Author

Topic: Swedish ASIC miner company kncminer.com - page 1019. (Read 3050071 times)

-ck
legendary
Activity: 4088
Merit: 1631
Ruu \o/
November 19, 2013, 06:44:54 AM
Well, here's a binary which should hopefully address the known memory leak from cgminer itself based on the latest git master 009511c995458c00a5f46426c8e80ddb94c20cbb.

http://ck.kolivas.org/apps/cgminer/kncminer/cgminer
hero member
Activity: 635
Merit: 500
November 19, 2013, 06:35:40 AM
I think it was not Knc's fault.

But 'Orama, could somebody call Ups from Knc and ask what happened? They give info only for Knc....

This is only history but I'd only like to know what happened.
copper member
Activity: 2898
Merit: 1465
Clueless!
November 19, 2013, 06:34:05 AM
I just got off the phone with UPS. The lady told me that the shipment is very (very) unlikely to be delivered today (Germany) and I shall take this up with KNC (which I just did). The sender is the only one able to get UPS to investigate the issue. She said it seems the paperwork got shipped but the actual boxes were left behind.

It's sad that KNC has not already responded to the issue (it's noon in Sweden). Their decision to not sending the upgrades out Friday will result in half a week of lost mining. All I can say to this (from the bottom of my heart) is: Thank you SO much KNC. You really, REALLY rock. This is by far the fairest solution for all miners.

Totally unrelated: Congrats to those who decided not to trust KNC with the delivery and pick the boards up locally. I should have done the same.

Because this is KnC's fault, how?! ALL boxes were collected by UPS yesterday. They took them all in plastic packaging crates. UPS arrived here at 4pm. The boxes were ready, so no delay from us whatsoever.

Because of this it is your fault:
to be fair to everybody that ordered within those 7 minutes, they are all being shipped at the same time on Monday.
You could have shipped them last week (maybe not all...but a significant number). They were advertised that way you know....


prob would not matter whatever the issue with UPS and the boxes would probably have happened then as well...of course may have saved a day in untangling the mess but then again the weekend so probably not  imho just saying (like I really know anything about this  but seems likely)
hero member
Activity: 532
Merit: 500
November 19, 2013, 06:14:49 AM
I just got off the phone with UPS. The lady told me that the shipment is very (very) unlikely to be delivered today (Germany) and I shall take this up with KNC (which I just did). The sender is the only one able to get UPS to investigate the issue. She said it seems the paperwork got shipped but the actual boxes were left behind.

It's sad that KNC has not already responded to the issue (it's noon in Sweden). Their decision to not sending the upgrades out Friday will result in half a week of lost mining. All I can say to this (from the bottom of my heart) is: Thank you SO much KNC. You really, REALLY rock. This is by far the fairest solution for all miners.

Totally unrelated: Congrats to those who decided not to trust KNC with the delivery and pick the boards up locally. I should have done the same.

Because this is KnC's fault, how?! ALL boxes were collected by UPS yesterday. They took them all in plastic packaging crates. UPS arrived here at 4pm. The boxes were ready, so no delay from us whatsoever.
-ck
legendary
Activity: 4088
Merit: 1631
Ruu \o/
November 19, 2013, 06:01:02 AM
  • as ckolivas says its purportedly due to the web interface, which makes me think its a memory leak in the API calls? ck, maybe you could clarify, would using miner.php have the same issue?
Miner.php wouldn't have this issue. It's allegedly due to using the json commands to the cgminer RPC API instead of the standard API calls as in use by miner.php and elsewhere.

Cool, I hate memory leaks, tricky to track down, especially if theres multiple suspects. All I can say is that 3.4.0 with their firmware 0.96 doesn't seem to suffer from the same problem.
That's because that version reported nothing via API for the device since that predates when I wrote code for it, and they never wrote any code for the device specific information.

Ok so I've replaced the 3.4.0 with the latest 3.8.2 and I'll run with that for 24-48 hours and see how it goes, same miner firmware that seems to work well for me, but just replace cgminer.
But that will keep using the API which causes the problem if you're using the web interface... that's not a solution unless you move to ssh/putty monitoring.
member
Activity: 76
Merit: 10
November 19, 2013, 05:46:51 AM
I just got off the phone with UPS. The lady told me that the shipment is very (very) unlikely to be delivered today (Germany) and I shall take this up with KNC (which I just did). The sender is the only one able to get UPS to investigate the issue. She said it seems the paperwork got shipped but the actual boxes were left behind.

It's sad that KNC has not already responded to the issue (it's noon in Sweden). Their decision to not sending the upgrades out Friday will result in half a week of lost mining. All I can say to this (from the bottom of my heart) is: Thank you SO much KNC. You really, REALLY rock. This is by far the fairest solution for all miners.

Totally unrelated: Congrats to those who decided not to trust KNC with the delivery and pick the boards up locally. I should have done the same.
copper member
Activity: 2898
Merit: 1465
Clueless!
November 19, 2013, 05:46:42 AM

        enablecores just turns on any cores that were turned off from hardware errors, not sure how useful it is if using latest firmware as cgminer turns the cores on and off itself, but if going back to earlier firmware (pre 0.98) then is a must[/li][/list]


    ok ...I thought enablecores turned them all ON all the time....if enablecores just copies cgminer turning them off and on as needed I guess I don't need the enablecores.bin patch

    thanks

    Searing

    My understanding is that the enablecores just turns them all back on as a one time shot in the arm type of thing, they can still be turned off afterwards through either the later firmwares or via a version of cgminer that supports that facility.
    [/quote]

    so no way to change the number of error before this happens it always says "detected 10 errors shutting off core"  kinda thing

    would be nice if you could set it for like 20 errors (or maybe I'm missing something on cgminer

    anyway if it just turns them all on for a time not needed for me they just toggle on/off fine on their own whim now and it is only 2 that do so every few hours and another 2 that maybe do it 2 or 3x a day

    anway thanks
    Searing
    legendary
    Activity: 1372
    Merit: 1022
    Anarchy is not chaos.
    November 19, 2013, 05:29:18 AM
    Guess we shall see Mota...because I'm on an aggressive re-investment strategy....




    I have exactly one problem with your strategy. Pure jealousy. I missed out, due to not being able to raise the money.

    I think reinvesting in the mining equipment is a great long term strategy. I also think that if it were me, my first reinvestment would be in off grid power. Probably a combination of solar and hydro, since here in Idaho you can get remote land for very little money that has a stream or more running through it. Water wheels are cheap, and alternators aren't much more expensive. Batteries are the biggest investment, and they have a pretty good life these days. Getting the electric cost to near zero would make even the initial miners valuable for a long time, as they would simply continue to add to your hashpower as you added in more equipment 9both miners and electric generation).

    This is the strategy I want to follow, assuming I can turn the financial disaster that is my current life around.
    legendary
    Activity: 3234
    Merit: 1221
    November 19, 2013, 05:27:45 AM

          enablecores just turns on any cores that were turned off from hardware errors, not sure how useful it is if using latest firmware as cgminer turns the cores on and off itself, but if going back to earlier firmware (pre 0.98) then is a must[/li][/list]


      ok ...I thought enablecores turned them all ON all the time....if enablecores just copies cgminer turning them off and on as needed I guess I don't need the enablecores.bin patch

      thanks

      Searing
      [/quote]

      My understanding is that the enablecores just turns them all back on as a one time shot in the arm type of thing, they can still be turned off afterwards through either the later firmwares or via a version of cgminer that supports that facility.
      legendary
      Activity: 3234
      Merit: 1221
      November 19, 2013, 05:25:36 AM
      • as ckolivas says its purportedly due to the web interface, which makes me think its a memory leak in the API calls? ck, maybe you could clarify, would using miner.php have the same issue?
      Miner.php wouldn't have this issue. It's allegedly due to using the json commands to the cgminer RPC API instead of the standard API calls as in use by miner.php and elsewhere.

      Cool, I hate memory leaks, tricky to track down, especially if theres multiple suspects. All I can say is that 3.4.0 with their firmware 0.96 doesn't seem to suffer from the same problem.
      That's because that version reported nothing via API for the device since that predates when I wrote code for it, and they never wrote any code for the device specific information.

      Ok so I've replaced the 3.4.0 with the latest 3.8.2 and I'll run with that for 24-48 hours and see how it goes, same miner firmware that seems to work well for me, but just replace cgminer.
      copper member
      Activity: 2898
      Merit: 1465
      Clueless!
      November 19, 2013, 05:25:08 AM
        enablecores just turns on any cores that were turned off from hardware errors, not sure how useful it is if using latest firmware as cgminer turns the cores on and off itself, but if going back to earlier firmware (pre 0.98) then is a must[/li]

      [/quote]

      ok ...I thought enablecores turned them all ON all the time....if enablecores just copies cgminer turning them off and on as needed I guess I don't need the enablecores.bin patch

      thanks

      Searing
      newbie
      Activity: 13
      Merit: 0
      November 19, 2013, 05:24:22 AM
      From the very beginning I wished they used FedEx. There are much less mistakes with FedEx and if you pay $300 for shipping on anything with FedEx they'll buy lunch for you that day. I want my god damn shipping charges refunded back to me. Pathetic.


      Really?? Do you know how close KnC were to being over a week late in being able to create any machines?

      FedEx delivered the majority and all of the first batch of Beaglebones to the wrong country and then spent three days lying about the whereabouts in the final week of September. Marcus refuses to use Fedex ever again, for anything ever again. He hates them with a passion for this.

      It was only because Beaglebone themselves worked through the day and night for three days solid and paid DHL to deliver expedited that there was any. Even though FedEx were well aware of the problem and Beaglebone found out and restarted manufacture, DHL still beat them.

      The Beaglebones arrived Saturday, the chips Sunday, that's how close to line this project was!!

      My UPS tracking shows my boards left Sweden, arrived in Germany and there were sent back to Sweden. As others are in the same position we'd appreciate it if you could you provide a clear indication if this is dodgy UPS tracking or if there is another issue?

      cheers
      -ck
      legendary
      Activity: 4088
      Merit: 1631
      Ruu \o/
      November 19, 2013, 05:14:38 AM
      • as ckolivas says its purportedly due to the web interface, which makes me think its a memory leak in the API calls? ck, maybe you could clarify, would using miner.php have the same issue?
      Miner.php wouldn't have this issue. It's allegedly due to using the json commands to the cgminer RPC API instead of the standard API calls as in use by miner.php and elsewhere.

      Cool, I hate memory leaks, tricky to track down, especially if theres multiple suspects. All I can say is that 3.4.0 with their firmware 0.96 doesn't seem to suffer from the same problem.
      That's because that version reported nothing via API for the device since that predates when I wrote code for it, and they never wrote any code for the device specific information.
      legendary
      Activity: 3234
      Merit: 1221
      November 19, 2013, 05:09:33 AM
      • as ckolivas says its purportedly due to the web interface, which makes me think its a memory leak in the API calls? ck, maybe you could clarify, would using miner.php have the same issue?
      Miner.php wouldn't have this issue. It's allegedly due to using the json commands to the cgminer RPC API instead of the standard API calls as in use by miner.php and elsewhere.

      Cool, I hate memory leaks, tricky to track down, especially if theres multiple suspects. All I can say is that 3.4.0 with their firmware 0.96 doesn't seem to suffer from the same problem.
      -ck
      legendary
      Activity: 4088
      Merit: 1631
      Ruu \o/
      November 19, 2013, 05:00:09 AM
      by the by you've been mining a long time...gonna sell all your BTC and buy and Island or something ..myself I've made 21 coins in a month with a Jupiter and the temptation to take the profits and run amuck with a sports car or and/other mid life crisis is strong..have no idea if you have like 100 plus coins what the last week feels like.
      Funny thing about this, I've never been a big time miner... most of my btc has come from donations or paid work in btc supporting the ecosystem. Also I happen to have a full time career that has nothing to do with computing, software or bitcoin.
      hero member
      Activity: 532
      Merit: 500
      November 19, 2013, 04:57:13 AM
      From the very beginning I wished they used FedEx. There are much less mistakes with FedEx and if you pay $300 for shipping on anything with FedEx they'll buy lunch for you that day. I want my god damn shipping charges refunded back to me. Pathetic.


      Really?? Do you know how close KnC were to being over a week late in being able to create any machines?

      FedEx delivered the majority and all of the first batch of Beaglebones to the wrong country and then spent three days lying about the whereabouts in the final week of September. Marcus refuses to use Fedex ever again, for anything ever again. He hates them with a passion for this.

      It was only because Beaglebone themselves worked through the day and night for three days solid and paid DHL to deliver expedited that there was any. Even though FedEx were well aware of the problem and Beaglebone found out and restarted manufacture, DHL still beat them.

      The Beaglebones arrived Saturday, the chips Sunday, that's how close to line this project was!!
      -ck
      legendary
      Activity: 4088
      Merit: 1631
      Ruu \o/
      November 19, 2013, 04:57:05 AM
      • as ckolivas says its purportedly due to the web interface, which makes me think its a memory leak in the API calls? ck, maybe you could clarify, would using miner.php have the same issue?
      Miner.php wouldn't have this issue. It's allegedly due to using the json commands to the cgminer RPC API instead of the standard API calls as in use by miner.php and elsewhere.
      hero member
      Activity: 635
      Merit: 500
      November 19, 2013, 04:53:27 AM
      'Orama!

      Could you ask somebody at Knc what the fuck happened with the delivery?

      Ups only give info to Knc cause they paid for shipping.

      I'm not angry just broke a few chairs... Smiley

      I've asked the local Ups customer service, they told me I'll receive them tomorrow so not a big mistake but......
      legendary
      Activity: 3234
      Merit: 1221
      November 19, 2013, 04:42:46 AM

      Yep I go back to .96 all the time because none of the newer firmwares do anything to improve my Jupiter except make it use more power and then after a few days lose more GH/s

      So normally for me, its reflash 0.96, reboot, flash enablecores, reboot, flash bertmod (don't reboot)

      You do realize the dwindle-down effect is from using the Bertmod....?...Right?
      Remember the memory issue?
      You are partially responsible for helping identify the leak.
      The memory won't get gobbled up if you stay off the GUI, putty, and bertmod..... and just use the pool to monitor...
      Just checking...

      I think its just the GUI and bertmod (seeing as it becomes pare of the GUI) putty is fine according to ckolivas.

      However I've also found that it only seems to affect the newer versions of cgminer in the new firmwares, although it appears fixed in .99

      The firmware I use has cgminer 3.4.0 which doesn't seem to suffer from the memory leak regardless of if you are using the web interface or not.  Huh
      legendary
      Activity: 3234
      Merit: 1221
      November 19, 2013, 04:35:05 AM
      well now that my miner has been running more then 24hrs on .99 and my 'at pool' hashrate on slush has improved 25gh plus ...was running a very stable 555gh on .95 so waited a bit to upgrade....
      (shocking what 24hrs of burn in can do)...anyway my questions:


      1) enablecores download on the knc site..... do you install this the same way you install update firmware like .99 binary?

      2) any danger to using this (overclocks?) core temps are higher etc? for my setup since .99 upgrade I have like 4 cores poping off/on once in a great while 1 every 1/2h or so..should I even bother?

      3) the 'memory leak" people talk about when slow creep lose of hash rate (minuscule I guess) by running the knc screen all the time on the miner or putty on all the time..anyway has this leak been fixed on new version..of cgminer?   thoughts? or so lame in gh loss no big deal?

      4)  as to being better dumb and lucky vs any planning on all this is $15,750 usd (21btc at 750/coin) a good haul for 31 days of mining (hell yes) and being as smug as we all are (was it unacceptable who has 11 Jupiters ...the glory...the mind boggles) ....will I ever be able go back to "bitcoin miner peasant status" if *gasp* BTC goes back down to only $200 bucks a coin..(still a good return from when I started my month of mining at $150..ah Good old BTC days or yore....where you say stuff " the good old days of btc at $150 and you mean 1 month ago*

      5) Is all this BTC up Drama etc weirder then heck? as I write this www.mtgox.com is at $793 USD

      my 2c worth
      Searing


      • Yes, just like a firmware upgrade
      • enablecores just turns on any cores that were turned off from hardware errors, not sure how useful it is if using latest firmware as cgminer turns the cores on and off itself, but if going back to earlier firmware (pre 0.98) then is a must
      • as ckolivas says its purportedly due to the web interface, which makes me think its a memory leak in the API calls? ck, maybe you could clarify, would using miner.php have the same issue?
      • Get some sleep man you are rambling Smiley
      • See 4 Cheesy
      Jump to: