Pages:
Author

Topic: Cointerra AIRE Miner 16nm PreOrder - page 4. (Read 19889 times)

member
Activity: 98
Merit: 13
October 27, 2014, 04:17:23 PM
Entertaining to say the least-

Fact's-

Only devices using 16nm technologies are cell phones, using TINY amounts of voltage and amps, thus, low heat.

Intel's 14nm processor is suffering from massive overheating issues, they are now triyng to work in 3 dimensional space, as the 14nm / 16nm resistors generate so much heat that it melts the wafers.

No 16nm chip produces has had the type of heat running through it like these ASIC's will produce (unlike 20nm, use in processors, and GPU's)


Speculation-

ITS VERY BAD, that no processors, or GPU chips (high heat) have been made with this technology.  This chip will fail considering intel, with 2 years, and 50million dollars invested haven't been able to produce an efficient 14nm processor chip, I doubt these under qualified engineers have a chance in hell at doing anything that will work.  

Do NOT PRE-ORDER, it may take several months to a year before 16nm chips are ready for ASIC production.

Sources - ACTUAL ENGINEER, consulting with engineer friends who work on the 14nm intel chip production plant in Chandler, AZ....

clueless newbie - saying that you spoke to an actual engineer isn't quoting a source.  to do that you need to say exactly who you spoke to.

anyhow... thats ridiculous to say 16nm isn't ideal for bitcoin chips and only useful for mobile chips.  what a ridiculous thing to say.   the reality is that anything thats good for mobile chips, which are low voltage and low power, is exceptionally good for bitcoin mining chips as well since they too need to be low voltage and low power.  in fact, i think you'll find that bitcoin chips run at even lower voltages, on the whole, than mobile phone chips do..!

one thing I'm pretty sure of, is that in 2015, a large amount of bitcoin mining will be done on 16nm chips, and by 2016, probably most of it will be 16nm.  its just so much better than 28nm (and yes, even 20nm) that there's no reason not to use it for everything.

btw, nvidia has already announced they're using 16nm for their next graphics chips.

cointerra may have the first 16nm to be ready for tapeout, but i guarantee within the space of 6 months, a lot more 16nm are coming.  its the obvious next step for the majority of the asic companies that want the lowest possible power and smallest possible silicon area (= more dies per wafer).   sure, some smart companies can eek out a bit more juice out of 28nm and survive for one more generation in 28nm, but ultimately in 2015 they will all be using 16nm,.. cointerra's just the first, thats all.  they might be a quarter or two ahead of the pack, but the whole pack is going there!

(btw, 16nm tsmc and 14nm Samsung/global foundries are pretty much similar in specs, so for our purposes are interchangeable)


Lol you obviously don't understand hardware at all do you?  My point was, only mobile processors / applications are using 16nm technology currently.  Also, you're argument for Nvidia going to 16nm is flawed at best.  First off, they are over a year out, second, they recently switched to Maxwell technology, that allows them to use about 50% less power than older models.

My point is, NO ONE has produced an efficient 14nm or 16nm process (like CPU / GPU processor yet 5volts to 9volt power).  The reason is, the transistors are so close together on the wafer, it causes extensive heat, they literally melt the silicon between the transistors.  This is why Intel's 14nm tech has been in development for so long.  They are trying now to solve it by printing in 3 dimensions (more space between transistors for cooling) buts its going to be another 6 months to a year, and Intel isn't going to share there process for 2 more years.

Obviously you have no understanding or clue what is involved.  You say the ASICS can benefit from low voltage, while that is true, they can't achieve the same type of processing potential running on 2volts the same way they can on 5volts.  Its simply physics and I'm sorry if I'm talking above your head.  Am I saying its impossible, absolutely not, but they technology they are working on is cutting edge.  If you think that butterfly labs had issues with 20nm process, just wait for CoinTerra trying to do this on the 16nm level.

Come revisit this threat in March of next year and will see how spot on the hardware issues are as they miss deadline after deadline.

legendary
Activity: 1666
Merit: 1185
dogiecoin.com
October 20, 2014, 05:13:19 PM
This company's rating has been updated in the Manufacturer Trustworthiness thread.

[This message won't be monitored, discuss your concerns in the thread.]
legendary
Activity: 2212
Merit: 1001
October 16, 2014, 06:52:12 PM
All right,he is a decent guy & I wasn't trying to piss anyone off,just trying to get a point across  Roll Eyes

How's this  Huh

hero member
Activity: 857
Merit: 1000
Anger is a gift.
October 16, 2014, 03:09:47 PM


Better?
One too many 'are's. Does that make it arse?

Dammit... It makes me an arse.
-ck
legendary
Activity: 4088
Merit: 1631
Ruu \o/
October 16, 2014, 09:02:25 AM


Better?
One too many 'are's. Does that make it arse?
hero member
Activity: 857
Merit: 1000
Anger is a gift.
October 16, 2014, 08:49:45 AM


Better?

Edited because I cannot proof-read like an adult.
legendary
Activity: 2128
Merit: 1073
October 16, 2014, 07:27:46 AM

Could somebody please fix this meme? The text should be "Pre-order ASIC Prices Are Too Damn High!". Jimmy McMillan is too much of a gentleman to be dropping f-bombs.
hero member
Activity: 924
Merit: 1000
October 16, 2014, 04:41:26 AM
If they have all this money, and are such a great company, why do they need PREORDER money?
I bought a tesla X and he did not need me to pay for it.   I will pay for it when (and if) he delivers it.

It's really quite obvious if you've ever preordered hardware. They just want more money.

They know that hardware be selling from stock for ~$0.2/gh by time they are finally shipping so they would rather have people pay current rates instead.

Further can you really trust Cointerra given the last miner they put out on the market and the amount of complaints, refund issues and so and so on. I suggest steer clear of them on that alone.

Looks like they completely removed the entire forum today also.

Oh no.. that is a clear red flag there.
legendary
Activity: 2212
Merit: 1001
October 16, 2014, 04:33:05 AM
Do not fund these people.

Pre-ordering is trusting someone else with your money.

Don't do it.


Totally agree with Josh here.  We have seen enough scams/failed promises to know that pre-ordering can no longer be considered an option.
Avoid pre-orders!

full member
Activity: 148
Merit: 100
October 16, 2014, 04:25:26 AM
Do not fund these people.

Pre-ordering is trusting someone else with your money.

Don't do it.

Totally agree with Josh here.  We have seen enough scams/failed promises to know that pre-ordering can no longer be considered an option.
Avoid pre-orders!
legendary
Activity: 3878
Merit: 1193
October 09, 2014, 12:19:15 PM
If they have all this money, and are such a great company, why do they need PREORDER money?
I bought a tesla X and he did not need me to pay for it.   I will pay for it when (and if) he delivers it.

They do preorders so that all risk is on the customer. There's no reason to risk your own money when there are thousands of idiots out there that will risk their money instead.
hero member
Activity: 714
Merit: 500
October 09, 2014, 11:59:48 AM
I never said pre-ordering in general or from this company specifically is a good idea.

In fact most of the time you can buy cheaper hardware from in-hand vendors by the time your pre-order should arrive.
More than 4 months is obviously a very long time in Bitcoin land and nobody should lock in money for miners that long.

I was mainly commenting on the technical feasability aswell as giving some background information.
hero member
Activity: 770
Merit: 509
October 09, 2014, 11:34:23 AM
If they have all this money, and are such a great company, why do they need PREORDER money?
I bought a tesla X and he did not need me to pay for it.   I will pay for it when (and if) he delivers it.

It's really quite obvious if you've ever preordered hardware. They just want more money.

They know that hardware be selling from stock for ~$0.2/gh by time they are finally shipping so they would rather have people pay current rates instead.
legendary
Activity: 2478
Merit: 1020
Be A Digital Miner
October 09, 2014, 11:25:56 AM
If they have all this money, and are such a great company, why do they need PREORDER money?
I bought a tesla X and he did not need me to pay for it.   I will pay for it when (and if) he delivers it.
legendary
Activity: 2030
Merit: 1076
A humble Siberian miner
October 09, 2014, 11:16:15 AM
DO NOT BUY FROM COINTERRA! you have been warned!
Thanx, I will not. )
hero member
Activity: 702
Merit: 500
October 09, 2014, 06:46:51 AM
I was asking about the Terraminer when you replied about 0.23W/GH. Everything is clear now.


here's some assumptions...

lets assume that they've produced circa 20+ PH of terraminers.  lets say its 50:50 of the older 1.6TH and newer 2.0 TH boxes. 

cointerra's own network is circa 12 PH, according to organofcorti stats... is probably some old and mostly newer terra miners, since their network only started after the last customer pre-orders were delivered.

this needs to be tested by someone that actually has boxes...  but i would guess its in the 0.7-0.85 w/gh range- certainly not going to get as low as the newer generation miners from others that are in the 0.5-0.6 range.   suspect the newer 2.0 TH terraminer has more control over its voltages and could potentially use less power than the earlier model at the lower en of the range as well.

of course, a major factor is where the miners are and how much power costs at that location.   for instance, a power inefficient miner in a cheap power location (e.g. washington, iceland etc) could be more profitable than a more efficient miner in an expensive place (california..?  israel?  china?  take your pick).

in the short term, minimising power cost by moving miners to where they can operate cheaper, and thus, for longer, is the right thing to do... and in the long-term, its to use more power efficient miners, that have the margin to become even more power efficient when the need is there, so that they can be utilised for at least a year without having to change the miner.

-- Jez
hero member
Activity: 714
Merit: 500
October 09, 2014, 06:36:02 AM
You didn't understand my reply. I was trying to say that big money doesn't mean that they are safe from default or from fail. I wasn't implying that the are not making money.

Also I wanted to know what's their best efficiency for their already deployed MW. I think aerobatic replied to me about their 16nm miner, but I'm not interested in that. I wanted to know their best efficiency on their deployed miners with the GoldStrike1 ASIC chip.
Around 1W/GH, maybe 0.8. Any lower than that the voltage won´t allow the chip to work. (assuming the chip desing wasn´t changed)
It also means that you would only achieve around 10TH / rack on the standard customer design, but they either have access to cheap space or they have redesigned the enclosures for higher density.
legendary
Activity: 1904
Merit: 1007
October 09, 2014, 06:34:13 AM
They are building their 20MW facility (it is not finished / operational as far as I know).
They have added at least 1MW of capacity every month since April, and those first added PH were obviously the most profitable ones.
(Customers at 3$/GH have more than broken even twice, so you can see that Cointerra at internal cost of ~0.5$/GH has been shitting gold).

In August they had around 15MW of own / outsorced capacity, populated with miners.
Additionally, they sold cloudhashing contracts with a massive margin (and delivered around 5000 miners until June afaik).

I agree that Bitfury is more impressive, but Bitfury´s internal hardware cost on the 55nm chip is above or equal to 0.6$/GH, while cointerra should have been below 0.5$/GH from the start of their sales.

The claimed efficiency seems entirely within the realms of possibility, especially if you see that current spondoolies chips manage around 0.55W/GH on 28nm, and possibly even better with additional optimisation.
That doesn´t mean that they will achieve the claimed efficiency, especially with their track-record though.


You didn't understand my reply. I was trying to say that big money doesn't mean that they are safe from default or from fail. I wasn't implying that the are not making money.

Also I wanted to know what's their best efficiency for their already deployed MW. I think aerobatic replied to me about their 16nm miner, but I'm not interested in that. I wanted to know their best efficiency on their deployed miners with the GoldStrike1 ASIC chip.

the 0.23 W/GH is the 16nm 'aire-miner' they're offering for pre-order now.  its not their old 28nm based 'terraminer' !! no one has a 28nm based product that can achieve 0.23w/gh... at this time.

I was asking about the Terraminer when you replied about 0.23W/GH. Everything is clear now.
hero member
Activity: 702
Merit: 500
October 09, 2014, 06:33:13 AM
 

Fancy and complicated answer for a simple question. How low can TerraMiners go with efficiency was my question. From your answer I understand that they can go as low as 0.23W/GH which I don't believe it. Maybe you were referring to the chip efficiency, not the whole system, but even so I still find it hard to believe it. I want to see it!


the 0.23 W/GH is the 16nm 'aire-miner' they're offering for pre-order now.  its not their old 28nm based 'terraminer' !! no one has a 28nm based product that can achieve 0.23w/gh... at this time. and sure, we both know people are working on it and hoping to get there.  and it may happen in 2015...  but 16nm has a lot further it can go.  the 0.23w/gh that cointerra are claiming isn't the lowest power mode it can support.  its the most cost effective one (capital and operating costs are balanced over a 12 month period).  when it needs to be lower power, it could be well draw lower than 0.2 w/gh when running in low power mode (at lower volts).  you can't test that til silicon comes back, as its not easy to simulate at low volts.

i actually have no idea how low a terraminer can go.  suspects its not a long way under 1w/GH but ya never know!  perhaps someone that has one would like to put it on a killawatt  on it and select the lowest power stepping mode and give it a try for us (mine are hosted and i don't have access to them).   there's two models of terraminer.. the 1.6 TH one that shipped for the first few months and the 2.0 TH (that shipped in the last few months).  the latter one is more power efficient, as the extra performance was achieved without more power and just by using more efficient dc/dc converters.
hero member
Activity: 714
Merit: 500
October 09, 2014, 06:28:49 AM
They are building their 20MW facility (it is not finished / operational as far as I know).
They have added at least 1MW of capacity every month since April, and those first added PH were obviously the most profitable ones.
(Customers at 3$/GH have more than broken even twice, so you can see that Cointerra at internal cost of ~0.5$/GH has been shitting gold).

In August they had around 15MW of own / outsorced capacity, populated with miners.
Additionally, they sold cloudhashing contracts with a massive margin (and delivered around 5000 miners until June afaik).

I agree that Bitfury is more impressive, but Bitfury´s internal hardware cost on the 55nm chip is above or equal to 0.6$/GH, while cointerra should have been below 0.5$/GH from the start of their sales.

The claimed efficiency seems entirely within the realms of possibility, especially if you see that current spondoolies/bitmain chips manage around 0.55W/GH on 28nm, and possibly even better with additional optimisation.
That doesn´t mean that they will achieve the claimed efficiency, especially if their track-record is anything to go by.

However, they are currently the only ones that offer next-gen hardware for pre-order, and I am sure they will get some people / bigger investors to spend a lot.
Pages:
Jump to: