Topic: Swedish ASIC miner company kncminer.com - page 1960. (Read 3049501 times)

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.

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.

I cannot believe no one else has picked up on this as well.  KNCMiner's die size is 2.5 INCHES squared.  That's . . . . Titanic sized.

I hate do to be debbie downer here, but that means that on a standard 300mm wafer they will not get many chips, and that's assuming a 100% chip yield from the wafer.  Wafer's always have defects which cause some chips to go down.  The yield decreases exponentially as a the die size goes up (i.e. it only takes a tiny imperfection in the wafer to make the entire chip a paperweight).

Even if they get chips, the yield is going to be absurdly low, I would bet less than 25%.  They can probably recoup some investment by selling chips that are spec'd lower and have some "cores" disabled in them, but this is hit or miss depending on where in the chip the defect occurs.

Wow they had those running since march and april of last year? Yet its taken this long to prepare to be able to sell units?

-MarkM-

BS. You are totally lying now. Welcome to the list of trolls.(bdpuke)
Bitfury's 110gh rack is HUGE, and consumes 10 kilowatts!!!!
http://www.bitfury.org/bitfury110.html

28nm is tiny, not huge. same as in my 7970

if what you're saying is true, and this other site is real, then hats off to bitfury, but ill stick with knc..... I'm skeptical of the post, because its not on bitfury's site even. bitfury.org doesn't mention it.

You are just digging your grave of incompetency even deeper.

Only shaders on the GPU are repetitive. All the I/O: clocking, PCIe interface, memory interface, video interface and most of all the supervisory execution logic are non-repetitive.

That's why I compared it to GPUs.  Which, like ASICs, are dumb repetitive.

BS. You are totally lying now. Welcome to the list of trolls.(bdpuke)
Bitfury's 110gh rack is HUGE, and consumes 10 kilowatts!!!!
http://www.bitfury.org/bitfury110.html

Not sure if you are actually trolling now.....

Bitfury's chips can go up to 5 GH, but their power consumption goes WAY up.

They choose to keep their power levels down to around 1W/GH which only allows the chips to get maybe 2.5-3 GH I think last I checked

I MUCH rather have a KnC Chip that goes 100GH and uses a little more power then it be throttled to 50GH and use only 50W.

This discussion has already been made, the cost of power vs the amount you are getting from mining is negligible currently at even what you just stated.

If you are worried about the future, it will be a long long while before power costs = mining revenue (By then there will be many more generations of these chips).

Your analysis is 100% correct and 100% inapplicable to the SHA-256 hashing chip.

You cannot compare yield of a supremely-complex OoO CPU like x86 with a dumb repetitive SHA-256 hasher.

The bigger the dies, the lower the yield, the more you have to charge to cover your costs on the wasted dies from a wafer.

This is the best piece of info in this entire discussion.  Unfortunately Fabs almost never release their defect density and yields, unless you are a customer under a Non-Disclosure Agreement.  In almost two decades following CPU/IC production I have never willingly seen a fab release this info.  Very rarely you can get it leaked from a former employee, but then it is always tainted and uncertain if that person has an axe to grind with his former employer or not.  Alternatively, sometimes, you will get vague impressions from the executives during financial disclosures, but never the detailed info you want.

Now, the only thing apparently known about KNC is that the package size is about 3000 mm2.  Let's assume I am wrong, would not be the first time, and that they have a relatively small chip for the package size.  Say 15-20%.  That would be a die size of 450-600 mm2.  That would still be BELOW the lowest curve on the above diagram.  That's really scary for yields, but not impossible to obtain working product by history.  But any customer with a die size that large would have to expect very low yields, and they probably had to sign a waiver from the foundry about guaranteed yield due to "design constraints".

Here are some historical data points for 3rd party fab yield problems with some of the larger GPU dies, for bigger customers:
http://www.geek.com/games/ati-and-nvidia-have-troubling-40nm-yields-from-tsmc-815501/  which were "fixed" a year later:
http://www.dailytech.com/TSMC+Says+40nm+Problems+Resolved+Preparing+28nm+Fab+Production+/article17355.htm
http://www.eetimes.com/document.asp?doc_id=1261006  which were "fixed" about 10 months later:
http://www.pcmag.com/article2/0,2817,2411985,00.asp

Now, a TSMC/UMC is going to work with an AMD/NVidia to improve yields because they are producing 10-100s of millions of chips.  For a smaller companies, they will not go to great lengths to help them revise designs and processes to improve yields.  You get what you get.

Traditionally GPU dies have been some of the largest dies, with the high-end (low yield) parts topping out around mid-500 mm2 in size.  And the parallel structure of a GPU is what most closely resembles the SHA256 ASICs being designed and produced by companies today (Avalon, BFL, KNC, BitFury, etc.).  So from a manufacturing standpoint, we can probably infer a lot from large ASIC production issues by looking at issues that GPU manufacturers have had in the past.

The bigger the dies, the lower the yield, the more you have to charge to cover your costs on the wasted dies from a wafer.  Therefore it is always good to make your dies as small as possible.  Just 10 mm2 per die, when factored over an entire wafer, can mean tens of thousands of dollars in improve yield per wafer.

For comparison, here is a good illustration from the 45 nm days, with Intel - who historically have much higher yields than the rest of the industry (the invest more in their process and refinement).



Again, I hope KNC brings to market a good product, and in a timely fashion.  The more competition, the better we all do in terms of up-front pricing for our miners.  I just am concerned that we haven't see any substantial data about the processor (number of SHA256 cores, clock frequency, die size, heat production, etc.).

Considering the 10+ million dollars of orders they have had, you would hope they would publish this info to keep their investors/customers happy.

Please do, We're tired of hearing you whine about nonsense. no other company can compare. I like the coolaid, it's a winning recipe, thank you.

The why you should care is because low chip yields = fewer chips = slower production = delays in getting to customers.

You can have the most efficient design in the world, but if you cannot reliably produce it or make enough to sell in any significant quantity, it's worthless.

You think they can't do what they say?    I believe in them (KNC)100%

The why you should care is because low chip yields = fewer chips = slower production = delays in getting to customers.

You can have the most efficient design in the world, but if you cannot reliably produce it or make enough to sell in any significant quantity, it's worthless.