Topic: Swedish ASIC miner company kncminer.com - page 1391. (Read 3050074 times)

Sorry I should have been more specific..
it's electrical supply is 240 volt and they are measurements at the wall taken with a kill-a-watt meter.

Users are adding fans.  Why not go with 12vdc fans that accept the 4 pin connector.  Put in a splitter before the 4 pin molex in.  Drive 12vdc fans from that external splitter.  When unplugging the miner leave the fans plugged in draining any potential on the 12v line in the power supply.

A Jupiter can't use 1200W.  The DC regulators are only good for 320A max (and running them at max current 24/7/365 is a good way to destroy them).  So lets assume 80% duty cycle.  320A * 0.875V * 80% = 224W per board or 896W total.  Not even sure the chip can handle that kind of power but even if it could the DC regulators can't provide more than that.  

Before buying a second PSU and a 1600W "bonus" one have you spend $20 on a kill-a-watt to see what kind of power you are pulling from the wall.  If you are pulling 500W from the wall then buying 2400W+ of power supplies for each rig is going to do absolutely nothing expect waste money.

Like, those power supplies are designed to go into a computer, closed.  One doesn't have access to the plugs and even then the motherboard stays plugged in.

Or here's another consideration.  You can win a bet by showing how a capacitor has voltage memory, charge a cap, short it's contacts and bet that there is still voltage, take the short away, wait a few seconds, measure voltage.  

User shuts off the power supply, UNPLUGS the miner, say to take the top off - leaving the supply unplugged.  Remember, in its proper application, the motherboard stays plugged in.  So, the supply is unplugged and no current flow.  The voltage memory of the caps brings back a potential but no current flow there is no effective voltage control.  The voltage might get very high.  Plug the connector in and BAM.  We use to induce magnetic fields into slugs by building up charge in capacitors then discharging the caps thru coils around the slugs that then became permanent magnets.  That use to be one hell of a ZAP.

(pssst - it's a joke/irony/sarcasm that they haven't given us any updates about anything)

Thanks Avenger, had been following the thread closely but skipped the last page.

Upon further examination though, it is blank (the quote)

In line with the general feeling in the topic towards KNC*




*No hard feelings guys, just trying to lighten up the mood a little bit  

LOL! Who created that deserves a reward.

PS:

All info we have is posted here:
https://bitcointalksearch.org/topic/m.3319431

In my books, they have until the end of October to deliver all the October units. October 15th is again an implicit perversion of the "November 15th minus 1 month" assumption.

500MH/s

lolololol!!

aka  the KNC bork bork bork eruptor!

In line with the general feeling in the topic towards KNC*




*No hard feelings guys, just trying to lighten up the mood a little bit  

Could the blown capacitors have been the result of users experimenting, trying to get higher hashrates by unplugging the power supply live from the miner then inserting the live plug?  They might not tell if they did.

Probably multiple workers in hosting.  If due to ease of access a worker(s) did the same and with the same result, could they have kept it secret so as not to look like an idiot(s) before the other workers?

I have still have a day 2 low 400 order for hosting still not hashing. 11 days past the deadline. Very disappointed. 😞

My guess is that (KnC) under specified on the power requirements by saying 860 watts was enough to power a Jupiter.
These Jupiters can easily use more than 1000 watts, probably more like 1200 watts.  Even if then can upgrade firmware to be more efficient, at moment each of the 4 cards on my Jupiter is drawing between 285 watts and 305 watts. In total that's 1150 watts to 1220 watts so in reality you would need minimum 1350 watt power supply, but I would go a 1500 - 1600 watt just to be safe.  Even if they state they could make it more efficient you still need the head room to be safe for testing out past, present and future firmwares.

At moment I can't run all cards at once as my 860 watt power supply over loads and shut down, but I will be more sure hopefully tomorrow after I get another 860 watt power supply to add with the existing one or I may buy a 1500 - 1600 watt to run Jupiter on its own if I can find one tomorrow. 

Again, this is my initial thoughts and I'll know more next couple days as I do more testing.  Also just guessing here but this may explain problems at the data center because if you can remember they bought hundreds of the 860 watt Corsair power supplies and they must be experiencing similar problems.

I don't think it's a huge problem, as all is required is more power and after power problems are worked out then they can focus on driver/software problems to fix hardware errors etc. It wouldn't make sense to do it the other way round.

I'm not sure if the amount of VRM's has anything to do with it, as my cards have 4 VRM's not 8.

it's really too bad you guys are having to go thru all this trial and error to make up for what KNC should've taken care of.

in HF's case, they hired the water cooling company itself to do heat simulations using all the Babyjet parts except for the actual chip itself.  they used a heating element to put out 250 W of nominal heat production to make sure the cooler could handle that first.  then they took the wattage up to 350 W of heat to make sure the cooler could handle that amount at what is meant to be an overclocked situation.  no problems according to the results from the water cooling company.  they may have even taken it higher but don't quote me on that.  i won't bother to tell you what hashing rate that took the chip up to b/c HF doesn't want to overpromise but it would knock your socks off.