Author

Topic: Swedish ASIC miner company kncminer.com - page 1690. (Read 3049528 times)

sr. member
Activity: 476
Merit: 250
let's have some fun
September 10, 2013, 06:47:44 AM
One thing to watch out for is that KNC for some reason decided to use 4 PCIe connectors instead of EPS12V connector + 3 PCIe connectors.  On the 860 Corsair 2 of the PCIe connectors are on the same PSU connector (i.e. single set of wires with 1 PSU connector on one end and 2x PCIe connector on other end).  Without testing it there is no way to know for sure what the overcurrent limit is for that connector (good PSU limit both overall current and current per wire).  The PCIe standard only mandates that two 8 pin connector supply 300W (1.25@ @ 12VDC).  Now the PSU probably can supply a lot more than that but just by looking at it there is no way to know for sure.  Most reviews only show total load not max load per connector and certainly not at amperages beyond what a PC would use anyways.  
I believe such a limit would violate the EPS12V specification (see Table 28 of version 2.91, but other versions have a similar requirement) which specifies that overcurrent protection may not trigger until a rail reaches, at minimum, its specified peak current.

For those wondering what JK is referencing:
Quote
Table 28: Over Current Limits
Voltage  Over Current Limit (Iout limit)
+3.3 V  110% minimum; 150% maximum
+5 V  110% minimum; 150% maximum
+12V1  Peak current minimum; 20A maximum
+12V2  Peak current minimum; 20A maximum
+12V3  Peak current minimum; 20A maximum
+12V4  Peak current minimum; 20A maximum (22A maximum for 750W-800W)
http://www.pcpower.com/downloads/EPS12VSpec2_91.pdf

IIRC later version of the spec removed the requirement for multiple 12V rails and with it the max current on the 12V rail.  

Joel, I think what may be unclear Joel is that in the PSU linked above while the PSU has 6 PCIe connectors it puts TWO connectors on one strand of wires. So each pair of PCIe connectors is limited to 22A.  With three pairs of PCIe connectors at least one rail will need to drive TWO PCIE connectors and thus the load on that rail is half of a Jupiter.  To be clear this wouldn't be a problem if a PSU with four PCIe connectors connected each one to a seperate rail but for cost reasons they don't.  If you look at the wiring of all ATX PSU you will find each set of cables has two PCIe connectors in parallel.  So multi-rail PSU are going to be bad news.   

However it isn't quite that bad.  Later versions of the ATX spec removed the requirement for multiple 12V rails and today almost all high end PSU use a single massive (100A+) single 12V rail.  It is cheaper, more efficient, and easier to manage.  However despite all the connectors attaching to the same rail for safety no PSU is stupid enough to allow 100A to flow down one connector.   100 amps is a massive amount of power enough in a short circuit situation to cause a fire or death.  So while technically a PSU can deliver 100A+ on any connector (PSU connector not downstream PC standard connectors) no PSU is allowed to do so.  It is just a lawsuit waiting to happen to design a product than if a short circuit occurs will allow 100A+ to flow to wires and connectors not designed for that kind of current.  All single rail PSU employ over current rotection at the connector level to prevent the full PSU current from unsafely flowing down one connector.  They will trip if "excessive" current is going down one connector.  It might be 20A, 30A, 40A all depends on the design and configuration.  There is no way of just looking at the PSU what the limit might be.  To put it into perspective the PCIe spec only require 150W per 8 pin connector so a set of wire with two PCIE connectors in series "normally" won't pull more than 25A.

Lets imagine a Jupiter pulls 790W (65A @ 12VDC) that means the current at each PCIe connector is 16.25A. No problem there.  However if there are two PCIe connectors on a single "strand" and both are used then the current on the strand is 32.5A.  If the overcurrent protection for the connector is 40A well there is no problem but if it is 30A?


Thanks for background info, very appreciated!

So I guess the corsair is quite a good choice:
* dedicated single +12V rail with user-configurable virtual "single rail" and "multi-rail" software modes
* features a massive 71.6 Amp (859.2 Watt) single +12V rail
* can also be configured as a multi-rail device with individual PCI-E over-current protection (OCP) trip points
=> http://www.corsair.com/us/blog/ax860i_technical_details/
sr. member
Activity: 322
Merit: 250
September 10, 2013, 06:42:51 AM


Awesome.

I am looking forward to Over Delivery.
hero member
Activity: 692
Merit: 500
September 10, 2013, 06:12:25 AM
hero member
Activity: 798
Merit: 1000
September 10, 2013, 05:36:17 AM
And for those eagle eyes, yes, it does say "for a Jupiter running at the expected speed of 500 GH/s". That's today's newsflash  Grin

this is what I call a good news (if confirmed)

Tongue
As stated by KNC staff:
http://forum.kncminer.com/forum/main-category/hardware/107-psu-recommendation-jupiter-the-largest-planet?p=1050#post1050
legendary
Activity: 1260
Merit: 1008
September 10, 2013, 05:32:56 AM
And for those eagle eyes, yes, it does say "for a Jupiter running at the expected speed of 500 GH/s". That's today's newsflash  Grin

this is what I call a good news (if confirmed)

Tongue
hero member
Activity: 798
Merit: 1000
September 10, 2013, 05:24:37 AM
Regarding power supply, KNC Carl stated this on their own forum today:
"As some have stated above, having a margin on the energy consumption does reduce the risk of over heating and increase the efficiency of the PSU (at an increased cost for purchase). But you can opt for two 600W PSUs instead of a 1200 PSU or some other combination. According to our engineers an 850W, 80+ Gold certified PSU should be enough for a Jupiter running at the expected speed of 500 GH/s. "

So it seems you could buy a couple of 600w (or reuse ones you already have) rather than spending mega dollars on one massive overspecced one.

And for those eagle eyes, yes, it does say "for a Jupiter running at the expected speed of 500 GH/s". That's today's newsflash  Grin
newbie
Activity: 43
Merit: 0
September 10, 2013, 04:59:44 AM
Not to really take this off topic but how can it not suit anybody. its free power. looks like a win win senero on that

Well to my opinion nothing comes for free.
hero member
Activity: 574
Merit: 500
Mining for the hell of it.
September 10, 2013, 04:51:37 AM
Not to really take this off topic but how can it not suit anybody. its free power. looks like a win win senero on that
legendary
Activity: 804
Merit: 1002
September 10, 2013, 04:51:13 AM
I never had good luck with oil -- heat transfer coefficient is 3 to 4 times worse for oil, it's messy, it wicks, is more viscous (doesn't flow as well) etc., etc.  Oil is also a pain to circulate well in the tank itself, especially around heatsinks designed for forced air, and don't forget that it wicks.  And it wicks.  Drop a wire into the tank, and have a puddle on the floor in the morning.  4realz.  Anything dipped in oil becomes a pain to rework, no matter how much solvent you use (trust me, it will get on the board even if the board itself is not submerged.  And it *does* have a smell.  There's fun stuff i've only read about -- boiling point ~30 - 40 C, but it costs way too much.  Something like that.

It just needs to be properly planned and executed. The wicking is easily overcome by gravity and wick stops. Usually stuff you plan to submerge, you don't plan to bring out unless it fails. You can get it in bulk, like $2 a gallon or something, even less in some places. As for flow, inexpensive aquarium powerheads do the trick with ease. Mineral oil, and variant oils, are used in high end server racks all over the world. Just needs to be done well.


I wonder if that would happen - heat dissipation from the circuits -> oil -> glas should be sufficient if you use a fishtank. oil has a heat dissipation of ~0,13 (a fifth of water) and glas is at around 0,76. If you take the surface of a fishtank to be around 1 m² the temperature should even out at max. 65°C if your room temperature is around 20°C

It still needs a method of transfer. Air bubbles, generous surface movement, fans blowing on the glass, heat exchanger such as  radiator/etc. Otherwise with units like these, it'll cook itself. Without it, the heat doesn't dissipate, it collects in the bath. You can put a mild CPU (think single, maybe dual core, few years old) in there, but that's it. Not ASIC chips. You'd need a generous thermal exchange plan.

Oil is great if you plan it out. Super quiet and neat, and easier than watercooling systems (when dealing in bulk large pieces of equipment like this). For just a few chips I might consider water cooling. If I was serous about something other than air cooling for many units, I'd just design a simple submerge tank for oil.

I've always wanted to do a mineral-oil to salt water pool cooling datacenter hahaha. I have a 45,000 gallon pool which can easily be used to absorb heat from some racks, just need to have the oil heat exchanger inside of a closed circuit of the pool's system. Plenty of thermal capacity, and it'll warm the pool up for the crisp winter use.


The thermal conductivity coefficient is based on the thickness of the material, not the surface area.  That's why it's measured per meter and not per square meter. You have to do calculus to figure out how much energy will be transferred (but in the case of a flat surfaces it's not difficult)

Anyway, you would definitely need to keep oil circulating in order to keep things cool.  And I think anyone dunking a brand new KnC in a vat of oil is pretty stupid.

You should be able to use standard CPU coolers for this so just get some waterblocks if you don't want to go with air cooling.

It boils down to personal preference. I thought about both, like I mentioned before. If the waterblocks that I have fit then it would be no problem whatsoever. If they don't - mineral oil.
sr. member
Activity: 1176
Merit: 265
September 10, 2013, 04:37:24 AM
how does this affect possible roi for day 1 delivery ?

is this worst case or just probable ?

https://bitcointalksearch.org/topic/estimate-of-asic-pre-orders-13-to-15-phs-diff-18b-to-21b-by-end-of-2013-283820

brings tears to your eyes...  Cry


don't worry bro

quoting the relevant part:


Code:
Presales Total 2013:        6,281 Thash
Announced Total 2013:       1,750 Thash
Running Total 2014:         7,380 Thash
--------------------------------------------------------------
Combined Total:            15,411 Thash

Relationship between difficulty and hashing power
Code:
1 TH/s = 0.14 mil difficulty
1 PH/s = 140 mil difficulty
1 million difficulty = 7 TH/s
1 billion difficulty = 7 PH/s
1 trillion difficulty = 7 EH/s


so in march we will have 15.5 PH/s, that implies more or less 2 billion diff.
i've quickly input those data into http://www.bitcoinx.com/profit/ using 2.1billion difficulty and guess... what you'll mine  0.0958 BTC per day @400GH/s.

it's not that bad at current values that means almost 360 USD.

I know you should have taken into account pool fee and electricity cost but I left it as exercise to the readers Tongue

and as last point, hopefully by march you already get back our initial investment
 



Now that bit is what puzzles me, what KnC planned to do in march? That's gen 2 time isn't it? Going to need some serious kit to deal with that situation.
That guy says we can assume that people will stop mining once electricity costs aren't covered, but I don't think that's quite right...maybe when it's cheaper to buy BTC than mine them they will, but that's not the same at all is it? It's not free to buy BTC for a start.

I found a UK company who fit "free" solar panels yesterday, the catch being that you don't own them or get any feed in money...but you DO get to use as much as you can of what you generate which would be handy for a zero upfront cost reduction in lecky bills. IT wouldn't suit me because one day soon I want my own panels, and it won't suit anyone not owning their home...but it may suit some people quite nicely.
full member
Activity: 238
Merit: 100
September 10, 2013, 03:28:06 AM
I never had good luck with oil -- heat transfer coefficient is 3 to 4 times worse for oil, it's messy, it wicks, is more viscous (doesn't flow as well) etc., etc.  Oil is also a pain to circulate well in the tank itself, especially around heatsinks designed for forced air, and don't forget that it wicks.  And it wicks.  Drop a wire into the tank, and have a puddle on the floor in the morning.  4realz.  Anything dipped in oil becomes a pain to rework, no matter how much solvent you use (trust me, it will get on the board even if the board itself is not submerged.  And it *does* have a smell.  There's fun stuff i've only read about -- boiling point ~30 - 40 C, but it costs way too much.  Something like that.

It just needs to be properly planned and executed. The wicking is easily overcome by gravity and wick stops. Usually stuff you plan to submerge, you don't plan to bring out unless it fails. You can get it in bulk, like $2 a gallon or something, even less in some places. As for flow, inexpensive aquarium powerheads do the trick with ease. Mineral oil, and variant oils, are used in high end server racks all over the world. Just needs to be done well.


I wonder if that would happen - heat dissipation from the circuits -> oil -> glas should be sufficient if you use a fishtank. oil has a heat dissipation of ~0,13 (a fifth of water) and glas is at around 0,76. If you take the surface of a fishtank to be around 1 m² the temperature should even out at max. 65°C if your room temperature is around 20°C

It still needs a method of transfer. Air bubbles, generous surface movement, fans blowing on the glass, heat exchanger such as  radiator/etc. Otherwise with units like these, it'll cook itself. Without it, the heat doesn't dissipate, it collects in the bath. You can put a mild CPU (think single, maybe dual core, few years old) in there, but that's it. Not ASIC chips. You'd need a generous thermal exchange plan.

Oil is great if you plan it out. Super quiet and neat, and easier than watercooling systems (when dealing in bulk large pieces of equipment like this). For just a few chips I might consider water cooling. If I was serous about something other than air cooling for many units, I'd just design a simple submerge tank for oil.

I've always wanted to do a mineral-oil to salt water pool cooling datacenter hahaha. I have a 45,000 gallon pool which can easily be used to absorb heat from some racks, just need to have the oil heat exchanger inside of a closed circuit of the pool's system. Plenty of thermal capacity, and it'll warm the pool up for the crisp winter use.


The thermal conductivity coefficient is based on the thickness of the material, not the surface area.  That's why it's measured per meter and not per square meter. You have to do calculus to figure out how much energy will be transferred (but in the case of a flat surfaces it's not difficult)

Anyway, you would definitely need to keep oil circulating in order to keep things cool.  And I think anyone dunking a brand new KnC in a vat of oil is pretty stupid.

You should be able to use standard CPU coolers for this so just get some waterblocks if you don't want to go with air cooling.
legendary
Activity: 994
Merit: 1002
September 10, 2013, 03:25:48 AM

so in march we will have 15.5 PH/s, that implies more or less 2 billion diff.
i've quickly input those data into http://www.bitcoinx.com/profit/ using 2.1billion difficulty and guess... what you'll mine  0.0958 BTC per day @400GH/s.

it's not that bad at current values that means almost 360 USD.


360 USD per month is not bad, and I doubt the difficult will go up to 2 billion next year in march ...
legendary
Activity: 1260
Merit: 1008
September 10, 2013, 03:18:36 AM
how does this affect possible roi for day 1 delivery ?

is this worst case or just probable ?

https://bitcointalksearch.org/topic/estimate-of-asic-pre-orders-13-to-15-phs-diff-18b-to-21b-by-end-of-2013-283820

brings tears to your eyes...  Cry


don't worry bro

quoting the relevant part:


Code:
Presales Total 2013:        6,281 Thash
Announced Total 2013:       1,750 Thash
Running Total 2014:         7,380 Thash
--------------------------------------------------------------
Combined Total:            15,411 Thash

Relationship between difficulty and hashing power
Code:
1 TH/s = 0.14 mil difficulty
1 PH/s = 140 mil difficulty
1 million difficulty = 7 TH/s
1 billion difficulty = 7 PH/s
1 trillion difficulty = 7 EH/s


so in march we will have 15.5 PH/s, that implies more or less 2 billion diff.
i've quickly input those data into http://www.bitcoinx.com/profit/ using 2.1billion difficulty and guess... what you'll mine  0.0958 BTC per day @400GH/s.

it's not that bad at current values that means almost 360 USD.

I know you should have taken into account pool fee and electricity cost but I left it as exercise to the readers Tongue

and as last point, hopefully by march you already get back our initial investment
 

legendary
Activity: 2338
Merit: 2106
September 10, 2013, 02:01:45 AM
how does this affect possible roi for day 1 delivery ?

is this worst case or just probable ?

https://bitcointalksearch.org/topic/estimate-of-asic-pre-orders-13-to-15-phs-diff-18b-to-21b-by-end-of-2013-283820

brings tears to your eyes...  Cry
hero member
Activity: 574
Merit: 500
Mining for the hell of it.
September 10, 2013, 12:47:30 AM
Its the only thing that can be done untill mid october.  Cry Wish we did know more info but we just dont. So gota sit and wait!
eve
full member
Activity: 210
Merit: 100
September 10, 2013, 12:39:56 AM
Hi @KNC, out of first thousand miners you make, may I know, how many of them you,

1. keep to yourself?
2. host for others?
3. ship for others?

Am I asking too much? Smiley


They won't know or won't tell you. Where are your chips, has they been produced? - They won't tell you either, instead the shills and the trolls will asked you to shut up and be patience.
sr. member
Activity: 405
Merit: 255
@_vjy
September 09, 2013, 11:50:25 PM
You can take your order number from your My Orders page and paste it into the order tracking search box on the main page to see when your miner is going to ship.

It just says paid... That dont tell me much. Let alone when it will be shipped... Huh

Yeah, not much details. I'm also tracking 2 group-buy orders.
hero member
Activity: 574
Merit: 500
Mining for the hell of it.
September 09, 2013, 11:42:57 PM
You can take your order number from your My Orders page and paste it into the order tracking search box on the main page to see when your miner is going to ship.

It just says paid... That dont tell me much. Let alone when it will be shipped... Huh
sr. member
Activity: 405
Merit: 255
@_vjy
September 09, 2013, 11:36:20 PM
Hi @KNC, out of first thousand miners you make, may I know, how many of them you,

1. keep to yourself?
2. host for others?
3. ship for others?

Am I asking too much? Smiley
legendary
Activity: 994
Merit: 1000
September 09, 2013, 09:13:51 PM
Did the specs change recently on the Saturn or Jupiter? I could have sworn I was getting 400 GH/s but my Jupiter says it is a 200. Has it always been a 200?

Also from KNC's website in the product description it says won't ship until mid November. I thought these were suppose to ship in September unless that means if you order now.

Over 400 pages to read here and now they have their own forum gets crazy. That's exactly what BFL did. Instead of coming here they opened up their own forum. Similar minds think alike?



Jupiter = Minimum of 400 Gh/s
Saturn = Minumum of 200 Gh/s

If you purchase a new one today, it won't ship until November.  If you purchased prior to last week it will ship before the middle of October.

You can take your order number from your My Orders page and paste it into the order tracking search box on the main page to see when your miner is going to ship.

Please don't ever compare KNC to BFL again. You are only making yourself look uninformed and reactionary.
Jump to: