Author Topic: NEW KNC TITAN FIRMWARE - 'GenTarkin's Mod' OFFICIAL TOPIC  (Read 15824 times)

Offline GenTarkin

  • Jr. Member
  • **
  • Posts: 94
NEW KNC TITAN FIRMWARE - 'GenTarkin's Mod' OFFICIAL TOPIC
« on: September 09, 2015, 07:12:45 PM »
UPDATED 3.3.16 - RELEASE of GenTarkin's Mod ~ Custom KNC Titan Firmware - v1.0.0
------------------------

wait for it.....    Wait for It............  WAIT FOR IT......................

RELEASE of GenTarkin's Mod ~ Custom KNC Titan Firmware v1.0.0 IS FINALLY HERE!!!!!!!!!!!!

Previous customers: head to http://gentarkincustomtitan.pcriot.com/custom-knc-titan-firmware-myaccount/ to obtain your FREE UPGRADE and license key(s)!!

New / Unverified previous customers: head to http://gentarkincustomtitan.pcriot.com/custom-knc-titan-firmware-shop/ to purchase!!

RELEASE NOTES:

Major Features, Additions, Fixes & Changes:
Added the long awaited “Energy Saver” feature – automatically tunes die voltages for highest effeciency while keeping the die stable as possible & hw% low. ~~MORE INFO http://gentarkincustomtitan.pcriot.com/energy-saver-faq-help/ ~~
DCDC Over-Current protection – any DCDC’s detected above 46A will be issued a die reset, if that fails then die speed is clocked down significantly to prevent DCDC failure.
Long Term Stability checking for all dies – bumps voltage up on a die if it becomes unstable more than 2x in a period of 2 days.
Per RPI Cube Tracking – if cubes are added, removed or moved around their last known clocks and volts on a given pi will be restored.
Internet outtage / dead pool detection to prevent miner from constantly resetting itself.
Slight webgui redesign.
Product license activation.
Full Copyright & end user license information.
Much smaller installation file.


Additional Features, Additions, Fixes & Changes:
Full translation of webgui pages to 80+ different languages.
Logging file output format refined and also includes more timestamps now.
Optimized timing of die / cube soft and hard resets.
Fixed format of advanced.conf file not being updated properly when user changes/removes/adds cubes or manually sets dies OFF.
ADVANCED page now shows which dies(if any) are being throttled because of DCDC over temp threshold.
ADVANCED page now auto refreshes every 5 minutes.
Noticed in rare cases dies are configured with positive voltage – wrote correction code for this.
For “Energy Saver” implemented cross reboots internal data array tracking capability.
Ability for user to save Log Files to a file to be posted or emailed for diagnostic purposes. (This is part of the Energy Saver options).
Added BALANCE strategy to MINER page – this may work better than LOAD-BALANCE for some users.
Removal of BFGminer version option – in testing 5.1.0 works best with Titans – can re-add later on as an optional installation file if demand is high enough.
More fixes to KNC’s shoddy code.
And many more small fixes / improvements too numerous to list.

All the previous firmware features and improvements are included!

Now … you’re probably asking yourself “yeah yeah, but how much do all these wonderful new features cost?” Simple: $75 first full copy, $35 for additional licenses.

Verified Previous Customers: The upgrade to this firmware is completely FREE for the number of Titans the previous version was purchased. Additional Titans require additional license purchase. If you created an account here previously then you will find license keys issued to your account. **NOTE** If you have not created an account on this site and verified your previous purchases, please contact me - http://gentarkincustomtitan.pcriot.com/custom-knc-titan-firmware-contact/
--
New Customers: Prices as detailed above.

***IMPORTANT INSTALLATION NOTICE*** – Upon upgrading to and activation of this firmware, its EXTREMELY IMPORTANT to keep track of which license was used on your KNC Titan(s). An easy way to do this is create backups of your Titan(s) after activating, then keep record of which backup went to which Titan.

***PRODUCT NOTICE*** – upon purchasing / obtaining this software, the end user agrees to the license terms set forth here(this file is also contained in my firmware): gentarkincustomtitan.pcriot.com/gentarkin_mod/LICENSE











---------------------------------


Hey all! I recently found out word of my firmware had made it to this forum. I didnt realize this forum even existed till someone informed me! HA!
Anyways about me a bit: Ive been in crypto since mid 2011, its been a hobby of mine ever since in mining. Its opened my eyes to many things & Ive learned a bunch! I modified the good ol bfl single firmware back in the day to work better & faster .. I hadnt done anything on coding in miners since then, till I bought a Titan a few weeks back.
When I bought the titan, it had the classic "sleeping" or "dropping" dies issue which would require babysitting the miner throughout the day and doing a full power cycle to bring those dies back. After looking at the source code KNC put out, and seeing that the DCDC's used were programmable, I had a strong suspicion that if I could issue a power cycling to a Dies respective pair of DCDC's then restart bfgminer ... then the die would come back up and in a reliable manner. Sure enough this proved correct and I implemented this "hard reset" of dies into KNC's original watchdog script.
I announced my findings & success to the community via the KNC thread over at bitcointalk.org ... I had some big donations come in and the project continued to grow and grow w/ more and more ideas to make the Titan more reliable & safe then ever.

So, without further ado.... I give you my custom firmware:
https://github.com/GenTarkin/Titan  - "GenTarkin's Mod"
Code: [Select]
Based on 2.0 firmware for KNC KNCminer Titan, My modificatoins:

*Coded w/ RELIABILITY & SAFETY of the Titan in mind...

New method of "Dead Die" handling - power cycles cube w/ "Dead Die" to attempt to get it working w/o manual power cycling.

Added ability to switch between STOCK 5.1.0 BFGminer version vs newer 5.2.0 ver.

Added DCDC temp monitoring & corrective action of 25mhz dropping on dies till DCDC temps are below threshold specified.

In addition to the above, dies which were previously throttled will raise their clocks once under the threshold temperature setting by 10%.

Added more stats to STATUS page.

Added miner name to not only STATUS but MINING & ADVANCED page.

Added "donation status" to reflect which donation plan ur Titan is running.

Added Mod version info on STATUS page.

Added "Upgrade" button to STATUS page to perform "OTA" updates to my latest releases!!"

Added option under SYSTEM to disable lighttpd access loggin(saves SD card life).

When creating backup, it now saves clocks, voltages & miner name(if specified).

Redesigned all branding graphics & information for webgui.


--------

*NOTES:*
DCDC threshold settings & BFGMiner version is settable in webgui under ADVANCED tab.
Current DCDC mod settings & running BFGMiner version are shown in ADVANCED tab.

Ive implimented 2 payment options, details are below:

1. Upfront payment of $50 for version which has daily donation fee disabled.. see  this post for details & questions:
https://bitcointalk.org/index.php?topic=170332.msg12194406#msg12194406  (vegasguy is handling processing & distro of the img for prepaid version)
TO PURCHASE go here: http://bitcoinlasvegas.net/kncminer-titan-custom-firmware-mod-paid-membership-page/

2. The version available for download as an img here on github in my releases, this has a built in .75% DAILY donation where your Titan mines for me roughly$

NOTE: The fee's / sales are ONLY for my custom modifications to the original firmware. Im not "selling" the original firmware.



DISCLAIMER: In the unforunate even of ur Titan catastrophically failing while my firmware is installed... I am in no way responsible for the damages.

*I have not coded anything harmful to the Titan, Ive largely created "smarter"(more reliable & safer) running condition handling of the Titan.

The firmware can also be purchased on ebay.

Also, I noticed someone mentioned "no speed improvements" the reason for that is, on Titan any form of overclocking beyond 325mhz is a huge risk to the power circuitry. 325mhz already runs the DCDC's 24/7 over their RMS limit of 40Amps. Not to mention the electrical limit of the 6pin PCIE connector @ 12v is around 270W.

Feel free to pm me w/ any questions or purchasing questions =)




As far as code improvements in the FPGA, Im pretty sure the FPGA firmware is closed source, I have nowhere the skills to reverse engineer it. If someone wants to attempt that, go ahead =)
« Last Edit: March 04, 2016, 06:36:36 AM by GenTarkin »

Offline GenTarkin

  • Jr. Member
  • **
  • Posts: 94
Re: NEW KNC TITAN FIRMWARE - 'GenTarkin's Mod' OFFICIAL TOPIC
« Reply #1 on: September 10, 2015, 04:03:56 AM »
An update coming down the pipes for next release....
Fan failure protection w/ webgui output in big red letters when a potential fan failure has been detected!!!!

Up till now, my monitoring script has dynamically adjusted clocks up n down depending on environmental temperatures that increased / decreased gradually. It would attempt to clock the unit down as temps rose but this happened at the rate of KNC's monitoring script which is 1 poll per minute. A fan failure would easily go up in temperature much quicker than decreasing 25mhz ever 1 minute on respective dies would account for.
In the process of coding this I also found fault w/ KNC's waas command not actually being able to set a die OFF, rather bfgminer would have to be restarted and the advanced.conf file held a null value of any desired dies to be set as "OFF".
So... what Ive come up with is a small rewrite of setting dies to OFF logic(now all dies set to off while mining will invoke a bfgminer restart).
Next, onto the fan fail protection conditions...
1. If either of the DCDC temps of any die have detected as being above the user set threshold by 10C or more then the dies will be turn OFF.
OR
2. If either of the DCDC temps of any die have reached 115C or higher then that die will be set OFF.

After all dies which meet the condition above have been set to OFF, bfgminer will be issued a restart command.
A warning in big red letters on the STATUS page of webgui will warn about a possible fan failure detection.

To clear the warning the pi must be rebooted(I figured in the case of a genuine fan failure, the pi will be turned off along w/ the miner to service the cube w/ the failed fan).

*note.... if the 115C threshold ends up being an issue for anyone running their titans this hot well... A: your Titan will not last long B: I can always change this upward, but Im willing only to go up to 120C or so.

I will post when this release will be live.

Offline Mattzees

  • Sr. Member
  • ****
  • Posts: 704
Re: NEW KNC TITAN FIRMWARE - 'GenTarkin's Mod' OFFICIAL TOPIC
« Reply #2 on: September 10, 2015, 08:08:46 PM »
So, for my own benefit, and for others here, can you tell me how to get the log files, and what log files you need to see in order to diagnose/debug what's going with my rig when I use your firmware?

I may need to bypass auto-configuration of some dies, but don't know what dies those are until we look at logs.

Offline GenTarkin

  • Jr. Member
  • **
  • Posts: 94
Re: NEW KNC TITAN FIRMWARE - 'GenTarkin's Mod' OFFICIAL TOPIC
« Reply #3 on: September 10, 2015, 09:12:44 PM »
So, for my own benefit, and for others here, can you tell me how to get the log files, and what log files you need to see in order to diagnose/debug what's going with my rig when I use your firmware?

I may need to bypass auto-configuration of some dies, but don't know what dies those are until we look at logs.

Yeah, I have planned a release that will do smarter tracking of die conditions and if dies get issued resets up to a certain amount of times within a short period of time then they will be auto bypassed for checking.
For now, if you need bypassing of checking of dies then ... you can ssh into the pi, run cat /var/log/monitordcdc.log
It will show you in the ASIC# DIE# lines which dies are constantly being issued resets.
you can then add those dies to the knc-asic/bypass_dies.conf file in the form of
1 2

the above would mean ASIC#1 DIE#2

Offline GenTarkin

  • Jr. Member
  • **
  • Posts: 94
Re: NEW KNC TITAN FIRMWARE - 'GenTarkin's Mod' OFFICIAL TOPIC
« Reply #4 on: September 11, 2015, 12:48:08 AM »
A small preview of this coming release:



This release (v.99b) will be available shortly for those who have paid up front either through ebay or bitcoinlasvegas (site handles my sales as well) or payment to me directly.

The donation version of v.99b will be avail to download as a release when I no longer flag it as a prerelease(probably at least a few days).

Offline Mattzees

  • Sr. Member
  • ****
  • Posts: 704
Re: NEW KNC TITAN FIRMWARE - 'GenTarkin's Mod' OFFICIAL TOPIC
« Reply #5 on: September 11, 2015, 05:42:41 PM »

Yeah, I have planned a release that will do smarter tracking of die conditions and if dies get issued resets up to a certain amount of times within a short period of time then they will be auto bypassed for checking.


You might want to include check boxes on the advanced page to manually bypass certain dies.

Offline kilroy

  • Jr. Member
  • **
  • Posts: 46
Re: NEW KNC TITAN FIRMWARE - 'GenTarkin's Mod' OFFICIAL TOPIC
« Reply #6 on: September 11, 2015, 07:35:05 PM »
I agree about bypassing certain dies with a check box because a bad die can cause a long period of substandard performance on reboot as your approach restarts the whole process so if we know for sure the die is bad and unrecoverable, why disrupt services of good die?  Alternatively if it is possible to just have BFGMiner rescan without restarting, not sure if it has the same effect on the die but seems it would be less intrusive.  Also as mentioned previously, passing temps to BFGMiner is a HUGE +!  I will be getting with you next week for the 10pack licensing, keep up the good work and thanks much for your contributions to the project.


Offline GenTarkin

  • Jr. Member
  • **
  • Posts: 94
Re: NEW KNC TITAN FIRMWARE - 'GenTarkin's Mod' OFFICIAL TOPIC
« Reply #7 on: September 11, 2015, 08:12:36 PM »

Yeah, I have planned a release that will do smarter tracking of die conditions and if dies get issued resets up to a certain amount of times within a short period of time then they will be auto bypassed for checking.


You might want to include check boxes on the advanced page to manually bypass certain dies.

I have thought about this and its more difficult to impliment then it seems. At least in the ways Ive thought about it =./ ... if I can think of a way Im certain of being able to do, I may do this.
Currently there is a way to manually bypass dies, thats by editing the knc-asic/bypass_dies.conf file  .

Offline p4fg

  • Newbie
  • *
  • Posts: 1
Re: NEW KNC TITAN FIRMWARE - 'GenTarkin's Mod' OFFICIAL TOPIC
« Reply #8 on: September 11, 2015, 08:15:54 PM »
Please feel free to add my pimped TITAN-UI (kncminion) to your firmware-build!

http://shellcode.se/programming/introducing-kncminion/

Best regards
p4fg

Offline GenTarkin

  • Jr. Member
  • **
  • Posts: 94
Re: NEW KNC TITAN FIRMWARE - 'GenTarkin's Mod' OFFICIAL TOPIC
« Reply #9 on: September 11, 2015, 08:22:27 PM »
I agree about bypassing certain dies with a check box because a bad die can cause a long period of substandard performance on reboot as your approach restarts the whole process so if we know for sure the die is bad and unrecoverable, why disrupt services of good die?  Alternatively if it is possible to just have BFGMiner rescan without restarting, not sure if it has the same effect on the die but seems it would be less intrusive.  Also as mentioned previously, passing temps to BFGMiner is a HUGE +!  I will be getting with you next week for the 10pack licensing, keep up the good work and thanks much for your contributions to the project.
No, there is no way to simply update bfgminer while its running. If a die requires a "hard reset" (power cycle) then, in my experience, a bfgminer restart was required.

"why disrupt services of good die?" - As it stands ... a "perfectly good die" will not cause any sort of resetting..... So Im kind of confused bout the overall statement .. lol Also, my code never issues reboots, only bfgminer restarts if dies need a "hard reset"(power cycling)

Im currently working on my next release probably v.99c which will have auto bypass code. So far, it works like this, if a die is issued 5 soft resets, each of those resets within 15mins of the previous reset, then the die will be added to a "auto bypass file" ... on the next loop of the script, it will see that the die is listed in the "auto bypass file" and no longer issue any sort of reset command to that die.
I will have a message displaying what dies are currently being bypassed on the advanced page of webgui and a button to issue a "reset bypassed" dies which will wipe out that auto_bypass file.

"Also as mentioned previously, passing temps to BFGMiner is a HUGE +!" ... what does that mean exactly?

Offline GenTarkin

  • Jr. Member
  • **
  • Posts: 94
Re: NEW KNC TITAN FIRMWARE - 'GenTarkin's Mod' OFFICIAL TOPIC
« Reply #10 on: September 11, 2015, 09:25:02 PM »
Please feel free to add my pimped TITAN-UI (kncminion) to your firmware-build!

http://shellcode.se/programming/introducing-kncminion/

Best regards
p4fg

HAHA! Nice job man! Ive installed it and will include it in my next release as well =)
I may even put a link somewhere that points the kncminion page =)

Offline Mattzees

  • Sr. Member
  • ****
  • Posts: 704
Re: NEW KNC TITAN FIRMWARE - 'GenTarkin's Mod' OFFICIAL TOPIC
« Reply #11 on: September 12, 2015, 01:46:58 AM »
Please feel free to add my pimped TITAN-UI (kncminion) to your firmware-build!

http://shellcode.se/programming/introducing-kncminion/

Best regards
p4fg

HAHA! Nice job man! Ive installed it and will include it in my next release as well =)
I may even put a link somewhere that points the kncminion page =)

Yes.  KNCminion is the bomb.  It will be nice to see that added to the UI.


Offline kilroy

  • Jr. Member
  • **
  • Posts: 46
Re: NEW KNC TITAN FIRMWARE - 'GenTarkin's Mod' OFFICIAL TOPIC
« Reply #12 on: September 12, 2015, 07:07:12 PM »
Quote
"why disrupt services of good die?" - As it stands ... a "perfectly good die" will not cause any sort of resetting..... So Im kind of confused bout the overall statement .. lol Also, my code never issues reboots, only bfgminer restarts if dies need a "hard reset"(power cycling)"

Sorry for the confusion and maybe I am missing something but here is my example, This Titan I am testing hashes avg over 300MH steady but has a bad die or two.  It seems like when I run the .99a code it goes through about 1000 shares or so and then does a BFGMiner reset (because of bad die) and then starts back up again, does this over and over.  My avg hashrate up until the BFG reset is around 50MH lower (affecting good die performance) and that doesnt count the time lost in restarting each time.  Of course eventually it stops resetting and runs at the expected hashrate.  So the ability to disable die I know are bad would mean they would not cause my good die to under-perform for 1.x hours every time I reboot.

For now I can just manually configure in the conf file you pointed out if it becomes too much of an issue.

To me the temps for BFG is probably the biggest value as I manage multiple ASIC through Awesome Miner and really prefer to manage/monitor from a centralized interface like this.  Your temp/clock regulator definitely at least prevents catastrophe where KNC's code pretty much left this high dollar equipment quite vulnerable.






Offline GenTarkin

  • Jr. Member
  • **
  • Posts: 94
Re: NEW KNC TITAN FIRMWARE - 'GenTarkin's Mod' OFFICIAL TOPIC
« Reply #13 on: September 13, 2015, 01:39:59 AM »
Quote
"why disrupt services of good die?" - As it stands ... a "perfectly good die" will not cause any sort of resetting..... So Im kind of confused bout the overall statement .. lol Also, my code never issues reboots, only bfgminer restarts if dies need a "hard reset"(power cycling)"

Sorry for the confusion and maybe I am missing something but here is my example, This Titan I am testing hashes avg over 300MH steady but has a bad die or two.  It seems like when I run the .99a code it goes through about 1000 shares or so and then does a BFGMiner reset (because of bad die) and then starts back up again, does this over and over.  My avg hashrate up until the BFG reset is around 50MH lower (affecting good die performance) and that doesnt count the time lost in restarting each time.  Of course eventually it stops resetting and runs at the expected hashrate.  So the ability to disable die I know are bad would mean they would not cause my good die to under-perform for 1.x hours every time I reboot.

For now I can just manually configure in the conf file you pointed out if it becomes too much of an issue.

To me the temps for BFG is probably the biggest value as I manage multiple ASIC through Awesome Miner and really prefer to manage/monitor from a centralized interface like this.  Your temp/clock regulator definitely at least prevents catastrophe where KNC's code pretty much left this high dollar equipment quite vulnerable.

Ok, well those bad dies should be set OFF.... in the advanced page ... just like running stock firmware ... bad dies that dont hash at all should be set OFF. My firmware will then not issue resets to those dies. Unless of course I messed up that code somewhere, Ill look it over again, but I can confirm it at leas used to function that way =)

Offline kilroy

  • Jr. Member
  • **
  • Posts: 46
Re: NEW KNC TITAN FIRMWARE - 'GenTarkin's Mod' OFFICIAL TOPIC
« Reply #14 on: September 13, 2015, 07:26:04 AM »
I did not manually set them to off, and it does not appear as though your firmware sets them to off once it determines them bad.  I realize though, it is still good to check every now and then in case there is a false positive for a bad die or some other extenuating circumstance.  Not sure if there is need for action here on your part based on that tidbit which I really should of assessed prior as a matter of practicality.

M.Thanks