual electricity.