Hacking The KNC Firmware: Overclocking

[elasticband]

I have an October Saturn currently running 612/589ghs through cgminer and 585 through KNC's gui. This is far and away the absolutely highest numbers I've seen (for me and this machine), by almost double.

Thing is, it's doing no good. Error must be bad. WU is only 1450ish.

Poolside reporting is tragic -- looks like about 185ghs.

If someone can help me figure out how to make use of this apparent speed/reduce errors, I'll leave you a nice tip. I don't have enough knowledge and must resort to bribes.

You need to tune each h board. Send me team viewer details if you would like me to take a look

[1713574600]

1713574600

[elasticband]

i tweaked many things - https://bitcointalk.org/index.php?topic=313978.msg5175175#msg5175175

they are running fine, with under 1.5% HW. 740 + 750GH, going on a couple of ours now. Still buzzing...... but oh well just going to roll with it.

Well 24hrs later, still getting an extreme buzz from both 8vrm jupiters, but 750 + 740 GH with bellow 1.5% makes me keep going, really everything seems fine.

[j03]

I have an October Saturn currently running 612/589ghs through cgminer and 585 through KNC's gui. This is far and away the absolutely highest numbers I've seen (for me and this machine), by almost double.

Thing is, it's doing no good. Error must be bad. WU is only 1450ish.

Poolside reporting is tragic -- looks like about 185ghs.

If someone can help me figure out how to make use of this apparent speed/reduce errors, I'll leave you a nice tip. I don't have enough knowledge and must resort to bribes.

Not every die is going to enjoy being overclocked. Use something like Bertmod to see which dies have disabled cores due to hardware errors. Using the per module, per die script, return troublesome dies to their original clock setting of 1D1 or whatever yours was. There's no strict rules. You have to experiment to see what works best for your machine. If you PM your Bertmod per module stats and your script, I can recommend clock settings.

[mekadeka]

you'll have to post some pics (and let the miner run for 15 mins first)

what you posted doesnt really make sense

Sorry for the delay. It might all be moot, because after a restart I'm unable to browse to an IP, although the machine is still hashing, according to the pool. I might need to restart everything.
Here's a shot from last night:

https://i.imgur.com/DN9x2Yl.jpg

http://imgur.com/DN9x2Yl

[mekadeka]

You need to tune each h board. Send me team viewer details if you would like me to take a look

Thanks, Elastic. Still working on being able to get to the miner GUI. I started re-overclocking it last night after a new router install. I don't have everything in place. Thank you for the offer. I'll take you up on it if I can get rolling again, and if you're still willing.

[mekadeka]

Not every die is going to enjoy being overclocked. Use something like Bertmod to see which dies have disabled cores due to hardware errors. Using the per module, per die script, return troublesome dies to their original clock setting of 1D1 or whatever yours was. There's no strict rules. You have to experiment to see what works best for your machine. If you PM your Bertmod per module stats and your script, I can recommend clock settings.

Hey, j03. Hope your weekend is going well.

I did try to use the cyper script you pointed me to several days ago, but I don't think I ever got it running correctly. It didn't seem to hash faster, but I'm certain my procedure or my tweaks were not correct. The bizarre speeds I got recently were from a single-line edit, but I did do something a little differently.

I really appreciate your offer to help, along with elastic's. I'll take you both up on it when I can get settled again. I had to put in a new router last night (before I started futzing with the Saturn) and not everything is perfect yet. Also, last night I did something with the miner (I don't think it was OC'ing) and I can't get browse to its IP. It is still hashing, although more slowly than stock. Hell, maybe it did mess something up.

[mekadeka]

j03, elastic,
FWIW, I got the miner up and running ok again. I have it going with just 211 until I get a chance to catch up with either/both of you. Thanks!

[elasticband]

Woke up today and founf one jupiter dead. btcguild shows it has been offline for 7 hours, so just round about when i went to bed.

The miner status page just shows STOPPED and no hashrate, restart does nothing, SSH screen - r returns no screen.

edit: so about 30mins later, several resets, and a downgrade to 0.981 and it is hashing away again currently at 580GH... phew!

edit: tried to go back upto 0.992 or 0.991 and the machine won't hash and just hands when trying to go to the advanced page. No screen is found when trying screen -x


edit: tried 0.098, same bad results. back to 0.0981 and miner hashes away. any ideas how i can get this back upto 0.0992?

thanks in advance!


update: th problem seems to have been narrowed down to the control board. when i run 3 boards, it works, no hangtime, shows the advnaced page, everything works. I connect the 4th board and i just get hangtime and no hash and no stats.

I took the board and flat ribbon cable and connected it to another jupiter with extra spaces on the ctrl board and everything worked fine.

update: board back in original jupiter, same problem. unplug board and jupiter works. plug board in, unplugg another board and miner works.
It realy does not like a 4th board.

[elasticband]

Well the above has finally sparked me into doing what I have wanted to do for several weeks now.

[mekadeka]

Well the above has finally sparked me into doing what I have wanted to do for several weeks now.

I don't think your machine is working correctly. The picture shows only 2 coins produced.

JK. Nice setup!

Your issues sound somewhat similar to mine. I did the SD card reset. It works, but the machine seems to like me messing around with it less and less, because the "can't browse to machine IP" issue happened again relatively quickly.

[corrow]

Your issues sound somewhat similar to mine. I did the SD card reset. It works, but the machine seems to like me messing around with it less and less, because the "can't browse to machine IP" issue happened again relatively quickly.

How to do a SD-Card reset? Seems I have a bricked KnCMiner...

[Powell]

First evening of an October Jupiter with 4 VRM's so many things I need to go back through.  I get the buzz noise even with adding voltage as well.  It calms them down but still HW errors.  I was only doing 201 and 221.  I need to buy some more fans tomorrow and retry.

[Biffa]

First evening of an October Jupiter with 4 VRM's so many things I need to go back through.  I get the buzz noise even with adding voltage as well.  It calms them down but still HW errors.  I was only doing 201 and 221.  I need to buy some more fans tomorrow and retry.

HW errors is normal, its the % of HW errors that is important, anything <1% is nothing to worry about. Also HW errors often start high and settle down over time, so give it a day and see what the status is.

[Powell]

Yeah it was that and the buzzing noise it started to make that I was more afraid of.  I didn't want to post until I was 100% certain I couldn't get it to stop.  Going to put a few fans to help cool the VRMs tomorrow and see how far she can go.

Also I had a lot of 10HW failures core disabled on a few more than I thought I should see.

[elasticband]

anyone got any clock settings above 271?

291 and going strong, HW dropping currently 1.9% after 5 min uptime  820/780/770

power is about 205w per board for a total of 820w

is this to much?


edit: 1.4 HW  790GH 

[DPoS]

anyone got any clock settings above 271?

291 and going strong, HW dropping currently 1.9% after 5 min uptime  820/780/770

power is about 205w per board for a total of 820w

is this to much?


edit: 1.4 HW  790GH 

I think over 200watts per board is asking for toast...   I might try for 185-190w soon...right now I am around 170watts running 221 clock setting

[elasticband]

I have been running 231 for a few weeks, then 271 for the last few days. I guess i will let you all know how i get on tomorrow morning.

I do have extra cooling and temps range from 41 - 48, i do think over 200 is pushing it. I am trying to tune it down slightly but it comes at a sacrafice of hashrate

[temen]

I have 1 asic on clock settin 371, other varies die by die from 281 to 361. One asic is better than the other. Power draw from bertmod is currently 456 watts. Temps 43 and 50 degrees. Remember hex = 0 1 2 3 4 5 6 7 8 9 A B C D E F so if you want to go higher from 291 then the next would be 2A1. After F you change first digit and start from zero. Like 2F1, then 301.

Currently 490gh/s stable november saturn 2 modules.

Changing to 381 doesnt give any extra gh/s for me, at 391 cores start to go off, didnt want to see how long it would have survived=)

[elasticband]

I have 1 asic on clock settin 371, other varies die by die from 281 to 361. One asic is better than the other. Power draw from bertmod is currently 456 watts. Temps 43 and 50 degrees. Remember hex = 0 1 2 3 4 5 6 7 8 9 A B C D E F so if you want to go higher from 291 then the next would be 2A1. After F you change first digit and start from zero. Like 2F1, then 301.

Currently 490gh/s stable november saturn 2 modules.

Chancing to 381 doesnt give any extra gh/s for me, at 391 cores start to go off, didnt want to see how long it would have survived=)

Your boards are november, mine are october, i don't think they can be pushed as far.

I was more worried about the PCI cables and extenders, but they are nice and cool.

[Powell]

I added my fans but even at 221 and tried different SPI voltage but stuck around 600gh/s but I'm sure there is more sometimes I'll get buzzing and a lot of hw errors.

4 VRM October Jupiter on .99.1-Tuning

I know all are different on OC settings but can't should ignore the random buzzes instead of stopping them.  With my Avalons and Antminers they are usually not next to me so if they make noise I don't kno lol.