Topic: Swedish ASIC miner company kncminer.com - page 63. (Read 3050071 times)

Its the name of the game man! You have people willing to line up and shell out big bucks for shitty products ... any company will take advantage of that and lead their customers to the slaughter. Just look at apple.

The sheep are just even more retarded in crypto world cuz everyone will sacrifice everything to get more hash.

NOTE: Rant Below! Title of Rant: "How to be the best evil asic genius type of company you can!" You don't like my rants put me on /ignore I don't mind



Yeah again from what I understand following the old KNC forums (others correct me if this is incorrect) is the following:

The original Titan was 250mh but due to delays they then upped it to 300mh (standard) it was only AFTER they had the bad die issues etc that they told folk
it was OK to run them at 325mh or over clock them in my case in KNC forums PM with tech staff 1 month after I got my Nov 6th unit more or less.

....so due to late delivery imho and the fact they were 'designed' again imho with an over capacity
at 250mh.....well......the initial push to 300mh was do'able ......they could up the speed a bit and toot their own horn as a result in that it was 'overbuilt' on the
original 250mh speed....then they got stuck with bad product design and no slack in the setup when they pushed it beyond that point when they had bad die
issues and other problems with the Titans....a step to far so to speak

But with problems with 1 batch dies and such out of the gate for the Titans...they then upped that an'ti again and made the (in reality) the previous OVERCLOCKED setting of 325mh
the standard settings on the dies was an act of desperation to make up for the lateness of the delivery..... by touting the increase in speed....and the crappy construction
of the 1st batch units especially (late) and the first couple months....they of course had to push it as far as possible to play 'catch up"

To top it all off...knc probably figured it was no big deal to do so also from an RMA point of view.

They originally claimed (a change from 1 year not touted) that TitansONLY had a 90 day warranty...so this would have worked well on them not having RMA issues and the customers being stuck...but they neglected to change their warranty info on their web page for a month (still said 1 year) and of course the firestorm of the 1st batch titan units being crappy out of the gate ..they backtracked and made the warranty 1 year again. Said it was from the kindness of their hearts was really because of all the first batch units ordered that first month they had prob 1/2 90 day warranty and 1/2 1 year..so they made everyone 1 year..but again it was from the mistake on the web page on ordering and people had proof so they were kinda stuck.

I always keep in mind these raspberry pi piece of crap 8th grade science project units ...were 'designed' to only last 90 days or a bit more.....The fact my Titan(s) both
Nov 2014 units probably will go beyond 1 year in 2 weeks is a frigging asic miracle imho (you have to take what you can get from all this positive wise)

So to sum it up there was no slack left in the Titan setup was kicked up from 250 to 300mh .....then in desperation due to all the crap i mention here with the problems they had with the first batc... to the max of 350mh...so it was a big change in policy.....ie run the baby full out over clocked from day one and let crap burn out as it may..if you got over 300mh on your hacked together overclocked unit ....no RMA if only 1 bad die per cube ..your prob still over 300mh (in my case 306.1 with 1 dead die each of 2 cubes) you are out of luck too

I know for a while they would NOT RMA when you burnt out your cube for not using a Y adapter or other issues..they MAY have changed that but I remember a lot of angst
in KNC forums on KNC saying it was a PSU issue etc on your end

It is a frigging miracle the PSU plugs for power on the cubes don't  burn out once a week in that they made standard settings/speed imho for 250mh unit and now it is  100mh more then they designed the units for at 250mh (with probably the proper 'over build' of components at that setting vs the totally 'inadequate' way they were move up to)  for KNC  to cover their ass...and paper over the crap they were pushing. (don't get me started on the first couple months where the firmware sucked a whole nother rant!)

When I got original Titan Nov 6th 2014... this 325mh setting was NOT the norm..... I was only informed of this when 2 dies on 2 different cubes died then I was told this was OK to do by knc techs via knc forum email)..so was not the norm for me .....but due to again the above issues they had with the 1st batch it became the norm......

It also worked out for KNC in that when i tried to get an RMA for 1 dead die on each of 2 cubes they said I still got over 300mh so they would NOT RMA (probably
due to the swamp or RMA's for the issues I've stated above) but that was based on the overclocking of the other cubes to the 325mh setting on advance page..so they
scam'd their way out of doing RMA's on what should have been the standard speed of 300mh by the customers eating the problem and attempting to burn out the
power connectors and other issues.....evil asic geniuses they are indeed!

In my case being told after having the unit a month it was OK to over clock my 300mh settings to on adv page on the Tiitan (then gee guess what I had 2 dies 1 on each cube go dead) then being told by KNC that I should overclock for dead dies the other dies at 325mh ....so I did..it worked...but then I said I want to RMA my cubes... they say we don't RMA cubes with only 1 dead die each and you said you get 306.1mh with your 2 dies set to Off anyway...nor do we do so for any Titans getting over 300mh ...no RMA's (if I would have had 3 dies dead on 3 cubes then as it would be under 300mh then I'd get them RMA'd etc)


So talk about having your cake and eat it too......they tell you to overclock the titans to 325mh settings from their own standard of 300mh and barely designed yet to do and they flake or burn out connectors on your psu cable connector or your cube connector etc or what happens in my case maybe the dies? or you have legit RMA issues at 300mh with 2 dead dies on 2 cubes and then they revise the rules on what is standard (no longer 300mh now 325mh overclocking)or not to tweak the warranty and rma stuff in their favor. Pretty slick way to dodge any real reasponsiblity imho.


Last point in this rant ..they initially said you did NOT have to use Y adapters for your PSU's either .. 300mh on adv page was before they shipped ..just 16 awg wire on PSU's.
after folk started blowing the power couplings up..they then said they recommended such and finally said they would come with your titan....I got 3rd party Y adapters for the psu
before even turning it on.....I got the Y adapters from KNC laying around here that arrived 3-4 months later if I remember correctly....man they really really did not give a damn about customers my guess is they knew at that point (pumping Neptunes in their farm) that we were the end of the home miner consumer chain and they could do as they damn well
pleased I guess...they are their only customer now with the 16nm chip and their newest data hall

well played knc well played (they really do have this whole 'evil genius' thing down don't ya know!)

on a side note  the only evil genius flubs that KNC has done imho is 1) why did they keep knc forums up as long as they can before pulling the plug? (I sure wish I would have archived it) and 2) why the hell have they not 'locked' this thread up yet? Other then that ....everything else pretty much seems to have gone their way ..being evil works well I guess

Not sure if you really ripped them off. At one time if I remember correctly they 'claimed' that the warranty follows the device..if that is correct it was probably 'safer' to mention it in the manner you did (hey they may have just been saying that) but on knc forums (now down) I seem to remember them saying that (of course they said a lot of stuff before the titans shipped on knc forums)

but still ....myself if I ever had the same situation as you ....I would do it in your manner...nice to know you had some 'good luck' ...... BUT maybe they are keeping their word

(I know it is unlikely..but then again you got quick fast RMA service from KNC so perhaps the age of miracles are upon us!)

Anyone else remember the same claim from knc forums ...that the warranty follows the device? Anyway what I think I remember for what it is worth

Again good for you...a win!

As posted above: PER DCDC its doing this calculation: P * 1.15 + 5    (Measured power , 85% eff, add 5w for DCDC power dissipation)

It will  add all 8 of the DCDC's power as guestimated by above formula to show the cube consumption.

Thats the new cube formula, so its still an estimate, but its a lot closer to actual cube power consumption.

Awesome work, Tarkin! 

Were you also able to correct the incorrect power reported for each cube?

Ok, well I just finished up coding it all, now theres an additional section in the SYSTEM page that allows you to specify PSU effeciency, ranges from 50-100%
It will update the % on the advanced page whenever its changed.
It defaults to 85% if not specified prior(ie: before upgrading to new version)

These updates will be in next release =)

I wonder where the scryp midstate n crap is calculated ... on the controller board? if so then that power usage will spike up a tad when a new block is found or a work restart.

Yeah, really really low. Guess the CPU isn't getting much of a work out!

That's 0.45A * 5V = 2.25W 

WTF?! thats insanely low ROFL! thats 2.25W or 2.25A? LOL =P

Yep. Shame that KNC didn't spend an extra $10 per cube to put the 50A VRMs and a second PCI-e connector. 

Just measured the controller for ya. It's using 2.25w on the 5V rail, and nothing on the 12V rail. That's with 5 cubes connected and the LCD screen. I'm surprised by how low that is!

Looks like my calcs are spot on!
Except for that first cube, Im gonna assume the increase in power on that cube is because of the DCDC's running in the 42+ AMP range... man effeciency must PLUMMET drastically beyond the 41 AMP range!

I must say, Its fucking crazy that *everyones* PCIe connectors arent melting ALL THE TIME! Thats 27.5A ... thats way above the 23A rating, its even above the good ol "10% specs deviation rule for electronics" which in this case is 25.3A
~ only if we could make 14V go through the PCIe connectors instead =P ... would take a decent amount of load off them and probably make the DCDC's work a tad more effeciently. LOL!

Fun times!

Any way you can measure the power leads going to the controller board from PSU?(if thats how its powered)

Here ya go, Tarkin:



I can't get a reading on the 5th cube for now as it's temporarily running on a Seasonic ATX PSU. I can't get my clamp on the 12V wires since it's sleeved.

Would you also like a reading of a cube running 325MHz at default voltage, (-0.0366v) or is that unnecessary?

That sounds much much better! I just got home so I'll get the measurements soon for ya, along with a screen shot of the advanced page for all cubes.

The only way someone could get to 90% DCDC efficiency is if the underclock the cubes, right? Doubt anyone is doing that. I wouldn't worry about it too much if I were you. What's more important is getting the proper reading for normal use cases.

Ok, here is my updated formula:
PER DCDC its doing this calculation: P * 1.15 + 5    (Measured power , 85% eff, add 5w for DCDC power dissipation)
Which comes out to about 330w per cube. (on my titan)

Then for wall watts its:

Cube power * 1.15 + 5 + 25 (Total cube power as per each DCDC added together from above equation, 85% eff PSU, add 5w for pi, add 25w for controller board)

Gives my total wall watts of 1530w, which is EXTREMELY close to my measured 1515W at wall.

I dont really know how much the controller board consumes, so thats a pure guess LOL!

If I run the SAME guestimate through ur 5 cubes and ur 94% PSU, it comes out at ..... 1779W!! Which is also very close.

What ya think?

The only thing I cannot account for is all the different effeciency levels of the DCDC, so if someone is running the DCDC at like 90% eff, my calcs would still assume 85% =P

In regards to custom effeciency setting, I was thinking, for sake of cleanliness I would just add it as another setting under the SYSTEM tab, thoughts?

In that case, ours is on page 9 =) ... very close to the same.

WOW! 360w?!?!?! ARE YOU SHITTING ME?! I wanna see ur measurements ASAP =) haha
naw, the 6 pin PCIE is rated 75 & 8 pin rated 150w for PC applications, its the standard in the video card / PC industry. But its not necessarily the max of the CONNECTOR(not the cable) itself. The wire gauging used by PC PSU manufacturers puts the 75/150w limit on the PCIe power cable / connector overall because some PSU manufacturers use way too thin of a guage of wire to safely sustain the connectors rated max.
Molex.com has CURRENT ratings for the 6pin PCIE connector.
http://www.molex.com/molex/products/family?key=megafit_power_connectors&channel=products&chanName=family&pageTitle=Introduction&parentKey=power_connectors

23A * 12V = 276W, I round it to 275 for ease of remembering =P


ALSO, I just stumbled across something else, the other missing piece of the puzzler perhaps.
If Im thikning about this correctly, power dissipation would be additional to the effeciency rating of these DCDC's?
If so then thats another 5w added to the power needs of each DCDC, so its roughly 85% eff + 5w PER DCDC, thats 8 DCDC's per cube, which translates into another 40watts used per cube on top of the 85% eff... HOLY SHITBALLS!

Or is the dissipation part of the effeciency metric? I dont know, in the docs it shows 2 seperate graphs , 1 for effeciency at given current/voltage and another graph for power dissipation for given current / voltage.

If I'm not mistaken, that VRM PDF you linked is not the right one. That 50A unit was used in the Neptune only. KNC cheaped out even more by using a 40A unit in the Titan.

I think you're a little behind on the news LOL! It's (I thought so anyways) well known that both Neptune and Titan cubes pull approximately 360w to 400w from that poor PCI-e connector. A 6 pin PCI-e connector is rated at 75w and an 8 pin is rated at 150w. I'm not sure why, since all it does is add 2 ground wires while only leaving 3 12v wires. Where did you get the 275w figure from a 6 pin from?

AWESOME!

and you know what, I just figured out!!!! NO WONDER peoples PCIe cables get fried left and right on thse things.
Now that weve exposed cube power consumption is much closer to 300w ... its a no brainer! The poor lil PCIe 6pin power connector, as per molex's website can deliver up to 275w safely @ 12v .... these poor things are running on the verge of meltdown 24/7 essentially! LOL
Not to mention we already run the DCDC's beyond spec of their current deliver capabilities!  LOL!

Im sure glad I never figured out how to OC beyond 325mhz ... peoples cubes would be melting left and right!


FU KNC for being a bunch of cheap bastards! shoulda went w/ an 8pin connector or maybe even 2x 6pin per cube! ...

I'm out running errands right now, I'll get you the measurements in a few hours.