Swedish ASIC miner company kncminer.com

[Kuroth]

Dryer Outlets??     

[Mota]

Can someone show me how to use my dryer outlet? I can't find a psu cable.

Is that a joke? If you are not an electrician, don't do that yourself. That is a high voltage output, not something for amateurs to work on...

Don't do it. It is in fact low voltage (till 1000V AC / 1500V DC) but DON'T DO IT. You'll risk your life if something goes wrong. Or you will do a short, etc..

Watch out for sharp knives in the kitchen also ...

My favourite questions asked in electrical engineering at university was by a lecturer:
"So how many of you have been electrocuted?"
(lots of hands went up)
"How many of you didn't survive?"

That is the most stupid question ever. Do you know how many people die in Germany alone because of electirc shocks?
Germany has a standard voltage of 230V, "Starkstrom" has up to 1000V and is used by a dryer outlet. 600V can effectively reduce the resistance of the skin and increase lethality by a lot. And most people who died did not even have that high voltage...

240V here in Aus

You do of course realise that the higher voltage actually means you are less likely to die from electrocution right?
Cos the current is thus lower.
It's the current that kills you, not the voltage.

... I also remember my older brother sticking a screw driver in a power point when we were very young ... he's now >50

Yes electricity can kill you, so can a chainsaw, driving on the road, holding your breath under water, catching an aeroplane ...

You really think bitpop is a complete idiot? (I don't) Coz that's what's implied by the "DON'T DO IT" silliness.

Sounds as silly as a Starbucks hot cup warning ...

... a hot cup does not kill... I got electrocuted too and lived, but that is not the point. the point is that it CAN go very wrong and it does in about 3% of the cases with lower voltage...
The lethality of an electric shock is dependent on several variables:

    Current. The higher the current, the more likely it is lethal. Since current is proportional to voltage when resistance is fixed (ohm's law), high voltage is an indirect risk for producing higher currents.
    Duration. The longer the duration, the more likely it is lethal—safety switches may limit time of current flow
    Pathway. If current flows through the heart muscle, it is more likely to be lethal.
    High voltage (over about 600 volts). In addition to greater current flow, high voltage may cause dielectric breakdown at the skin, thus lowering skin resistance and allowing further increased current flow.
and that is from wikipedia, for your reading pleasure regarding higher voltage -> lower current.
Stromweg    Accidents whole     Acc. deadly    rel. Verteilung(insgesamt)    rel. Verteilung(tödlich)    Lethality
Hand-Hand    2891       82                             77,3 %                        48,5%                            2 ,84 %
Hand-Fuß             349           19                              9,2 %                         11,2 %                      5,44 %
Hand-Füße    294               18                            7,7 %                         10,7 %                      6,12 %
Hände-Füße    106               20                           2,8 %                            11,8 %                     18,67 %
Verkürzte Stromwege Oberkörper
(wie Hand-Brust, oder  Brust-Rücken)
                    108             30                             3,0 %                        17,8 %                       27,78 %
Insgesamt    3748    169                            100 %                          100 %                         4,51 %

I am too lazy to translate all of it, but you should get the gist...  DON'T try to play with currents if you have no practical knowledge of the matter...

[jmaccoin]

Dryer outlets in North America are just two 15 amp circuits with one neutral. 180 degrees out of phase so you can still run 120 volt items with the same wiring as if you were running them 240v. Perfect for running 4 jupiters.

[Vycid]

If you have ever seen a GPU die, then you know the shaders constitute the bulk of the chip.  > 80% of it.

Similar logic exists with ASICs, and based upon die shots of Avalon, BFL, and Bitfury processors, I would put the non-SHA256 components at 5% or so in total area.


How do you think different product tiers in GPUs exist?  7830 / 7850 / 7870 etc.  The are ALL the same die, just the lower-end products have various shaders turned off in post-processing because there are defects in there.  Even with this post-processing, the simple laws of IC manufacturing apply - bigger dies = BAD for yields.

So what that shaders consist of 80% of the area?

They are still on the one single clock tree and the one single JTAG chain.

Please stop digging yourself deeper. You've made a methodological mistake that invalidates your results. Just admit it to yourself.

I have historical evidence here to back this up.  BFL has run into these problems as well (even if you hate BFL for their business practices, realize they didn't manufacture the chip themselves).

It's why you see the wide range in GH/s out of their single chips.  It is also why you will note that since they started selling their chips, they have different bins at different prices.  We don't know what their yield is, but they have a substantial number of lower-grade chips that cannot hit design targets due to chip defects.

Certainly BFL has encountered issues with killer defects on their chips. Know what they do? They disable the regions with defects and bin them. Suddenly there's not much correlation between yield and die size anymore.

(Intel does this also, and their chips are *much, much* more complex. Then again, if you ever had to deal with those fuckers, you'd understand how they get their defect concentrations down. "Our spec is zero defects, hurr hurr hurr... Your recipe creates 2 defects per wafer..."

Sorry, sorta got an axe to grind with those guys)

Anyway, here's to hoping that KNC is at least as competent as BFL.

[erk]

Certainly BFL has encountered issues with killer defects on their chips. Know what they do? They disable the regions with defects and bin them. Suddenly there's not much correlation between yield and die size anymore.

(Intel does this also, and their chips are *much, much* more complex. Then again, if you ever had to deal with those fuckers, you'd understand how they get their defect concentrations down. "Our spec is zero defects, hurr hurr hurr... Your recipe creates 2 defects per wafer..."

Sorry, sorta got an axe to grind with those guys)

Anyway, here's to hoping that KNC is at least as competent as BFL.

Except when you read the threads in this forum where people have changed firmware and activated all the engines on their BFL chips, the results have been really good, so much for that theory!

[Vycid]

Certainly BFL has encountered issues with killer defects on their chips. Know what they do? They disable the regions with defects and bin them. Suddenly there's not much correlation between yield and die size anymore.

(Intel does this also, and their chips are *much, much* more complex. Then again, if you ever had to deal with those fuckers, you'd understand how they get their defect concentrations down. "Our spec is zero defects, hurr hurr hurr... Your recipe creates 2 defects per wafer..."

Sorry, sorta got an axe to grind with those guys)

Anyway, here's to hoping that KNC is at least as competent as BFL.

Except when you read the threads in this forum where people have changed firmware and activated all the engines on their BFL chips, the results have been really good, so much for that theory!

I assure you defects are quite real. I looked at one today under a scanning electron microscope. I think it winked at me. Or maybe that was the astigmatism...

Clearly, BFL is just as incompetent at binning their chips as they are at making deadlines.

[erk]

I assure you defects are quite real. I looked at one today under a scanning electron microscope. I think it winked at me. Or maybe that was the astigmatism...

Clearly, BFL is just as incompetent at binning their chips as they are at making deadlines.

BFL don't make the chips, and they are only 65nm,  I can just imagine what you would find on a 28nm chip.

[Vycid]

I assure you defects are quite real. I looked at one today under a scanning electron microscope. I think it winked at me. Or maybe that was the astigmatism...

Clearly, BFL is just as incompetent at binning their chips as they are at making deadlines.

BFL don't make the chips, and they are only 65nm,  I can just imagine what you would find on a 28nm chip.

The foundry is being too aggressive with the chip binning, then. They're probably defect mapping by using a tool like one of these:

http://www.kla-tencor.com/front-end-defect-inspection/puma-family.html

Any die with a defect gets knocked down a bin, and the corresponding engine gets disabled. Two defects, another bin... and so on. That's going to be overly aggressive, since not all of those defects will be killer.

What BFL SHOULD be doing is burning the chips in, and binning them based on performance. Who knows why they're not.

Obviously, a 90nm defect in a 28nm process would be a lot more devastating than it would be in a 65nm chip, so a higher proportion of the defects might prove to be killer. I always thought 28nm was an odd choice... why not something a little more mature, like 45nm? They'd have saved millions on the mask set alone.

[erk]

What BFL SHOULD be doing is burning the chips in, and binning them based on performance. Who knows why they're not.

I don't understand the point you are making at all. With the exception of the MiniRig, BFL don't run their chips flat out, that's why speed bumps are available over the base unit. The MiniRig has been reported at being 480GH/s instead of 500GH/s but I don't know how many people experienced that, I have only read the one report.

[Vycid]

What BFL SHOULD be doing is burning the chips in, and binning them based on performance. Who knows why they're not.

I don't understand the point you are making at all. With the exception of the MiniRig, BFL don't run their chips flat out, that's why speed bumps are available over the base unit. The MiniRig has been reported at being 480GH/s instead of 500GH/s but I don't know how many people experienced that, I have only read the one report.

Running your chips below their maximum stable clock is a very different thing from completely disabling functional engines. One may be an optimal trade-off with power consumption and chip lifetime; the other simply wastes potential.

[erk]

What BFL SHOULD be doing is burning the chips in, and binning them based on performance. Who knows why they're not.

I don't understand the point you are making at all. With the exception of the MiniRig, BFL don't run their chips flat out, that's why speed bumps are available over the base unit. The MiniRig has been reported at being 480GH/s instead of 500GH/s but I don't know how many people experienced that, I have only read the one report.

Running your chips below their maximum stable clock is a very different thing from completely disabling functional engines. One may be an optimal trade-off with power consumption and chip lifetime; the other simply wastes potential.

No kidding, that's why people are having fun unlocking the potential with firmware upgrades!

[Vycid]

What BFL SHOULD be doing is burning the chips in, and binning them based on performance. Who knows why they're not.

I don't understand the point you are making at all. With the exception of the MiniRig, BFL don't run their chips flat out, that's why speed bumps are available over the base unit. The MiniRig has been reported at being 480GH/s instead of 500GH/s but I don't know how many people experienced that, I have only read the one report.

Running your chips below their maximum stable clock is a very different thing from completely disabling functional engines. One may be an optimal trade-off with power consumption and chip lifetime; the other simply wastes potential.

No kidding, that's why people are having fun unlocking the potential with firmware upgrades!

And why I'm baffled, since BFL has obviously not binned them correctly.

BFL has been selling their round of loose chips at different prices based on the bin (A/B/C/D). If they are being too aggressive with their binning, they are actually getting paid less than they could be. That seems very un-BFL-like. Curious.

[minternj]

Back to the topic of KNC. Did they ever say how many chips they have on order? Wondering how they are still selling miners on their website saying shipping in september.

[Vycid]

Back to the topic of KNC. Did they ever say how many chips they have on order? Wondering how they are still selling miners on their website saying shipping in september.

Possibly they ordered a chip surplus with the intention of selling loose chips a la Avalon/BFL if they couldn't make up the difference with miners.

[minternj]

Interesting for them to go balls to the wall with chip ordering. I guess they are pretty damn confident in the design huh? Avalon and bfl loose chips  sales were not in their initial orders. They were placed after they actualyl had the chips in hand. Wondering how KNC is planning for current and future orders since with a 28nm its not like you could call up the foundry and get more chips in a week or a month. Maybe they should update thier shipping date on the website for new orders?

[erk]

Back to the topic of KNC. Did they ever say how many chips they have on order? Wondering how they are still selling miners on their website saying shipping in september.

It says "Shipment begins in September" when I just looked, which doesn't mean if you place and order now you will get it shipped in September, but it certainly suggests you wont get it before September.

[Vycid]

Interesting for them to go balls to the wall with chip ordering. I guess they are pretty damn confident in the design huh? Avalon and bfl loose chips  sales were not in their initial orders. They were placed after they actualyl had the chips in hand. Wondering how KNC is planning for current and future orders since with a 28nm its not like you could call up the foundry and get more chips in a week or a month. Maybe they should update thier shipping date on the website for new orders?

That is an interesting point. The NRE costs are out of the way (primarily the mask set)... so why would they risk even more by having all their chips made at once?

Maybe they figure that if the chips fail, they're out of business anyway...

[minternj]

Back to the topic of KNC. Did they ever say how many chips they have on order? Wondering how they are still selling miners on their website saying shipping in september.

It says "Shipment begins in September" when I just looked, which doesn't mean if you place and order now you will get it shipped in September, but it certainly suggests you wont get it before September.

I guess but would be nice to have an idea of their initial capactiy, which is why i was asking if anyone knew what it was.  I'd think they wouldnt be placing more chip orders until they at least get one batch in. So shipping starts in sept for batch 1 ?? 2014 for batch 2? who knows.

[erk]

I guess but would be nice to have an idea of their initial capactiy, which is why i was asking if anyone knew what it was.  I'd think they wouldnt be placing more chip orders until they at least get one batch in. So shipping starts in sept for batch 1 ?? 2014 for batch 2? who knows.

When you place a pre-order for a product that has never shipped before, you take your chances! If KNCminer are smart they would have ordered enough chips to keep some stock, I am sure that's what BFL tried to do, but it all fell apart when they discovered they would need twice as many chips per unit to meet the performance they promised!

Because it take a couple of months for a chip order to be manufactured, you would want to order at least two months more chip stock as you had orders in hand for, but the gamble is always will the chips perform as simulated, and is the order rate consistent hence predictable?

When the chips come arrive in September, and KNCminer test their prototypes, we shall know more, until then everything is just speculation, and probably a little pointless.

[minternj]

Not actually pointless, my question was simple although I did interject some extra speculation.  Wondering when you would get a product if you ordered now? Only info i see for new orders is that shipping starts in sept like you pointed out. Do they have enough chips in the first order of chips from the foundry to cover any new orders, and if so up to how many units does it cover past the initial?