Pages:
Author

Topic: Swedish ASIC miner company kncminer.com - page 51. (Read 3049511 times)

copper member
Activity: 2898
Merit: 1465
Clueless!
December 08, 2015, 01:56:58 AM

Yeah, I guess a decent netbook would maybe be enough compute power to do the job, but I would only use one as a backup and leave the controller on the titan till the controller dies.



I would use one netbook/SBC for each cube, and then be able to place the cubes wherever I wanted, without worrying about heating/cooling/power, etc.

FYI- what you're saying about the FPGA doing some algo prep work makes sense.  I have cubes that will not function properly when connected to the controller with other cubes, but if they are the only cube connected to the controller, they work really well.
Wow, ur willing to run 1 netbook per cube!? thats a significant expense LOL! Isnt there a way to multiplex the cubes for a single netbook somehow?

Another note, if you can figure out how to reverse engineer those signals, and get at least work submission working to the cubes , we would need the "ASIC metadata"(such as clocks & voltages) sniffed and figured out as well. With that, we could set the speed to anything we wanted =) ... the 350mhz overclock, for those who wanna burn their cubes and PSU's ... that will become a reality =)

not sure if this applies but any notebook that can't run windows xp anymore (not upgradeable) like my evon800v compaq i use to monitor titans with putty..that should work right..they are a dime a dozen or so ...folk are tossing them away...until i tell them to get unbuntu 12.04 they work great with that and have a 2nd life ..hell at least for my putty uses etc so perhaps a high end netbook is NOT needed..you can pick stuff up like that for 50 bucks on ebay and probably less in you local shopper classifieds...anyway what i've been telling folk that have been chucking such notebooks because they can't run anything beyond xp anymore ..anyway just tossing this out there as a cheap solution ..or hell a cheap solution to monitor your titan with putty on the dang old beastie



Ok, time for another question to all ya guys!!!!
If there was a way to keep track of last used volts / clocks for any given cube which was previously plugged in or swapped around on the pi ... would it be useful for yall?

I believe i can code something up that would do this, using the serial number of each cube to keep track of their previously used clocks & volts on a specific pi.

IE: If I coded this up, this is what would happen: lets say you have a miner with 2 cubes and you moved those 2 cubes to different ports or even swapped them around(on the same pi), the next time you boot the miner up ... the voltages & clocks would have moved automatically w/ the cubes to their new slots.

While I could see that being useful, I don't really see myself in particular needing it. I never unplug the cubes or move them around.

well as a guy who does move cubes about (6 cubes on 1 controller and 2 on another) I think I'd at least need the option (checkbox or whatever) to clean the config file as stated by glen so to speak etc...i'm sure i'm gonna have cubes die eventually (its a race difficulty making them doorstops or hardware death making them bricks) so such an option may be of help to all us old 1st batch titans limping past 1  year of use (who'd a thunk it?)

.......not bad for a raspberry pi 8th grade science project device that knc tried to foster on us with a 90 day warranty till they found out they forgot to change the www page on such from the neppies 12yr blurb on the sales page for a month ...so they had to make it 12 months.....(*but knc being evil and all you can tell by the design they really really did not think it would last past 4 months imho)

so yeah ...if just a way to reset the config file to get around this problem or anything else that helps that does not drive Glen Tarkin 'too crazy' (Just crazy enough to continue to muck with titan firmware...not sane enough to stop..nor crazy enough to be committed...a fine line indeed for glen Smiley its all about us here don't ya know Glen) Smiley

anyway frack ...probably should not move cubes about....wait for it something will now screw up so it makes sense to move cubes around

sr. member
Activity: 405
Merit: 250
December 07, 2015, 09:13:48 PM
Nice suggestion but I have no idea how to do that second part. Also, Im strongly against any sort of centralization of the miners, which is what ur proposed idea would do.

I get the centralization part. But that's just for those that choose to use it. It can be optional. I can help create the second part if you just feed and get cube data from a web API. Would be a very helpful option, and pretty good for statistics.
legendary
Activity: 2450
Merit: 1002
December 07, 2015, 09:01:30 PM
Ok, time for another question to all ya guys!!!!
If there was a way to keep track of last used volts / clocks for any given cube which was previously plugged in or swapped around on the pi ... would it be useful for yall?

I believe i can code something up that would do this, using the serial number of each cube to keep track of their previously used clocks & volts on a specific pi.

IE: If I coded this up, this is what would happen: lets say you have a miner with 2 cubes and you moved those 2 cubes to different ports or even swapped them around(on the same pi), the next time you boot the miner up ... the voltages & clocks would have moved automatically w/ the cubes to their new slots.

That would definitely be a selling point for me. I moved around my cubes a lot (aside from making it look like I was doing cool things).

What would surely sell me, is if you created a mini cloud webservice that would save cubes and their clocks/volts so that when I move them from controller to controller they would not need re editing. Would also be nice to see the average amount of cubes and their clocks/volts globally too.

Nice suggestion but I have no idea how to do that second part. Also, Im strongly against any sort of centralization of the miners, which is what ur proposed idea would do.
sr. member
Activity: 405
Merit: 250
December 07, 2015, 08:24:49 PM
Ok, time for another question to all ya guys!!!!
If there was a way to keep track of last used volts / clocks for any given cube which was previously plugged in or swapped around on the pi ... would it be useful for yall?

I believe i can code something up that would do this, using the serial number of each cube to keep track of their previously used clocks & volts on a specific pi.

IE: If I coded this up, this is what would happen: lets say you have a miner with 2 cubes and you moved those 2 cubes to different ports or even swapped them around(on the same pi), the next time you boot the miner up ... the voltages & clocks would have moved automatically w/ the cubes to their new slots.

That would definitely be a selling point for me. I moved around my cubes a lot (aside from making it look like I was doing cool things).

What would surely sell me, is if you created a mini cloud webservice that would save cubes and their clocks/volts so that when I move them from controller to controller they would not need re editing. Would also be nice to see the average amount of cubes and their clocks/volts globally too.
legendary
Activity: 2450
Merit: 1002
December 07, 2015, 03:58:49 PM
Ok, time for another question to all ya guys!!!!
If there was a way to keep track of last used volts / clocks for any given cube which was previously plugged in or swapped around on the pi ... would it be useful for yall?

I believe i can code something up that would do this, using the serial number of each cube to keep track of their previously used clocks & volts on a specific pi.

IE: If I coded this up, this is what would happen: lets say you have a miner with 2 cubes and you moved those 2 cubes to different ports or even swapped them around(on the same pi), the next time you boot the miner up ... the voltages & clocks would have moved automatically w/ the cubes to their new slots.
yeah, that would be a cool feature ... I'll hold off on some planned maintenance(as if I need an excuse to hold off on it)

could you have it ready by Christmas?  Smiley

Maybe, although I leave in a couple weeks for christmas vacation so.... autotune(Energy Saver) AND this feature would have to be ready for release, and not sure if autotune is ready, although its getting really close so will see, after some more tests.
legendary
Activity: 2450
Merit: 1002
December 07, 2015, 03:57:25 PM
Ok, time for another question to all ya guys!!!!
If there was a way to keep track of last used volts / clocks for any given cube which was previously plugged in or swapped around on the pi ... would it be useful for yall?

I believe i can code something up that would do this, using the serial number of each cube to keep track of their previously used clocks & volts on a specific pi.

IE: If I coded this up, this is what would happen: lets say you have a miner with 2 cubes and you moved those 2 cubes to different ports or even swapped them around(on the same pi), the next time you boot the miner up ... the voltages & clocks would have moved automatically w/ the cubes to their new slots.

While I could see that being useful, I don't really see myself in particular needing it. I never unplug the cubes or move them around.

Yeah, I guess many may not use this feature. Hrm..... will see what more folks say.
sr. member
Activity: 342
Merit: 250
December 07, 2015, 03:56:43 PM
Ok, time for another question to all ya guys!!!!
If there was a way to keep track of last used volts / clocks for any given cube which was previously plugged in or swapped around on the pi ... would it be useful for yall?

I believe i can code something up that would do this, using the serial number of each cube to keep track of their previously used clocks & volts on a specific pi.

IE: If I coded this up, this is what would happen: lets say you have a miner with 2 cubes and you moved those 2 cubes to different ports or even swapped them around(on the same pi), the next time you boot the miner up ... the voltages & clocks would have moved automatically w/ the cubes to their new slots.
yeah, that would be a cool feature ... I'll hold off on some planned maintenance(as if I need an excuse to hold off on it)

could you have it ready by Christmas?  Smiley
legendary
Activity: 1596
Merit: 1000
December 07, 2015, 03:16:54 PM
Ok, time for another question to all ya guys!!!!
If there was a way to keep track of last used volts / clocks for any given cube which was previously plugged in or swapped around on the pi ... would it be useful for yall?

I believe i can code something up that would do this, using the serial number of each cube to keep track of their previously used clocks & volts on a specific pi.

IE: If I coded this up, this is what would happen: lets say you have a miner with 2 cubes and you moved those 2 cubes to different ports or even swapped them around(on the same pi), the next time you boot the miner up ... the voltages & clocks would have moved automatically w/ the cubes to their new slots.

While I could see that being useful, I don't really see myself in particular needing it. I never unplug the cubes or move them around.
legendary
Activity: 2450
Merit: 1002
December 07, 2015, 03:08:07 PM
Ok, time for another question to all ya guys!!!!
If there was a way to keep track of last used volts / clocks for any given cube which was previously plugged in or swapped around on the pi ... would it be useful for yall?

I believe i can code something up that would do this, using the serial number of each cube to keep track of their previously used clocks & volts on a specific pi.

IE: If I coded this up, this is what would happen: lets say you have a miner with 2 cubes and you moved those 2 cubes to different ports or even swapped them around(on the same pi), the next time you boot the miner up ... the voltages & clocks would have moved automatically w/ the cubes to their new slots.
full member
Activity: 121
Merit: 100
December 07, 2015, 02:37:08 PM

Im also not sure if everyone is aware of this... but clocks & volts are bound to the ports on the controller, they ARE NOT bound to the cubes themselves. So, if you move a cube to a different slot, u would have to change the clocks & volts accordingly on the advanced page.


I did notice that, yes.  I always maintained my own records, and set everything up from a clean slate after a cube was moved.
legendary
Activity: 2450
Merit: 1002
December 07, 2015, 02:03:44 PM
Heads up to everyone. Im not sure if I posted this news yet in the thread.
I have come across a very serious "overlook" on KNC's part(in addition to you know... no fan fail protection, no DCDC overcurrent-overheat protection, etc... LOL!). It was wreaking havoc on my "Energy Saver" algo & confusing the crap outta Titans who had their cubes switched around / removed & added etc....
The problem was, the configuration file which holds the data for clocks & volts was never properly cleaned up when cubes were removed or added or switched spots on the controller. This lead to "ghost" entries being left in the config file.
So, as a fix and for my next release, The firmware will now update the config file w/ the current running settings of the cubes attached(each time the monitoring script is restarted) to the controller, hence removing the "ghost" entries being left in the config file.

Im also not sure if everyone is aware of this... but clocks & volts are bound to the ports on the controller, they ARE NOT bound to the cubes themselves. So, if you move a cube to a different slot, u would have to change the clocks & volts accordingly on the advanced page.

EDIT: just had an idea, if theres a way to identify each cube uniquely like a serial number or something, via code then.... a future feature may indeed be tracking per cube clock/die configurations!!!! That would be exciting.
IE, a pi would remember which cube was plugged in where and what the clocks & volts were for it, then if the pi senses that cube moved or plugged in again at some future point then it would set the clock & voltages it was last configured with. But, if theres no unique identification per cube then...this wont be doable. Time to research a bit =)

EDIT: FOUND IT! .... the eeprom has what looks like a unique serial number for each cube. This could be used to keep track of per cube volts/clocks =) so they would be bound to cubes. New SN's would default to default clocks / voltages.
legendary
Activity: 2450
Merit: 1002
December 07, 2015, 02:00:02 PM

Yeah, I guess a decent netbook would maybe be enough compute power to do the job, but I would only use one as a backup and leave the controller on the titan till the controller dies.



I would use one netbook/SBC for each cube, and then be able to place the cubes wherever I wanted, without worrying about heating/cooling/power, etc.

FYI- what you're saying about the FPGA doing some algo prep work makes sense.  I have cubes that will not function properly when connected to the controller with other cubes, but if they are the only cube connected to the controller, they work really well.
Wow, ur willing to run 1 netbook per cube!? thats a significant expense LOL! Isnt there a way to multiplex the cubes for a single netbook somehow?

Another note, if you can figure out how to reverse engineer those signals, and get at least work submission working to the cubes , we would need the "ASIC metadata"(such as clocks & voltages) sniffed and figured out as well. With that, we could set the speed to anything we wanted =) ... the 350mhz overclock, for those who wanna burn their cubes and PSU's ... that will become a reality =)
sr. member
Activity: 342
Merit: 250
December 07, 2015, 01:52:54 PM
It's not about bandwidth (I mentioned bus to refer to protocols and speed). Speed IS a factor here when dealing with an ASIC that has it's performance rely on speed. You don't think your miner would hash faster if the program that was feeding it work was on the controller board rather than your PC? I mean, it's a micro improvement but that's all it takes sometimes. It's about the microcontroller on the controller board. This is how companies keep hardware closed down so that they control the flow. Any ASIC, any FPGA. It's how we do things.

I suppose you could leak all the cables and just get obscure information, but that's slow and may not result to anything if everything is encrypted (knowing KnC, they would encrypt it). It would be much easier to reverse engineer a chip that we know the schematics on. I've been working on exploiting the cyclone on the board with some overflows and some old techniques, but i'll need some more time. Like I said, it's nothing like the old Cyclone chips.

**EDIT** I stand corrected. Tried sniffing the cables. Looks like building a software solution will be possible, unfortunately, speed would be the problem.

Hi qberty, I have been thinking about your bridge connector problem and it sounds very similar to the bad electrolytic capacitor problem I had on 2 cubes. Both caps went bad after power related issues. One was a power failure, the other was a burnt Y connector.

I am not saying this is the solution to your problem, but replacing the big electrolytic cap might be something to consider and a heck of a lot cheaper if it works
full member
Activity: 121
Merit: 100
December 07, 2015, 01:51:33 PM

Yeah, I guess a decent netbook would maybe be enough compute power to do the job, but I would only use one as a backup and leave the controller on the titan till the controller dies.



I would use one netbook/SBC for each cube, and then be able to place the cubes wherever I wanted, without worrying about heating/cooling/power, etc.

FYI- what you're saying about the FPGA doing some algo prep work makes sense.  I have cubes that will not function properly when connected to the controller with other cubes, but if they are the only cube connected to the controller, they work really well.
legendary
Activity: 2450
Merit: 1002
December 07, 2015, 01:45:54 PM
**EDIT** I stand corrected. Tried sniffing the cables. Looks like building a software solution will be possible, unfortunately, speed would be the problem.

In scrypt, on work restarts I think theres some compute heavy "higher level" of the algo that gets calculated prior to dispatching the jobs to the ASIC itself. Perhaps, this is what the controller is tasked with and does it a lot faster then a pi could.


It does this faster than a Pi could, but what about a netbook or a regular PC?  Even if those were not as fast as the FPGA, would they be sufficiently fast to eliminate the controller?  Currently, if your controller goes down, your Titan is dead.  There are no replacement controllers available.

Someone on Ebay is trying to sell a controller for $900.  For less than half of that, I could buy a netbook or SBC for each cube and be done with it.

Yeah, I guess a decent netbook would maybe be enough compute power to do the job, but I would only use one as a backup and leave the controller on the titan till the controller dies.
full member
Activity: 121
Merit: 100
December 07, 2015, 01:39:55 PM
**EDIT** I stand corrected. Tried sniffing the cables. Looks like building a software solution will be possible, unfortunately, speed would be the problem.

In scrypt, on work restarts I think theres some compute heavy "higher level" of the algo that gets calculated prior to dispatching the jobs to the ASIC itself. Perhaps, this is what the controller is tasked with and does it a lot faster then a pi could.


It does this faster than a Pi could, but what about a netbook or a regular PC?  Even if those were not as fast as the FPGA, would they be sufficiently fast to eliminate the controller?  Currently, if your controller goes down, your Titan is dead.  There are no replacement controllers available.

Someone on Ebay is trying to sell a controller for $900.  For less than half of that, I could buy a netbook or SBC for each cube and be done with it.
legendary
Activity: 2450
Merit: 1002
December 07, 2015, 12:28:35 PM
**EDIT** I stand corrected. Tried sniffing the cables. Looks like building a software solution will be possible, unfortunately, speed would be the problem.

In scrypt, on work restarts I think theres some compute heavy "higher level" of the algo that gets calculated prior to dispatching the jobs to the ASIC itself. Perhaps, this is what the controller is tasked with and does it a lot faster then a pi could.
full member
Activity: 121
Merit: 100
December 07, 2015, 09:17:37 AM

**EDIT** I stand corrected. Tried sniffing the cables. Looks like building a software solution will be possible, unfortunately, speed would be the problem.


Okay, I'm very interested now.  Tell us more.  I smell some crowdfunding coming your way. 

(My uneducated guess was always that this wouldn't be as hard to do as everyone thought it would.)
legendary
Activity: 2450
Merit: 1002
December 06, 2015, 09:44:54 PM
Knc is closed at factory
Or not?


No answer at email
No

News firmware for titan


Nothing

Um.... I thought u knew this whole time... LOL! KNC has effectively abandoned Titan users. Cept for the hand picked warranties they decide to honor here and there.
legendary
Activity: 2408
Merit: 1004
December 06, 2015, 07:57:20 PM
Knc is closed at factory
Or not?


No answer at email
No

News firmware for titan


Nothing
Pages:
Jump to: