KnCMiner Jupiter Miner First Impressions

[sickpig]

(so before luke implement the mechanisms to automatically disable and re-enable problematic cores),

I haven't looked at that implementation yet

here the link to the commit that introduce automatic en/dis core algo:

https://github.com/luke-jr/bfgminer/commit/8b690959dd4b107010217289fa69da82889e63cf

quoting comments:


 + If a core gets 10 HW errors in a row without doing any proper work
 + it is disabled for 10 seconds.
 +
 + When a core gets 10 HW errors the next time it checks when it was enabled
 + the last time and compare that to when it started to get errors.
 + If those times are close (50%) the disabled time is doubled,
 + if not it is just disabled for 10s again.

I have done similar test runs like you, and it is nice to just leave off the really bad ones and just keep on the rest, but it is also a wonder if even a 70% HW error core isn't worth just leaving on since it still provides 30% valid work

Does one terrible core effect another? Do HW errors tax the control unit?

hno on kncminer irc said that there colud be some kind of tourbolence, a sort of bad interaction between good cores and bad one if you keep the latter enabled. 

anyway this morning (UTC) I've discovered that my miners were running cgminer since sunday morning due to a temporary glitch in the power line at the colo that makes the miners restart. As soon as I realized that I've checked the cores status and guess what... all cores were enable again. this is strange.

Just after that I've stopped cgminer and restart bfgminer and after 12hs I get almost the same performance.   
 

If KNC would sell extra control units I would play around with a bad module but I am done taking downtime setting up different tests.  At some point I will RMA a few problem modules.

I do agree. It would be fantastic, imho

If you have good hardware you can run .96 at very nice Gh

I will give it a try, but I don't know if I really have good HW. I had a dead die on a asic slots for each machine. I was able to bring it back using Enablecore.bin on 0.95. This improved my hashrate at the pools from 480 to almost 520Gh/s at the pool.

I've tried 0.97 on one of them but the higher hastrate was balanced by a way higher HW error rate (10-15%)   

I put 4 of my best modules in one miner and get 545-565 on avg at btcguild

wow it would be good to gain another 25Gh/s per jup.

This is the list of the thing that I've to try next:

- (SW) ckolivas cgminer version
- (SW) bfgminer automatic en/dis mechanism
- (SW) test 0.96 to really see if I could gain another 20GH/s
- (HW) mod the heatsink fans as you did (lowering them)
- (HW) add two small fans in the front to improve the intake at BBB/VRM level
- (HW) improve the VRMs airflow take into account what "The Avenger" share here:
            https://bitcointalk.org/index.php?topic=170332.msg3375668#msg3375668
            maybe using something like trepex did here:
           



 

[1714626963]

1714626963

[1714626963]

1714626963

[1714626963]

1714626963

[AFox]

0.98 isn't working out for me.
Back to 0.97.

All the firmware's were run at least 24 hours without any reboot.
For that reason firmware 0.90 to 0.93 aren't listed.

--------------------	0.94	0.95	0.96	0.97	0.98
Average CGminer	270Gh/s	274Gh/s	280Gh/s	284Gh/s	278Gh/s
Average Pool	275Gh/s	263Gh/s	278Gh/s	283Gh/s	269Gh/s
Consumption	485 Watts	305 Watts	313 Watts	313 Watts	348 Watts
Temputure	54 & 62°C	42 & 47°C	42 & 47°C	40 & 44°C	44 & 50°C
Rejected	1.5%	0.22%	0.38%	0.48%	0.61%
HW	2.81%	0.79%	1.08%	1.47%	3.31%
WU	3975/m	3717/m	3933/m	4021/m	3825/m

0.94	0.95	0.96	0.97	0.98

[CYPER]

0.98 isn't working out for me.
Back to 0.97.

Do you have any idea why KNC removed 0.97 from their website?

[AFox]

0.98 isn't working out for me.
Back to 0.97.

Do you have any idea why KNC removed 0.97 from their website?

I wasn't aware of that
It's a good thing I keep a copy on my computer. Here's a link for those how want to download firmware 0.97 : http://1t4pro.1fichier.com/en/index.html

[CYPER]

0.98 isn't working out for me.
Back to 0.97.

Do you have any idea why KNC removed 0.97 from their website?

I wasn't aware of that
It's a good thing I keep a copy on my computer. Here's a link for those how want to download firmware 0.97 : http://1t4pro.1fichier.com/en/index.html

Maybe they renamed it to KnCMiner firmware 0.9.6.1?
Or KnCMiner firmware 0.9.6.1 is a different one?

https://www.kncminer.com/pages/support

[AFox]

Firmware 0.961 is a different one. It came out before 0.97.

[lenny_]

My Jupiter is using 700-800 Watts at the wall. The ASICs report using ~400 W. Anyone have an issue like this?

No.

In my Jupiter with FW 0.98, Bertmod reporting 572W, it's taking 668W at the wall, with OCZ 1000W Stealth PSU. Before, on FW 0.97, it was around 500W on BertMod and 580W at the wall.

What FW you're using?

[shmadz]

My Jupiter is using 700-800 Watts at the wall. The ASICs report using ~400 W. Anyone have an issue like this?

No.

In my Jupiter with FW 0.98, Bertmod reporting 572W, it's taking 668W at the wall, with OCZ 1000W Stealth PSU. Before, on FW 0.97, it was around 500W on BertMod and 580W at the wall.

What FW you're using?

0.98, Bertmod reports 555 W DC-DC. I use two PSUs, a 750W and 800W. Hoping to read maximum efficiency at 50% load.

Has anyone tried using 240V?

Haven't tried 240, but I'd recommend a nice 1000W PSU over using those other 2 *(not sure what they are, not sure if an 800 could power all four modules)* - but I'm pretty sure you're pulling more power from the wall than you would with 1 really good 1000W PSU.



Personally, I'm running a Seasonic 1000 platinum and my Jupiter is now reporting 564 at the web interface (been running 4 days now, was at 563 for the last couple days) -- pool reports anything from 520 to 600.

Back on topic, I have no idea how much I'm pulling, (I'm running 0.81) but the PSU barely even gets that warm. (If I had to guess I'd say around 10-15 degrees above ambient - regular mode, never tried the hybrid mode)

Bottom line. Even though I can't test the power at the wall, the low temperature of the PSU leads me to think that it is extremely efficient at whatever load it's pulling. I highly recommend this PSU as a great choice to power a Jupiter. (just my personal experience, and no, I'm not getting paid by Seasonic)

just a couple reviews:
http://www.hardocp.com/article/2011/11/01/seasonic_platinum1000_1_kilowatt_power_supply_review#.UnmnlhAyxaQ
http://www.guru3d.com/articles_pages/seasonic_1000w_platinum_power_supply_review,1.html
http://www.tweaktown.com/reviews/5097/seasonic_platinum_1000w_80_plus_platinum_power_supply_review/index.html
http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story&reid=264

[il_minatore]

overclock tips?

thanks

[RoadStress]

overclock tips?

thanks

I think 140GH/s per chip is the maximum, so not much room for overclock.

[Rampion]

overclock tips?

thanks

I think 140GH/s per chip is the maximum, so not much room for overclock.

I'm getting 144GH/s per chip without overclock.

[Sitarow]

overclock tips?

thanks

I think 140GH/s per chip is the maximum, so not much room for overclock.

I'm getting 144GH/s per chip without overclock.

Stability is also important.

Thus far firmware v0.9.8 has been stable.

[lenny_]

My Jupiter is using 700-800 Watts at the wall. The ASICs report using ~400 W. Anyone have an issue like this?

No.

In my Jupiter with FW 0.98, Bertmod reporting 572W, it's taking 668W at the wall, with OCZ 1000W Stealth PSU. Before, on FW 0.97, it was around 500W on BertMod and 580W at the wall.

What FW you're using?

0.98, Bertmod reports 555 W DC-DC. I use two PSUs, a 750W and 800W. Hoping to read maximum efficiency at 50% load.

Has anyone tried using 240V?

I am on 240V Europe. 668W on 1000W PSU on mine and yours are using 700-800W. You should read buy better PSU.

[RoadStress]

overclock tips?

thanks

I think 140GH/s per chip is the maximum, so not much room for overclock.

I'm getting 144GH/s per chip without overclock.

Ok then that's the maximum. Someone posted a formula with cores and speed, but i don't have it at hand, but you cant push chips faster than that because it's already their maximum.

[Zelek Uther]

My Jupiter is using 700-800 Watts at the wall. The ASICs report using ~400 W. Anyone have an issue like this?

No.

In my Jupiter with FW 0.98, Bertmod reporting 572W, it's taking 668W at the wall, with OCZ 1000W Stealth PSU. Before, on FW 0.97, it was around 500W on BertMod and 580W at the wall.

What FW you're using?

0.98, Bertmod reports 555 W DC-DC. I use two PSUs, a 750W and 800W. Hoping to read maximum efficiency at 50% load.

Has anyone tried using 240V?

I am on 240V Europe. 668W on 1000W PSU on mine and yours are using 700-800W. You should read buy better PSU.

I'm on 240V in New Zealand, my Jupiter setup:

• SeaSonic Platinum 1000W ATX12V power supply
• KnCMiner firmware 0.98.1
• 640W at wall
• Bertmod reports total DC/DC power output: 567W
• 550GH/s on pool as I type this*

* I'm super happy, I don't know why it's going so well right now, but I daren't touch it.  On 0.97 I was getting about 330GH/s.  On 0.98.1 the last few days the hash rate has varied, and I'm taking the precaution to reboot every night (as I did with 0.98).  This is because the hash rate at the pool dwindles down over time.  I would normally reboot now, since it has been a day since last reboot, but I can't bring myself to reboot since the performance is so good.

[DeathAndTaxes]

My Jupiter is using 700-800 Watts at the wall. The ASICs report using ~400 W. Anyone have an issue like this?

That seems insanely high how are you measuring wattage at the wall.

[RoadStress]

My Jupiter is using 700-800 Watts at the wall. The ASICs report using ~400 W. Anyone have an issue like this?

That seems insanely high how are you measuring wattage at the wall.

Kill-a-Watt

What firmware are you using? Same for waldohoover.

[DeathAndTaxes]

That is weird that there would be so much variance in reported power.


Most units are in the ~600W +/- 10% range and then there are some >700W?  That can't be explained by differences in power supply efficiency.

[CYPER]

That is weird that there would be so much variance in reported power.


Most units are in the ~600W +/- 10% range and then there are some >700W?  That can't be explained by differences in power supply efficiency.

There are PSUs with 92% efficiency and there are PSUs with 80% efficiency, so that might be the case.

[BTClobsta]

how are you guys powering off your unit? i just got mine today. when i first got it going i was hashing at 667 and my pool also showed 667. i did a power shut down by turning off the psu and changed it to some other psu. then powered it back on and now i'm at 567 and pool is at 399. anyone having the same problem?