Topic: Swedish ASIC miner company kncminer.com - page 400. (Read 3050076 times)

ok this is just me looking at your board above... not my buddy who 'maybe' could fix it (yeah with the resources of NASA likely! heh)

(er sorry 'too soon')

anyway for my own info is/was anything in particular broken off/re-soldered replaced or whatever and if so a close up and
could you circle it...if there is anything particular we can see in the pic?

or is it cosmetically ok ..just plain won't hash 'well'?

but to my untrained eye I see zip (of course I know zip about electronics that is my buddies bailiwick)

Searing

edit: er duh I assume it is supposed to have those yellow block items surrounding the main chip (what exactly is on them..do they all say the same thing?) and again from what you can tell were they just 'mounted' on a kinda solder pad? I think I see?

and of course lastly if my 'guess is correct' do you have those items and can you show us a back side to them to see how they are attached (close up)

anyway he does not mine...so if he was to suggest how it is re-attached it would be from an electronics/TV background and again he is one of 3 to the state 'certified' to repair the robot made micro circuit LCD TV's now a days

anyway I'm sure others on here have better advice ..but again go to your local univ electronics dept (EE or EET) beer and pizza go a long way..even check the tech univ or local guys..

again hopefully others on here are more helpful

Searing

Repaired or Not...It WILL be hashing next week!!!!

Behave yourselves girls !

bring out the gimp

You're funny roadstress

these are working boards, just low, low hashing boards...too many pieces got ripped off these boards.(frankenshippment)  ya, cant send these back...getting 4.2 so far

You lack neurons. Really! I just told you that a 40nm chip has better performance than KnC's 28nm chip. This is real!

A 28nm chip will have better performance than KnC's 20nm chip. Are you that retarded to not get it?

You can get better performance from the 20nm chip.

http://en.wikipedia.org/wiki/AMD_GPUs#Volcanic_Islands_.28Rx_200.29_Series  (for example)

Are these extra boards or did you get the unassembled kit? I ask this because if you have extra boards, you can't RMA extra boards to KNC unless you ship everything(!) back. So you're just stuck with broken boards. If you just received the single kit, then you should RMA the broken boards since you have nothing to lose.

I have 4-6 boards that need repair, I havent been able to find anyone.  Im still waiting on beaglebones...I have 1.5th?? sitting(in knc)

You are so clueless. SP10's chip that is on 40nm is more power efficient than Jupiter's chip that is on 28nm by far. (0.85W/Gh vs 1.1-1.2W/Gh at system level) SP30 will be at 0.5W/GH at system level which is on par with your prediction which will turn out false. I feel that Neptune will be 0.7W/GH which is worse than SP30. So why pay the expensive 20nm mask if you can get the same performance from a 28nm chip? What's the advantage?

What makes you think there are no issues with the design?

Why could knc only achieve 1w/gh on 28nm where as bitfury could achieve 0.8w/gh on 55nm?

The Neptune can probably beat 0.5 very easily unless there's something wrong with the design/process, it's all about how much performance you have to sacrifice to reach a particular power target and if it makes sense at the time. Just look at the span on the bitmain 55nm chips, they can get anywhere from 1W/GH (probably a bit lower than that too) to 2,5W/GH depending on how high it makes sense to clock them and what voltage you need.

All things being equal 20nm will outperform 28nm for density/efficiency unless there are issues with the process/design, not saying a optimal 28nm design couldn't outperform a 20nm product, however as long as nothing is broken 20nm "KNC" will be better than 28nm KNC if you get what I mean so it ends up being better for them to rush 20nm.

The main reason that 20nm is no better than 28nm is the cost of the NRE, you get no performance/$ benefit and cant increase performance at a certain price point, hence why most of the industry is sitting back waiting.
KNC however took the easy route and had someone else fund their little enterprise so for them it made sense. The cost/transistor for 20nm will most likely drop below 28nm in 6-12months time and then KNC will have an advantage over the competitors who still are stuck on 28nm as long as they can match them on efficiency/density.

A few people received a pile of trash. Others received unassembled yet working miners once assembled. The luckiest received (poorly) assembled miners and extra ASIC boards. Many needed beagle/controller board flashing, heatsink repositioning, fan replacement, etc.

I'm going out on a limb here but I think most people got them working to some extent and some have them working near perfectly and have bonus hash from the extra boards.

I actually think much of the rage was because of the poor quality of the product assembly and the shameful packaging of the shipments. And everyone who received these abortions had to work way too hard to get them running. But ultimately most people got near the hash that they (over)paid for.


edit... I misread your post and gave a completely unrelated answer. Sorry!

i got a guy that supposedly can do it...only one of 3 in my state that can repair the newest of TV's *unfortunately it hardly pays to use a microscope to fix 'tossable' walmart LCD TV's etc made by robots etc

but then again you have to trust me trust him to try to do it ...trust he could do it....etc etc

i suppose if you just are at the point it is 'all junk' in your opinion or your techs you could send it to me

but might be like doing heart surgery by a vet on a stray cat you hit with a car....may not be worth the effort

pm me if it really really unsalvageable and you could ship it to me and we could look at it..but then again that whole trust issue

(all you guys have to move across town so we could have  beers and really really get a PH farm going! too much distance!)

anyway all I have..some? security that if my Titan shows up as a brick I have an 'option' with him getting it going..if he can't do so
well..that means you got something the local EE dept at your univ can't do either for beer and pizza

another option chase down the electronics dept at a local school/univ etc  beer and pizza an access to state of the art equip might
help pull something like this out...and you could get some newbie bitcoin recruits

all i got

good luck

Searing

Here is a link you might find interesting. It provides information on geometry, die size, transistor count and performance of some recent AMD GPUs from the R9 2XX series. It clearly illustrates how scaling geometry (40nm -> 28nm etc) effects performance and transistor count. Some of the latest 28nm designs contain over 6 billion transistors and provide over 5 tera-flops single precision performance in a 438 square millimeter  die size.  When AMD introduce a 20nm node GPU, it'll be a monster.

The same holds true for ASICS.

http://en.wikipedia.org/wiki/AMD_GPUs#Volcanic_Islands_.28Rx_200.29_Series

Jimmothy, that chart only tells a very small part of the story - the impact of node scaling on transistor cost given certain assumptions about wafer cost and yield that are typical of Nvidia's processes.

When each one of those transistors is producing BTC instead of drawing pixels on the screen, analysis of relative transistor costs is a secondary factor compared to hashing speed and power consumption.

They did mention re-submitting the code during the Christmas week... Which is part of the shuttle-service during the design phase; so I think Jimmothy is right...
they may use 20nm to 16 nm shrikage plan, lol....   wonder if it will employ Finfet 3-D as well.....   hmmmm
I'm also thinking that "cost effectiveness" won't be determined until we know for sure about our ROI possibilities.
Yeah, I know most think Neptune will never ROI, but you never know what the future holds until it has arrived.
KNC also have alot to "make up" for, and they just might "pull it off".
*Barring the FrankenJupiter mess*... That was Freakishly bad for PR, but now those guys are now hashing and have been hashing for how long now?
Anyone care to share what they've made BTC wize with FrankenJupiter so far? 

A better/smaller process node is equal to zero if the performance is worse than a bigger process node. There is no advantage there. Why would I choose a 20nm Neptune that has ~0.7W/GH (just an assumption) when I can choose the 28nm SP30 chip that has 0.5W/GH at system level?

Doesn't really matter since they used crowd funding to pay for the NRE (the big drawback of 20 vs 28nm), long term for KNC they got a win since they are on a "better" process node than the competition and the only ones that lost out on the development time were their customers (KNC were happily mining away with old gear anyway). Unless BTC price goes up considerably it is going to be really hard for another HW manufacturer to justify/raise the money to go 20nm once the mining market is saturated (at least until the mask costs comes down).