Topic: Hacking KNC Titan / Jupiter / Neptune miners back to life. Why not? - page 9. (Read 76589 times)

ill contact some people who got some broken neptunes, might be able to get broken api boards aswell.

Crap. Well, removed all the signal chips (z17 and TPS ones) and the thing does the same thing. Which  means that one of the supplies is dead shorted from +12 to the SCL bus, probably the telemetry chip. And since they're all on the same line no way to identify which one.

That sucks. Will think about it more tomorrow.

Meantime I am working on these other two dead boards: They're odd. They don't have the usual shorts on pins 4 and 6, but they have something different.

In these cases, they have 1k resistance on pin 4, open circuit on 6. That is odd. More odd is that when I plug this board into a controller with no 12v it does not show up (but doesn't shut down the controller). Power up 12v and the controller shuts down. So there is something amis on the 12v line on the telemetry circuits. Which makes no sense as 12v never goes to telemetry, that's all powered by the controller. Odd.

The second board had 30 ohms to ground on pin 8, a show stopper and nothing on pin 6, same oddity. Checked the board, shut down the controller even with no 12v. First thing was to figure out pin 8, looking at the 4 gateway 0ohm shunts around the chip I noticed that one of them (pin 8 is a signal line from the titan chip controllers, sda line) had a resistance of only 6.8 ohms to ground. THAT means that chip has a short between it's SCL line and the hashing engines. Fortunately you can isolate that die so I did. Result is a board that now doesn't crowbar the controller with no +12 and does with.

Odd, similar errors. No idea quite why. Will continue to research and see why applying 12v will shut down the controller. Maybe there is a common element somewhere....

C

They are quite similar, except of course that Neppies do SHA and Titans do scrypt. But a lot of the support tech is similar, I use Neptunes for parts and ways to trace out circuit paths.

are the titans similiar? i dont have titans, might be able to get more  half dead neptunes to dismantle, i have bearly seen titans in sweden knc using them all 

Another two titans, exceptionally badly burned power plugs. Fortunately I was able to get them off and new ones on. Back to full hashing and back to the sender.

A note: These both had the original thermal paste, and I got a chance to see it in action: It's like a rock, which is why as long as the cube isn't disturbed it can provide thermal connection to the dies. But if that heat sink moves the thermal paste shatters like ceramic, and you have little heat transfer. Likewise the lids put some down pressure on the heat sink, so always run them with the lids on.

Also: One of the heat sinks had copper pipes at the bottom that were not coplanar. In order words one of the tubes was lower than the other three, that would cause wicked overheating. Threw it out and used one of my neptune spares, but remember to check that; the bottom of the heat sink has to be perfectly flat.

Back to work on some controllers.

More stuff came in. Two badly burned and dead Titan boards, managed to get one running with 3 dies at full speed, the other had 2 dies recoverable. Blackened on the back, keep an eye on the heat levels on these things.

And another 4 cubes. All were dead, not responding, got two running at full power (65mh@42C, could go higher but that's good) and the other two are still dead. Trying to figure out what's up with those, but two fixed isn't bad for starters.

Never dull.

C

the Neptune controller and the Titan controller are the same card, interchangeable

BBB boots and loads faster than PI because it has some type of internal HD

BTW I have a bunch of spare Neptune/titan control cards,  BBB and LCD displays. if anyone needs some let me know.

In a nutshell, yes. Not sure if you recompiled everything for the firmware upgrade, but I had to do a complete recompile of the code specifically for BBB, then put in a number of hacks and symbolic links in order to map the bbb SPI ports and devices to what a Pi expects (the code was written to run on Pi, incomplete for BBB).

The result is a Neptune controller board with bbb to boot off SD card and fire up my code fixes so it can run a Titan. Technically it can recognize NE and JU type devices, but I didn't bother with hacking a universal one that would run neptune and titans on the same controller because it would have required 2 copies of bfgminer and is quite frankly insanely silly.

Also the light is blue when it comes up so you know it is my code.

C

Does this firmware allow bbb and a neptune controller to run a Titan? I ask because I think I have a customer who claims he is doing just that due to his original Titan controller dieing. He purchased my firmware, he said it "installed fine" ... I have no idea if all of the features of my firmware work or not which I told him. LOL! Im very interested to see if my firmware upgrade works "out of the box" on this sorta set up...
I personally, have no way to test this scenario.

You said it "runs faster" in reference to the bbb.... what do you mean by faster, I wouldnt think the Titan would hash faster. Maybe the OS runs faster?

Yup, rock solid and still faster then the Pi. They really muist not have been able to source many bbb at the time of the titan. Probably due to the neptune

ohhhh are you trying to run Titans or other cubes on the uploaded code we are talking about?

One thing I noticed a few days ago was that none of the pages were working. I was wondering what the hell was going on, then realized I forgot to turn on javascript on the page....

Everything worked fine. I think the process could theoretically hang, but the miners would keep mining and a simple reboot would fix it. So far though it's been dead solid reliable on the Neptune hardware.

C

I will play with those after my vacation i didnt have chance to play with it longer. All my cubes run perfectly fine. I bought 8 saturns/jupiters to tinker with controllers so im not in hurry, those will be my backups or to split sets or i will just sell them.

Meantime I fixed a couple of more miners, getting better at fixing the burned boards with one die working on them. Interesting.

No, afraid I haven't seen those. Are they FPGA based miners by chance?

lightfoot - this is a long shot, but have to happened to come across any Cairnsmore1 boards or Ztex that you have nursed back to health? If so I would be interested in buying a few for another non-mining purpose.

Mine has been running solid since the .img file was uploaded. I truncated the largest unformatted space on it. I ended up with about 4.2gb of space taken up on a sd card. I burnt it to the sd card using win32 disk manager and started the controller with cubes i knew worked in it and it hasn't failed me yet. I can do all the things i used to on the original. ssh in and bring up the miner screen ect ect.

Maybe try reflashing it to the sd card or even truncating the largest unformated space on the lightfoots original .img file again.

Also try just using one cube at a time. I say that because i have a cube or two that i plug in and it basically wrecks the whole system. Nothing posts at all however, if i unplug it from the controller or from the cube directly (the controller cable not the power) everything works fine again. (did you find a solution or cause to that lightfoot)

Hm. I wonder. Advanced page should bring up the various cubes unless something is stuck in the cube, try a cold reboot.

C

It kinda works, clock is set by default on 225mhz i think, but i cannot load advanced page to change those. So far i only tried on one cube and it did start but after couple restarts. Tested on Saturn controller dunno if theres any difference between that one and Neptune's or Jupiter's.