Topic: Swedish ASIC miner company kncminer.com - page 399. (Read 3050076 times)

Energy use is secondary now that hashrate is 99.5 Billion and rising.

Personal testing and guessing, I got mine to around 0,91-0,93w/GH at the wall while keeping it above 500GH, after that some of the cores started shutting down. So 0,7 should not be much of a problem since people underestimate how quickly efficiency diminishes/increases due to the exponential scaling with core voltage, all about how much performance you are prepared to sacrifice to get there.

I can check out how low I can get it when I have physical access to my jupiter in case I mess something up and need to do a hard reset Depends as well on how low you are able to go for voltage in software as well on the Jupiter, not really a fan of doing hardware modifications to prove a point.

and what asic company isn't? if you name one yer brain dead

There is no room in the case to put more chips so they can't do it. And even if they could why would they sell something underclocked? It means less chips for them and by now we all know that they are greedy cunts so they will never ship underclocked stuff.

If they can achieve 0.5w/gh at the wall, there is no reason they couldn't give you enough chips for 3th/s. And if they give 3th/s worth of underclocked chips we might see 6th/s overclocked neptunes.

It's pretty much up to knc how much hashrate they will give.

Although they have shown they don't really give a shit about their customers so who knows.

Oh didn't got the Jupiter part. Well ok, but even so...having a normal 3TH/s Neptune at 0.7W/Gh you will have ~2.2TH/s(or something close) if you downclock it to 0.5W/GH which would suck considering that it costs so much.

No he said that the jupiter chips are capable of 0.7w/gh at the wall.

If that is possible then 0.5w/gh at the wall on 20nm should be possible.

This comes from KnC's statement that the Neptune will require 30% less power than the Jupiter. No final/real numbers yet because there are no Neptunes.

To simplify what you said:

Jumping to 20nm using customer funds = great idea

Jumping to 20nm using your own funds = possibly a good idea

In terms of $/gh, knc will definitely have an advantage as soon as manufacturing costs for 20nm come down, but I am sure we will see plenty of competing 20/16/14nm and even some highly optimized 28nm chips by then.

Not sure how knc's 20nm chip will fare in terms of $/GH and W/GH but I doubt that they come out on top for both efficiency and cost. At least without a design improvement.

Also you mentioned that knc chips are capable of ~0.7w/gh at the wall. Got a source?

Ofc there is since the industry never stands still, only Intel (however their 14nm launches has been delayed) is currently moving over anything major to 14nm and they are usually 12-18 months ahead of the rest of the industry when it comes to production. Usually when foundries starts talking about a node you can be sure that their mass production is still 12 months+ away. Samsung will probably start their 14nm production next year, good luck getting access to their fabs for something like btc mining however. 20/16nm at TSMC/Globalfoundries will be as good as it's get for 2014/2015. The only opening I can see in that time frame iirc is that GF struck some kind of deal with Samsung on 14nm tech utilization, so they could be converting sooner rather than later. However converting a fab is not something that you do overnight exactly.

Also doesn't change the fact that for every new process node the costs have increased exponentially for RND while the gains in power reduction/performance increases have been decreasing. It used to be that you should jump on a new process node asap since you could offer a product with better performance metrics at a lower price, these days however it makes more sense to wait until the new tech has matured unless performance/density are your only concerns.

The whole argument has been about why you should go 20nm instead of 28nm when the NRE costs are higher while the cost  for performance x with efficiency y is the same on both processes. Currently it costs more to develop for 20nm than 28 while you gain almost nothing in terms of performance/$, this will change with time however when the process matures and price comes down for manufacturing.

However if you like KNC simply make someone else pay for your NRE costs it made a lot more sense to jump directly to 20 vs 28nm. They could have offered the same kind of performance at the same price with a new better 28nm design and delivered it sooner, they wouldn't have filled up their orders as easily however since "OMG 20NM SO GUD MUST HAVE" mentality. In the long term KNC will be the winners since costs of 20nm wafers will go down faster than the cost of 28nm so they will be able to fill their datorhall(s) with higher density chips paying less/GH, and we paid for it pretty much.

A 20nm GPU has a different design than a ASIC 20nm chip. I can see Nvidia having troubles designing and manufacturing their own custom 20nm chips, that's likely, but comparing them to KNC's chip is like comparing an apple to orange. There is an IC company that has or is in the process of completing a 15 or 14nm chip but it too isn't an ASIC or GPU design.
 

Did anyone really fall for this promise?

KNC always wait until the last minute, then make excuses to cover their lies.

They knew very well in early October (looking at how many units they had to ship and how many were coming off the production line each day) there was no way they could ship all Jupiters by the 15th October, like they promised. Rather than tell people when they knew it, they didn't announce this until after they missed this date - 16th October.

Likewise with November Jupiters, they didn't announce until the day they were supposed to start shipping (i.e. 15th November) that there would be 2 weeks delay.

They always mislead people into thinking everything is going just fine, then wait to the last moment to announce delays/fuckups. This will be especially true when it comes to them having to give away a single bitcoin to their competitors, i.e. customers.

So don't expect anything until the July 1st "unexpected delays" newsletter

They will not have any advantage because they were never interested in designing the most efficient machine, they were interested in taking money for pre-orders to do OTHER projects (aka datorhall).

What is the full part number on those Ericsson VRMs ?

Is it a BMR464 as shown here: https://github.com/KnCMiner/jupiter-hw-docs/blob/master/Ericsson_BMR464_Technical_Specification.pdf

Looks a lot like: http://www.mouser.com/ProductDetail/Ericsson-Power-Modules/BMR4642002-001/?qs=sGAEpiMZZMsc0tfZmXiUnRXPWXedK30lvt%2fDJN7MGhU%3d

I think to be fair at system level a Jupiter is a lot less than 1.1-1.2W/Gh

My sole surviving one sits at 840Gh at 800W at system level and thats obviously overclocked.

Firstly, what happened to Plan B and CD batch Neptune customers getting cloud hashing in "early June"?

Secondly, I don't know why I haven't done this before, but I went to the KnC site to get the measurements for Neptune so I could plan where I'm going to put it.  600 x 400 x 300 according to the Neptune Third Batch page here: https://www.kncminer.com/products/neptune-third-batch

300 mm high! 1 RU = 44.45 mm, so 300 / 44.45 = 6.7 RU!

Damn, that's a frikken huge box!

For comparison, Jupiter is 500 x 400 x 200.
200 / 44.45 = 4.5 RU, which is pretty big already.

Consider that an SP10 is 1.25 RU, you could stack 5 SP10s on top of each other (6.25 RU) and still not be the same height as a Neptune.

The Asic fairy may well deliver a special package to  you this week ;-) 

At a quick glance... it's missing 3 VRMs! Wtf?!

I'm not saying there couldn't be be issues with KNCs 20nm design, the whole argument has been why it made sense for KNC to jump to 20nm rather than making a new fully custom 28nm design when there was no initial cost/real performance/$ gains from doing so.


It has been speculated (confirmed?) that the Jupiter design is a hardcopy and not a full custom design (hence their fast design/delivery time and worse performance/effiency). Also the Jupiter chips are more than capable to get below 1W/GH (you can get down to about 0,7-0,8W/GH at 400-450GH) but why would you sacrifice the performance for efficiency when running them at higher speed but worse efficiency has been more profitable?

To properly compare 2 different chips you have to look at the given performance at a certain amount of chip area. Take 100mm2 of total chip area for 2 design then have a specific performance point and compare the best efficiency at that performance point, that's the only way you will get real comparison on how 2 different designs stack up to each other in terms of efficiency. Otherwise you can easily push the metrics in whatever direction you want.

So let's take a look at Bitmaintech and their S1s, they sold them at 180GH@2W/GH (a much worse chip in your world) and is clearly inferior to Bitfury by your logic. I could however take a S1 Antminer and probably get it below 0,8W/GH but you would only get around 70-80GH out of it. But by your logic I have just taken this inferior horrible inefficient chip and turned into something that is better/equal to Bitfury. Since Bitmain is on 55nm as well it would be really easy however to compare who has the lowest manufacturing cost to produce those results, so who out of Bitfury and Bitmain uses the least amount of die area (and hence lowest manufacturing costs) to produce say 100GH@0,8W/GH? Not the faintest idea, something for you to figure out I guess!

What my whole point is that the same chip can be used for a usually quite large range of performance targets depending on how you tune them.

The rule is however that as long as the designs are in the same ballpark for optimization it will take substantially more die area for a chip on a larger node to match the same performance/watt as a chip on a smaller node. Hence why saying that Bitfury 55nm is clearly better than the KNC jupiter is not as clear cut as you might think.

Unless your are selling into a specific market where performance/watt is EVERYTHING selling chips at their "best" efficiency specs doesn't make sense since the performance tradeoff to get there is to substantial. People pay for performance and the scaling of the chips will mean that as a manufacturer you are looking selling them at a performance point that makes sense and generates the most profit.

That is so crappy to send this kind of used crap to customers that paid 10-13Gs for equipment. That's why I hate these guys more than anything.