Topic: Hacking KNC Titan / Jupiter / Neptune miners back to life. Why not? - page 8. (Read 76589 times)

I know. I just fixed three more today, the Post Office is going to send me a Christmas card or something.

But these are problems I can deal with. Keep them running at about 1,000mhz (~70mh) total and they run pretty happily.

Heh, who'd a thunk in Jan/Feb 2017 at $4.15 LTC or whatever KNC Titans now at 50 bucks an LTC would be making below on

www.litecoinpool.org (used the calc for that site ..no fee 105% payout)

www.litecoinpool.org/calc

profits at $50 LTC at 10c kwh and current difficulty w/o elec considered, a 300mh KNC Titan with Gen Tarkin's Energy Saver setup for 1,250 watts.


24 hours   0.63360352 LTC   31.68 USD   2.94 USD   28.74 USD
7 days   4.43522461 LTC   221.76 USD   20.58 USD   201.18 USD
30 days   19.00810547 LTC   950.41 USD   88.20 USD   862.21 USD

sheesh....you are probably gonna get more busy lightfoot

Another day, more work. Titans keep on coming in, I've gotten to the point where swapping the plugs is quite do-able, and even the shorts on pin 8 (once a sure dead cube) are fix-able. Glad to see they keep on running, if you need something fixed let me know!

Possibly, however they do seem to blow up a lot and take things out. Haven't seen a BBB fail yet. And with the glut of Neptunes out there it makes sense to take a bbb and run my code on it (free. whee free and all that).

But running a Pi takes forever. I'd consider porting the code to new models of Pi's but that would be a lot of work for a solution that's still slower than a BB.

C

Knc techs at the time said a bbb cost 40% more also.

Update: The latest version of my beagleboneblack code is uploaded and now just needs a 4gb SD card or bigger to run:

https://www.dropbox.com/s/vfks74mxqlrfyf9/knc-lightfoot%204gb%20version.7z?dl=0

I also updated the link on post #2, that is where I keep all my links and hints.

HAHA, good on you mate. You are doing a great thing. Keep it up.

At last. Everything has been fixed, all the little boxes of repaired miners have gone out USPS Priority Mail, and I can take a break. Figured out how to fix MAT miner controllers as well; they run a version of Ubuntu that is similar to the installs for Beaglebones. So with a bit of work it's possible to get those un-wedged as well.

If you have anything broken, let me know. Should be a more quiet weekend, maybe I'll mow the lawn and things...

Fixed another 7 boards. This came from a guy that had heat sinks on all of the power supplies, the sinks do allow more power to be pulled but in this case he literally baked off the high side FETs on the supplies. Fortunately he included a bag of supplies, however failures included shorted supply controller chips, and just a general 2 day mess repairing the boards. However I did get 7 fixed, one was fine, one was totally shot.

Today I am working on another three that look like rejects from someone else's attempt to fix boards. Hint: A soldering iron alone will not take off a power supply and ripping up the vias is not impressive. I'm focusing on one board; the drivers were blown and one of the supplies shorted the control pin to 12v. This blew out every single supply on the board, I have pulled up and replaced five of them so far. This is, to be honest, insane.

Oh well. Forunately I have a supply of Titan boards to raid for supplies. Still a mess...

C

So more repairs coming in and I'm seeing something interesting on a few of the cubes.

Apparently someone else is trying to replace the power plugs. That's great, but these plugs are burning out as well and after looking at it I see the problem:

It looks like they are being taken off with a very hot iron without pre-heat. Do-able perhaps, but the problem is that much heat on the bottom is de-laminating the lowest layer of the board; breaking off the pads and causing the lower plane to be powered off the top plane through the vias. Not good, more heat.

More of an issue; the new plugs are being soldered on without warming the whole board, so the solder joints wind up being cold-soldered in the midde of the board. This causes resistance which is causing a lot of heat and baking the pins.

I recommend snipping the neutral pins first, then warming the whole board to around 150c as measured at the board with an IR thermometer first, then tinning the leads with rosin/lead solder, then removing the pins one at a time. With the top three off the bottom three can come off as a set when you tin them (which lowers the melting point of the solder from the very high ROHS levels to a more rational level) at which point you can remove the solder with wicking, test fit the new connector, then put some flux on the new connector pins, put it in, take it off the pre-heater and quickly solder the pins from the bottom. Then put it back on pre-heat, and check to make sure all six pins have solder when viewed from the top.

I can post pics if anyone is interested, but this is important. If the ground leads ever fail or go high resistance the board is destroyed by the resulting voltage differential between the chip hotel supplies from the ground on the controller board and the current demands from the hashing engines. It wrecks the chip and is the only thing so far I can't repair.

C

Noticed there seems to be a few people on this thread who may be interested in some replacement parts for the KNC Neptune..  I have a few fully functional cubes for sale if anyone here is interested!  Check the thread here...  https://bitcointalksearch.org/topic/knc-neptune-replacement-cubes-1923994

Thanks for looking!
J

Meantime back to cooking over a hot oven. Or rework rig. Whatever. :-)

Two more cubes in for repair, looking at them with interest. One had the normal blown drivers, fixed, testing at low speeds tonight for errors, then will ramp it up tomorrow. The second board is odd however. It doesn't crash a Titan controller board with Pi (man these things are so much slower than BB's) but it does fail to respond to PSU Spi requests with the power bus off. Might be a damaged supply screwing up the SPI bus, might be something else. Will review.

I've dealt with the owner of that outlet and i have to say he is a shifty man indeed. I would not trust anything that comes out of that website, not even the time. Good laugh tho yeah.

I'm referring to cryptocoins news

Saw that. Oddly enough the Titan motherboard with some small fixes would be a pretty good place to build a 800mh per cube sort of device. But that's probably not going to happen.

Ok, not directly about fixing Titans, but I figured this group would get a nice laugh out of this:
 https://webcache.googleusercontent.com/search?q=cache:MkMLMXxZkQEJ:https://www.cryptocoinsnews.com/knc-miner-returns-improved-mining-product-sale/+&cd=2&hl=en&ct=clnk&gl=us

Disappeared pretty quick, seems familiar...

I have a bunch of neptune controllers if anyone needs them. 

Boomin1

msg me

Ok, got all four of these cubes working. All were totally blown, all now fixed. The last one I got three full dies out of four, turns out power supplies 2,3 were both shorted, had to remove them both. I'm a bit better at replacing power supplies these days, but my belief is if you actually get to the point where you have 3 dies running at full power, quit while you are ahead :-)

200 watts per cube is a fully running cube. :-) Now to pack these up and work on the next batch. I'm still taking repairs, getting better every day...

C

Don't know. People have posted links in the thread in the past, read a few back and if you find it let me know; I'll put it in post 2.

Or just run them a bit slower. Overheating the supplies can cause a supply to fail shorted to the SCL bus, bit pain to find but I'm working on it.

C

Meantime busy here. This one cube which was totally dead has been interesting: Biggest problem was a short on pin 8, I found that die 2 was shorting the lines to the power supplies, so I pulled the jumpers. That allowed the board to be seen with power off, but not on. Which led me to think it was a problem with a power supply.

So I pulled the two supplies that ran die 2. Still no go. Then tried pulling the supply 1 (on die 0, the one closest to the front). That got the board to respond on the SPI lines, which means it is a short in a supply that can kill the board. Pulled the supply, fired it up, now hashing at 30mh with one fully working die at 325, one at 150 (running on the remaining supply) with a third die down for unknown reasons.

What a pain. Will see if this works for the other board.

where i can found inc cable splitter
i need 4 of these

have anyone have??