Swedish ASIC miner company kncminer.com

[soapmodem]

i have replace my cgminer binary with the cgminer-tune binary. renamed it and set the rights to 755. then /etc/init.d/cgminer.sh restart. till here everything fine. but after a reboot it seems that the binary gets replaced again by 3.8.1 . any way for a permanent replacement?

There was talk on the kncminer.com forums of a permanent replacement, with the binary included in a bin patch, but I don't know where that went or if it ever got off the ground. You'd have to rebuild the image with each firmware and cgminer update.. so probably not worth it.

It doesn't take more than a few seconds, so you're better off just doing:

Code:

wget http://ck.kolivas.org/apps/cgminer/kncminer/cgminer-tune
chmod +x cgminer-tune
screen -dm ./cgminer-tune -c /config/cgminer.conf

[Biffa]

Heads-up for tomo BTW - incoming...

Also want to add, my ambient temps are 30C - 86F.

so statements regarding VRMs needing heating and/or not liking cooling is NOT a fact covering ALL rigs/VRMs and even when they are hot(ter) they dont ALL fall back into line....

im sure ive mentioned this...

As for 'wiggling' the heatsink aka 'cooling tower' - i think ill give that a miss.

i personally dont want to hear 'warranty' is void due to negligence/abuse/possibly insane advice (though it might just work)

Regarding your temps, what temps are you getting? I noticed my rig was starting to drop performance no matter what firmware I used, even going back to my faithful 0.96 firmware didn't seem to make much difference. (circa 530Gh/s av)

However I had noticed that the ambient was getting pretty bloody cold here, so I moved it to a warmer location and performance jumped around 10Gh/s, then I reloaded 0.96 (always seems better for this rig, lower voltage, more stable) but wanted more realistic output from cgminer so loaded up ckolivas' latest 3.8.2 without tuning. And low and behold, the higher temps, the best firmware and more accurate cgminer and I'm back up to 550Gh/s avg over 24 hours, with spikes up to 600+Gh/s

Thing is edgar, these machines aren't perfect, they are enthusiast level, makes sense really for the market they are in. And as such they need a bit of TLC to get the most out of them. Yes some will be broken and need RMA'ing as a last resort, but there is no hard and fast rule that works for each one, and I think that finally KnC have realised this too, the tuning suite sounds more like you will be able to tune each ASIC separately for voltage, hopefully speed and maybe some other things.

This isn't going to mean that someone will give the "magic settings" that will work for everyone, if anything it means more work to find the best "tune" for your particular rig. Its a bit like overlcocking CPU's or GPU's no-one elses settings are quite right for your particular hardware.

I mean, I guess, they could have set a generic right-for-everyone 400Gh/s firmware but I think they knew that their hardware had more potential, probably a bad move on their part shouting about the upper limit of the October batches, but it sold alot of miners I bet!

I've been pretty happy with 530Gh/s but with difficulty increasing I'm looking to squeeze as much out of this puppy as I can, so today its remove the blue power cables and plug the Pcie's straight into the boards

This is an enthusiasts forum, and people are allowed to get enthusiastic, believe it or not Phoenix and the others in here have help loads of people get their machines working nicely. I for one think they are doing a great job and should continue. I salute their successes and apply what tweaks work for my rig, sometimes I even pass on things that have worked for me.

Thats the nature of forums like this, no offence, but it doesn't seem like it suits your nature. I just wonder what you actually want from this forum? Can you spell it out? Maybe bullet point it or something so that you have a record of the things you want addressed. Can't promise anything mind but it might help us understand why you seem so angry. 

[Biffa]

Sorry this is a little late but how many bad cores do you need before KNC will allow an RMA?

Was your ASIC completely dead before KNC would give you an RMA?

I have one ASIC with probably 15 bad cores which slows my hashrate...   

Or just live with it?

>>>As soon as the first report good numbers I'll go for it.
>>>Currently running stable on 3.8.2 with my 5/192 bad cores board, so I don't want to mess with it now.

felix is it 15 bad cores or 5?

You know that with the tuning cgminer it turns cores on and off to try and find a stable medium.

What are your HW error %'s?

I've given up on the firmwares that pump the voltage up (>0.97) and tune the cores, I'm sticking with 0.96 and the non tuning 3.8.2 cgminer and my HW errors are down from 4.5% to 1.3% and dropping.

[jelin1984]

Mining Status
CGMiner Status   Running (pid=1774)
Last Checked   Mon Nov 25 09:31:45 UTC 2013
Avg. Hash Rate   563 Gh/s
Difficulty Accepted   9,110,016
Difficulty Rejected   5,120 (0.1 %)
Hardware Errors   84,158 (0.9 %)
WU   7,852
Difficulty Stale   0
Network Blocks   146
Pool Rejected   0.1 %
Pool Stale   0
Best Share   2,235,713
Found Blocks   0


these is beta 0.98.1
is ok?

[-ck]

Mining Status
CGMiner Status   Running (pid=1774)
Last Checked   Mon Nov 25 09:31:45 UTC 2013
Avg. Hash Rate   563 Gh/s
Difficulty Accepted   9,110,016
Difficulty Rejected   5,120 (0.1 %)
Hardware Errors   84,158 (0.9 %)
WU   7,852
Difficulty Stale   0
Network Blocks   146
Pool Rejected   0.1 %
Pool Stale   0
Best Share   2,235,713
Found Blocks   0


these is beta 0.98.1
is ok?

Looks good, but every hardware is different with respect to dud cores. Some is golden and some is scat.

[jelin1984]

thanks ckolivas

[Biffa]

Looks good, but every hardware is different with respect to dud cores. Some is golden and some is scat.

ck, can you elaborate a little on the difference in behaviour & reporting between the 3.8.2 tuned and non-tuned on a KnC miner?

I haven't run the latest tuned version but in earlier ones it would turn the cores on and off according to number of hw errors (off) and time (on)

To see the status I'd run Update Asic Status on Bertmod and see what cores were turned on or off, these numbers would change over time obviously.

With the non-tuning version of cgminer, the cores don't get turned on and off.

With both versions of the software they report HW errors numerically and as a percentage.

Is there any correlation between the HW numbers and the different methods that the tuning/non-tuning versions of cgminer use to handle cores?

Or are HW errors irrelevant due to the way the KnC hardware reports them?

Cheers!

[TaggedYa]

I am not sure this is the right place for this but this seems to be a catch all thread so here goes.

I have 2 late October Jupiters.  One I was able to get to 550 with all cores running under 98.1 beta.  The other has 2 cores on one board that will not work until 73C or 74C and as soon as they start the whole board shuts down (I assume thermal shutdown) so it only works at around 480.

I was living with that (not economically feasible to give up 480 hashes for 2 weeks to rma the thing just to maybe get another 60 or so).

Now the problem.  I shut the full speed unit down for a few minutes to change locations and when I restarted it cgminer fails to start.  The bone boots and I can tty into it.  I can try to manually start cgminer and I get a "cgminer failed to start: Write Error".

I tried changing firmware.  Cgminer will run on 98 but not on anything after and on 98 I have all die 0s dead.

Anyone have any clues?

TaggedYa

[sickpig]

from #kncminer channel

10:45 < btcoco> any rig worth investment atm?
10:55 < hno> btcoco, depends on how you think the BTC price will evolve and associated risks. Unfortunately KnCMiner is all sold out at the moment, but there will
             be more available in some weeks I think.

for who doesn't know hno, he's an OrSoC engineer. This going to be big, potentially  

[Biffa]

I am not sure this is the right place for this but this seems to be a catch all thread so here goes.

I have 2 late October Jupiters.  One I was able to get to 550 with all cores running under 98.1 beta.  The other has 2 cores on one board that will not work until 73C or 74C and as soon as they start the whole board shuts down (I assume thermal shutdown) so it only works at around 480.

I was living with that (not economically feasible to give up 480 hashes for 2 weeks to rma the thing just to maybe get another 60 or so).

Now the problem.  I shut the full speed unit down for a few minutes to change locations and when I restarted it cgminer fails to start.  The bone boots and I can tty into it.  I can try to manually start cgminer and I get a "cgminer failed to start: Write Error".

I tried changing firmware.  Cgminer will run on 98 but not on anything after and on 98 I have all die 0s dead.

Anyone have any clues?

TaggedYa

Trouble shooting wise.

Make sure all the power cables are correctly attached.
Do a hard reset on the device:

HARD RESET
The miner is equipped with a hard reset button. This is located next to the power
input on the controller board. To perform a hard reset on the miner you must
press the button 5 times consecutively, wait 5 seconds, and then press the button a
further 5 times. Please note that the miner will not restart.

Also:
From the KNC support page: https://www.kncminer.com/pages/troubleshooting

Underperforming Miner

If you have a miner that seems to be underperforming you need to follow the below instruction.

To verify if an ASIC board is broken or the performance drop is due to software issues please follow these steps:

 1. Turn off and disconnect your PSU.
 2. Open your miner case.
 3. Unplug the flat ribben cable connecting one of the ASIC Boards (the circuit board that has the big fan on it) to the controller board.
 4. Unplug the PCI-Express power cable from the ASIC board.
 5. Ensure all other cables are connected properly.
6. Reconnect the PSU and power on the miner. Let the miner work for an hour to asses performance.  

If the miner is now performing as expected you should install the latest firmware.

7. Turn off the power and disconnect the PSU, reconnect all the PCI-E and ribben cables.
 8. Restart the miner and install the latest firmware.

Lastly, If you can isolate the bad board, then you can just return that bad board for an RMA and the miner can continue mining, you don't have to lose the entire miner and 480Gh/s just for one bad board.

[ncs0ne]

(...)
Now the problem.  I shut the full speed unit down for a few minutes to change locations and when I restarted it cgminer fails to start.  The bone boots and I can tty into it.  I can try to manually start cgminer and I get a "cgminer failed to start: Write Error".
(...)

I'm not able to find the string 'Write Error' (without caps as well) in cgminer source code.
Is it the exact error msg you get ?

If so I guess your issue is located somewhere in the OS.
Anything useful in the log file ?

*Note: journalctl is the replacement for a "less /var/log/syslog"
http://www.freedesktop.org/software/systemd/man/journalctl.html

[Biffa]

(...)
Now the problem.  I shut the full speed unit down for a few minutes to change locations and when I restarted it cgminer fails to start.  The bone boots and I can tty into it.  I can try to manually start cgminer and I get a "cgminer failed to start: Write Error".
(...)

I'm not able to find the string 'Write Error' (without caps as well) in cgminer source code.
Is it the exact error msg you get ?

If so I guess your issue is located somewhere in the OS.
Anything useful in the log file ?

*Note: journalctl is the replacement for a "less /var/log/syslog"
http://www.freedesktop.org/software/systemd/man/journalctl.html

write error could mean the beaglebone storage is full or corrupted.

in putty:
df -h will show the disk space
fdisk -l will show the partition info

[Madazam]

Ok the Noob , (thats me) got it working....

So i upgraded the modules and my saturn is hashing at a jupiter speed ...only one prob ...the front fans on casing have stopped working ...any idea why and solution ?

[btc_uzr]

from #kncminer channel

10:45 < btcoco> any rig worth investment atm?
10:55 < hno> btcoco, depends on how you think the BTC price will evolve and associated risks. Unfortunately KnCMiner is all sold out at the moment, but there will
             be more available in some weeks I think.

for who doesn't know hno is a OrSoC eng. This going to be big, potentially 

nothing new if in some weeks means March

7/18/2013 12:27:00 PM

We will ship no devices in December 2013, January 2014 or  February 2014. Meaning that once we have taken the difficulty up at the end of November we will not release any more hashing power for 3 months. We will then release our new generation of devices, which will begin shipping in March 2014. These devices will also have a much higher GH/$ rating than any of our current offerings.

[btc_uzr]

Ok the Noob , (thats me) got it working....

So i upgraded the modules and my saturn is hashing at a jupiter speed ...only one prob ...the front fans on casing have stopped working ...any idea why and solution ?

unplugged connectors ?

[-ck]

Looks good, but every hardware is different with respect to dud cores. Some is golden and some is scat.

ck, can you elaborate a little on the difference in behaviour & reporting between the 3.8.2 tuned and non-tuned on a KnC miner?

Reporting between the two is identical.

The tuned one simply changes the decision making process about when to turn a core off and back on. Firstly it turns all cores on every time it's starting. However it turns cores off very quickly instead of after many errors, but turns cores off in a staggered fashion a few seconds apart. Then it turns them back on after only a minute, but will only turn cores back on if no core on/off has happened for the last minute staggering them much more to turn them back on. My observations were that turning dud cores off always helps the hashrate so making it easier to turn cores off made sense, but with rapid changes to the number of cores on/off, the power going to the units changes wildly when in fact the device will be more stable if it only ever changes power requirements slowly, hence the changes. It also increases the duration cores are off every time they fail before disabling them permanently (for that run of cgminer).

Anyway I find the hashrate and consequently the hardware error rate keeps improving over time as only the decent cores are left hashing, and cores that are borderline turn on and off intermittently in a staggered way that doesn't change the power requirements dramatically.

[sickpig]

from #kncminer channel

10:45 < btcoco> any rig worth investment atm?
10:55 < hno> btcoco, depends on how you think the BTC price will evolve and associated risks. Unfortunately KnCMiner is all sold out at the moment, but there will
             be more available in some weeks I think.

for who doesn't know hno is a OrSoC eng. This going to be big, potentially 

nothing new if in some weeks means March

sure you're right if he meant to say a few months ... I suppose that the only thing we can do is waiting for official news. time will tell.

[Madazam]

Ok the Noob , (thats me) got it working....

So i upgraded the modules and my saturn is hashing at a jupiter speed ...only one prob ...the front fans on casing have stopped working ...any idea why and solution ?

unplugged connectors ?

No man ..checked already ......what could be any other reason ......

[Tigggger]

ck, can you elaborate a little on the difference in behaviour & reporting between the 3.8.2 tuned and non-tuned on a KnC miner?

Reporting between the two is identical.

The tuned one simply changes the decision making process about when to turn a core off and back on. Firstly it turns all cores on every time it's starting. However it turns cores off very quickly instead of after many errors, but turns cores off in a staggered fashion a few seconds apart. Then it turns them back on after only a minute, but will only turn cores back on if no core on/off has happened for the last minute staggering them much more to turn them back on. My observations were that turning dud cores off always helps the hashrate so making it easier to turn cores off made sense, but with rapid changes to the number of cores on/off, the power going to the units changes wildly when in fact the device will be more stable if it only ever changes power requirements slowly, hence the changes. It also increases the duration cores are off every time they fail before disabling them permanently (for that run of cgminer).

Anyway I find the hashrate and consequently the hardware error rate keeps improving over time as only the decent cores are left hashing, and cores that are borderline turn on and off intermittently in a staggered way that doesn't change the power requirements dramatically.

Excellent work ck, I have about 16 cores that have never worked in any firmware version, currently they just go on and off all the time and if this is impacting the rest of the system sounds like a good fix.

OT: 20130821 Uptime 96d

[demonmaestro]

not enough power?