Hacking KNC Titan / Jupiter / Neptune miners back to life. Why not?

[lightfoot]

Truth is the signal lines that short are on the outside edge of the CPU chip, right along the left and right sides. If I could jack the chip up a mm, I could get under it with some desoldering braid and pull off the solder around them. I would probably short the die's + and - supplies but who really gives a damn if the die is dead.

Trying this on some Neptune boards just to see what happens....

[1714790023]

1714790023

[1714790023]

1714790023

[1714790023]

1714790023

[Searing]

Your efforts to fix Titans are commendable...however you may be moving into the below territory (see meme)......just use a bit of caution!

(Those KNC products are 'dangerous' and the company as well....focus too much more attention on this and this could be your equivalent of the dreaded 'stapler obsession" ....)

Also this comes from Searing who you have a Titan Board with the very problem you have on hand yet you can't fix...

So do I 'sacrifice' your sanity by pushing you down this road to get a fully working cube back? Or man up and warn you off?

I am conflicted

Anyway, don't be this guy! (Come to think of it you probably are this guy...if so your efforts make a lot of frigging sense)

Remember folks...fixing POW miner equipment can effect your social life and shorten your life. (Evil they are...especially KNC and their design..made to drive techs crazy)




Crap..he has my address from my last cube fix....crap ..he is gonna track me down ..take my stapler and burn my house down

[dltrading]

hello steve, i tried disconnect all cubes, but the board present same error, in the lcd display kncspimux, i tried reflash and present same error, i tried downgrade and present same error,
the error in cgminer  say, devices not found,  whith one cube or two otr three or for connected same error, look like the board is broken in any place or anything else.

Ok. Questions:

1) Are you running the latest 2.00 Titan code from KNC? If not run it.
2) What does the light on the controller board do? It should flash white for a few seconds, then go off, then turn on green if you are running 2.00 code. If it flashes white quickly, then turns off then flashes white again the controller has a shorted FPGA. If no lights at all the FPGA and TMS chips may be bad.

The gold standard test is to ssh into the titan controller with username pi password whatevertheadminpasswordis, then type

io-pwr init

And watch the light. If it comes on green and there are no errors in your ssh session you're good. If you get errors, post them.

C

[lightfoot]

Didn't see your reply dltrader, just a quote.

Meantime found a new and interesting failure: This controller board that just came in was doing the flashing white light, but instead of a short flash it was a longer one, which was odd. Hooked up a Neptune BBB to it and did some diagnostics. Sure enough the TPS chip was coming up but failing on PGOOD timeout, which was new.

So I tried removing the display board. Controller comes up, all is well. Oddly enough it looks like the display SPI driver circuit is bad; that's a new one to me, normally the displays just go dead without shutting down the board. Repair averted, board is fine, sending it back....

On to the next project.

[Unacceptable]

Didn't see your reply dltrader, just a quote.

Meantime found a new and interesting failure: This controller board that just came in was doing the flashing white light, but instead of a short flash it was a longer one, which was odd. Hooked up a Neptune BBB to it and did some diagnostics. Sure enough the TPS chip was coming up but failing on PGOOD timeout, which was new.

So I tried removing the display board. Controller comes up, all is well. Oddly enough it looks like the display SPI driver circuit is bad; that's a new one to me, normally the displays just go dead without shutting down the board. Repair averted, board is fine, sending it back....

On to the next project.

[lightfoot]

Honestly this is as much of a keep my skills/brain in shape sort of thing as it is a business. There's no way I could recoup the time spent trying to figure some of this stuff out and I've been fortunate enough to have a platform that actually has value over time. Trying to learn how to fix an Antminer is kind of pointless; they go obsolete within a few months of release, not worth the effort.

I can fix GPUs, but there's not enough of a market in that. Thought about the Innos, but they seem to be a slightly odd bunch and I'll deal with that if it happens.

C

[lightfoot]

3 more controllers came in today. So far one's fixed with my inventory FPGA, two more tomorrow so I should be able to get them out on Tuesday.

In terms of what blows up FPGAs, one item is if you plug in the 10 pin connector to the controller board, then mis-place it on the Titan so it is only connecting on five pins. That blows up the controller (shorts all the FPGA driver lines, no buffers, sad), but also should blow up the cube in such a way that it appears on the console (once you replace that) but never hashes on any dies (internal gates blown)

I can fix that but I see less of those than blown controllers.

[lightfoot]

Busy weekend. Fixed four more Titan boards that came in, and parts on the way for two controller boards.

Did notice something very interesting: One of the boards was reported by the user as running "wild" and shutting down after a minute of hashing. I put the board on my jig and noticed that the heat sink was not properly engaging, even with the screws all the way down. Very odd. Measured the height of the top of the chip with my micrometers.

This board was 1mm lower than others. No evidence of bending, just a bit lower.

So in order for the sink to mate properly I had to put 1mm of washers under the screws. This is the ONLY time I would recommend this, but you can see if your heat sink is on properly by seeing if you can wiggle it when the screws are down. If it wiggles, it's way too lose. If not run the board at 100mhz, then 250 and watch the temps. If you go into the upper 40's then the sink probably is not mating properly.

Advice for the newcomers.

[boomin]

Just a good spot to say that if anyone is looking for Titan stuff....

https://bitcointalk.org/index.php?topic=1757548.0

thanks,

boomin

[lightfoot]

Ok I have run into a wall and need some help:

I got a Titan controller in from a guy in England, he bought the controller and cubes, and apparently the Pi was making "fizzing" sounds and the controller/cubes didn't work. Review here found the FPGA blown and the TPS power chip literally *fused* to the board. Likewise his bridgeboard is blown out as well, looks like the crystal is bad.

Since we do not know what the crystal is for these (and I have tried half a dozen different possibilities) I need to find a bridgeboard for this guy. If you have one that shorted out, I can fix it for him and sell him my spare board for the cost of a normal repair. However without a bridgeboard he's up a creek.

Anyone have a dead one that they replaced with a Qberty 1.0 board? I'll do a free repair for one because I want to help this miner out.

C

[Searing]

Ok I have run into a wall and need some help:

I got a Titan controller in from a guy in England, he bought the controller and cubes, and apparently the Pi was making "fizzing" sounds and the controller/cubes didn't work. Review here found the FPGA blown and the TPS power chip literally *fused* to the board. Likewise his bridgeboard is blown out as well, looks like the crystal is bad.

Since we do not know what the crystal is for these (and I have tried half a dozen different possibilities) I need to find a bridgeboard for this guy. If you have one that shorted out, I can fix it for him and sell him my spare board for the cost of a normal repair. However without a bridgeboard he's up a creek.

Anyone have a dead one that they replaced with a Qberty 1.0 board? I'll do a free repair for one because I want to help this miner out.

C

still trying to get you one blown bridgeboard it is coming from a buddy..but he found like 3 cubes he wants to add etc etc.....but thay are coming
I keep plugging away at getting them to you. (also only way my qberty bridge boards can ever be fixed is the slim chance a knc crystal will tell you something)

then again with LTC difficulty surging like 34% or some such in the last month..not exactly sure why I am bothering with this

dang bitmain L3's anybody else I could ease these knc titan's out into a 'soft landing' but bitmain is too damn big..they are gonna pop them puppies out like well......puppies

Hell at this rate I'm out in march 2017...well they ROI'd and it has been 2years 3 months by march... then ....I should not complain.....(but 'my precious' ack!)

[lightfoot]

You got it, it's here, and I fixed one of his Titan boards. Now I need something:

Does ANYONE have a Qberty 1.0 batch board that works? If so take the highest rev picture you can of the crystal and post it here. Got a theory.....

C

[Searing]

You got it, it's here, and I fixed one of his Titan boards. Now I need something:

Does ANYONE have a Qberty 1.0 batch board that works? If so take the highest rev picture you can of the crystal and post it here. Got a theory.....

C

Sent you a pm but it may take awhile .working 80hr weeks with OT and it is one of my 4 titan boards in the basement....may take me a bit to find out which
one .only way to tell is take each unit (titan) and shut it down ..then take the controller apart and look

you know how well that worked with the pins I sheared off the titan 2nd batch bridge I sent you...so not to thrilled with the idea 1 out of 3 on that attempt messed it up
(the clone bridge I sent you that did not work and I had to take off controller card) ...I'd send the whole card and use another but again can't tell which is the 1st batch
working qberty card unless i pull the bridge (dangerous)

qberty you'd think would be able to send you a pic I'd think, if he wants to contact us on here anymore .try him....

I'll get to it when I can ..but not the kinda thing I want to try to do in 2-3hrs before sleeping before working 0T again (and yes it is 7 days a week no break)

will do this as long as I can ...make hay while the sun shines Puts a crimp in my hobby stuff thou. Good thing they don't care what I do on the laptop at work

as I type this thou.

anyway sent you a pm

brad

[LongAndShort]

I always just figured the crystal on the bridge board was because it had to be the same freq as what the bbb spoke to the controller board. I really wish we could just get the titan firmware working on the bbb

[lightfoot]

Ok, here is what needs to be done to get Titan code to work on a BBB. Enjoy my raw notes and comments.

Note: It still doesn't work, there is something you need to do with bfgminer's configure to make Titans go. But this should get the whole thing up to the point where you can run a neptune with sha256 so all communications are working.

Note: I probably missed a few things like putting that ioboard rbf code in etc. Do that of course, the titan rbf code will work with a titan or a neptune.

Start with a clean BBB Debian install.
Create a user called pi, home directory /home/pi
Download the knc-asic directory from github's KNC project
Download the bfgminer code from github's KNC-asic
Download the spi-test directory from github's KNC-asic
Download the web code from github's knc-asic

do a make beaglebone in the knc-asic directory
do a make beaglebone in the bfgminer/knc-asic directory.

#!/bin/sh -xe
sudo ln -sf $PWD/raspi-blacklist.conf /etc/modprobe.d/raspi-blacklist.conf
sudo ln -sf $PWD/modules /etc/modules
sudo ln -sf $PWD/../program-fpga /usr/bin/
sudo ln -sf $PWD/../lcd-message /usr/bin/
sudo ln -sf $PWD/../asic /usr/bin/
sudo ln -sf $PWD/../waas/waas /usr/bin/
sudo ln -sf $PWD/../knc-serial /usr/bin
sudo ln -sf $PWD/../knc-led /usr/bin
sudo ln -sf $PWD/../io-pwr /usr/bin
sudo ln -sf $PWD/../../spi-test/spi-test /usr/bin/
sudo ln -sf $PWD/board-init.sh /usr/bin/
sudo ln -sf $PWD/factory_setup /usr/bin/
sudo ln -sf $PWD/default-config.sh /etc/init.d/
sudo ln -sf $PWD/ioboard.sh /etc/init.d/
sudo ln -sf $PWD/factory-setup.sh /etc/init.d/
sudo ln -sf $PWD/bfgminer.sh /etc/init.d/
sudo ln -sf $PWD/repartition.sh /etc/init.d/
sudo ln -sf $PWD/lcd-loop.sh /etc/init.d/
sudo ln -sf $PWD/lcd-loop /usr/bin/
sudo ln -sf $PWD/get_asic_stats.awk /usr/bin/
sudo ln -sf $PWD/monitordcdc.ericsson /sbin/
sudo ln -sf $PWD/../spimux-titan.rbf /etc/spimux.rbf
sudo ln -sf $PWD/i2c-loop /usr/bin/
sudo ln -sf $PWD/lcd-print_stopping.sh /etc/init.d/lcd-print_stopping.sh
sudo ln -sf $PWD/lcd-print_rebooting.sh /etc/init.d/lcd-print_rebooting.sh
sudo ln -sf $PWD/asic_data_cache.awk /usr/bin/

--I think all of these are done.


sudo cp inittab /etc/
sudo cp $PWD/monitordcdc.logrotate /etc/logrotate.d/monitordcdc

--Done as well

sudo update-rc.d default-config.sh start 11 S .
sudo update-rc.d ioboard.sh defaults
sudo update-rc.d factory-setup.sh defaults
sudo update-rc.d bfgminer.sh defaults
sudo update-rc.d repartition.sh start 11 S .
sudo update-rc.d lcd-loop.sh defaults
sudo update-rc.d lcd-print_stopping.sh stop 02 0
sudo update-rc.d lcd-print_rebooting.sh stop 02 6

Need to recompile bfgminer
Don't forget to compile it with --enable-script....
need to install uthash-dev
checking for JANSSON... no need jansson-dev, libi2c-dev

Ok to get apache going.
Link the www files to /var/www
Link the cgi-bin files to /usr/lib/cgi-bin

Only thing that doesn't work is seeing the cube. Something is still stuck there. Fuck it, use lighthttp--fixed


Ok, stats don't work. They are in /var/run
asic_stat_file=/var/run/stats.knc.$$
/home/pi/knc-asic/RPi_system/get_asic_stats.awk /var/run/.waas_cache >$asic_stat_file 2>/dev/null

./bfgminer --api-listen --api-allow="W:127.0.0.1/24" -c /config/cgminer.conf -Q 10 --scrypt -S titan:auto

Ok, now I see the /var/run/revision file is not updating at all. Manually putting it there fixes the web sort of but it's
supposed to be coming from waas

Hm. In Waas.c I see

#define DEFAULT_CONFIG_FILE     "/config/advanced.conf"
#define WAAS_CURRENT_FREQ       "/var/run/.waas_curfreq"--got this
#define WAAS_AUTOPLL_TIMESTAMP  "/var/run/.waas_t1"--not exist
#ifdef CONTROLLER_BOARD_RPI
#define EXPECTED_PERFORMANCE_FILE       "/var/run/expected_performance"--missing, exists on bb in /etc
#define REVISION_FILE                   "/var/run/revision"--missing
#else
#define EXPECTED_PERFORMANCE_FILE       "/etc/expected_performance"
#define REVISION_FILE                   "/etc/revision"--AH, it's looking here in /etc/revision.....
#endif

AH HA! THE FUCKING FILES ARE IN THE WRONG PLACE. Put links into /etc/expected_performance and /etc/revision to /var/run.
Dumbass system.


More annoying, it needs to be recompiled to fix this. Because the stupid links are not recreated.
Sigh.
Oh for fuck's SAKE! there is another KNC directory under the bfgminer that ALSO NEEDS TO BE RECOMPILED TO BEAGLEBONE!

Meantime you need to compile bfgminer with titan and kncasic. knc blows it out. But it still don't hash. Close...
 ln -s /etc/expected_performance /var/run/expected_performance
 ln -s /etc/revision /var/run/revision

WHAT A FUCKING JOKE!

[LongAndShort]

Wow! Looks like you had a painful time with that, like pulling teeth?!

[lightfoot]

Wow! Looks like you had a painful time with that, like pulling teeth?!

Indeed. Try something, doesn't work, try again. Keeps me busy wish we had things like documentation...

[lightfoot]

So in the meantime I sat down over the weekend and finally traced out the shorted chip problem. This is the one where when you plug a cube into a controller it shuts off the green light and crashes the controller. This is caused by a short in one of the die hotel circuit supplies and I have been hoping that cutting the lines to that die would clear the fault.

What I did was get the chip hot, then slide under the chip desoldering braid to remove the solder from the quarter of the chip that is shorted. Took awhile to do but I cleared out all the solder around the hotel circuits on the die that was shorted.

Board however was still dead. So I lifted the chip carefully and found when the short would go away. In a nutshell the hotel power circuits are connected together INSIDE the chip itself, so when one die goes, they all go.

Ergo, it is not possible to fix this type of fault. Drat. Can still fix just about everything else, but not that....

C

[LongAndShort]

compiling bfgminer should probably look something like this:

Code:

./configure --enable-scrypt --disable-sha256d --enable-titan --disable-other-drivers

[lightfoot]

Yup, didn't quite work but pls try to do it.

C