Topic: Details around setting up a replacement S9 controller (Read 261 times)

Jumper jp4 makes it boot from the sd card. You could test Braiins OS by booting that way, then you can mix and match all the hashboards you want, Braiins OS doesn't care, you do the tuning yourself by observing the hash board perform at different speeds and voltages, and stick with the best settings you want (ie. efficiency or power) for each of the boards.

Just make sure chip ideal rate and actual rate match, this might take several retries (and cgminer restarts) for every hash board, as they all respond a bit different to each other anyway.

---

OK, I read the thread.  I might have missed which post should have jumped out at me.  What am I missing that should have been obvious?

Don't care if your a newbie - read the instructions in the thread.  Even newbies can do that.

Jumper #4 (the outer most one), forces booting from the SDHC card.

Which controls what?  I'm very green here.

Sorry, only worry about the ONE jumper - jumper #4.

Thanks, What does the number 1 jumper do/control?

The hashboards are fairly interchangeable, although if your using stock firmware you want them to be of similar quality.  e.g.  (1) 13.5 and (2) 14.0s may require you to slow everything down to 13.5 to keep that one card happy.  The control boards, presuming they are less than a few years old, are interchangeable.

Only worry about the 1 jumper.  I've never found documentation on the others.

OK, as you can tell from my profile I'm still new at this.  I tend to jump in over my head and figure I'll get the hang of things.  I have 2 S9 miners.  One (S9-0000) has a label that has S9_13.5T. The second one (S9-0001) has a label that reads s9_14.0T. Are the control boards interchangeable between these two units? How do I Identify what control board I have? What do Jumpers 1, 2, 3, and 4 do on these boards? (I've found no documentation so far)  Are different controllers able to work peacefully with different hash boards? The hash boards for the 13.5 unit has blue on the components (I think they are compositors.) The 14.0 unit has pink/red markings.  How do I know which hash boards I'm looking at?  If you need more info I'll do what I can to get to you. 

Thanks.

Some hash boards cannot auto tune. With the latest, Nov 17, firmware I was able to mix one non tunable board with new hash boards. The older board just stays at the clock start frequency and the other two auto tune. Pretty slick.

The spares I received from Bitmain include the SD card slot so you can recover the system if needed. It was loaded with old firmware which I just updated immediately.

The shipping cost is very high, but it's much less per board if you order several in one shipment. I'd suggest you order several controllers, data cables and fans at one time to avoid delays and to reduce per unit shipping cost.

Great to know.  Alas controllers are right at the break point cost wise where they are almost not worth repairing.  Especially if such a repair would keep a miner offline for weeks vs. just swapping it out with a spare.  Still, once idled, it might be worth it if the shipping was in-country.

That's exactly where the ZYNQ oscillator Y1 is, its an external oscillator for peripheral timing and sync. The bad (hot) one I have is a black IC that appears(the markings are damn indecipherable, all letters and a "zero" maybe) to be a MEMS clock chip. On another working board It appears to be a 33.33Mhz (CMOS?) Metal cap Oscillator. Not sure if they're interchangeable. Installed I only get one led(3.3v power), pulled I get that plus the front error LED.

Are those Ideal hashrates? I think they're practical max hashrates. As for the difference I'm guessing they pulled back 500GH for stability/reliability reasons.

Thanks.

FWIW:  I upgraded the firmware 11/17 auto-freq version and the unit came up fine.  It auto-tuned and all three cards are hashing just great.

This is the first time since I've started this hobby back in 2014 that I smelled a burnt component.  I smell located that to the corner of the controller near the jumpers and SDHC card slot.  This controller made it right under 90 days.

Good news is that I know now what to expect when swapping a controller out with one of the ones I bought from Bitmain:  everything will default to 400 Mhz, which is a decent test to make sure all the hash cards are talking, ASICs are accessible, etc.  Once that is established, a quick flash and things go back to normal.

Still VERY curious what determines the "ideal" hashrate displayed.  e.g. Why do some miners have an ideal of 14 TH and others 13.5?  If it was pure board performance, I would expect that to be a spectrum, like 13.65, 14.1, 13.92, etc.

That's extremely interesting. There shouldn't be any difference between the controllers the only difference between the versions I have is the oscillator for the ZYNQ chip.(unless you have very early Altera FPGA based S9 controllers). One uses a MEMS oscillator and the other one uses a CMOS one.

     As far as other US-based repair outfits, after myrig is fanatic26 here(tho he's got his own farm don't worry about), he's also helping me become #3( or 4 or 5 or 6...there are lots of competent techs here) . That task will be complete if he can help me get an ASIC test controller. Without that I can only fix power issues and physically blown ASICs.

Hello all.

By way of introduction:  I run a ~100 node S9 farm that I brought up in early December.  Prior to that I ran a small cluster of around (8) S7s and S9s.  Got lucky and bought the bulk (80) of my machines during the brief 14THs sale.

The vast majority of the machines run just fine.  I've had a few APW3++ power supply issues, but swapping those out for new ones generally fixes those problems.

I have also had a few hash card problems:  things that new power supplies just don't fix, like this obvious one where a couple of chips are bad:  http://puu.sh/zmqcQ/be27eff6a1.png

Anyhow, yesterday one of my nodes had a card go offline.  Happens.  Normally a reboot and all is well.  Not this time.  When I rebooted, the system came up, but never started hashing.  Upon investigation I found the single board speed test seemed to be hanging... and I let one pass run over 12 hours to be sure.  OK, no biggie, standard process of disconnecting one board at a time and rebooting should identify whatever was wrong.  Nope.  Went through all 3 boards with no change!

Well, if it wasn't the boards, I figured it must be the controller... and fortunately I received some spares from Bitmain recently that I purchased just for this case.  For reference, the original controller card was version 1.20, and my replacements are 1.30... if that matters.

In any case, I swapped the controller and powered up.  All three cards are now running fine, but at 400Mhz.

Also for reference:

Hardware Version   16.8.1.3
Kernel Version   Linux 3.14.0-xilinx-gb190cb0-dirty #57 SMP PREEMPT Fri Dec 9 14:49:22 CST 2016
File System Version   Sun Jul 30 20:19:24 CST 2017

Interesting tidbits from the kernel log:

Chain[J6] has 63 asic
Chain[J7] has 63 asic
Chain[J8] has 63 asic
Chain[J6] has no freq in PIC, set default freq=400M
Chain[J6] has no core num in PIC
Chain[J7] has no freq in PIC, set default freq=400M
Chain[J7] has no core num in PIC
Chain[J8] has no freq in PIC, set default freq=400M
Chain[J8] has no core num in PIC

Which explains the 400M clock speed.

These leads me to my primary question:  How do I get the controller to run its speed test and auto-tune these back up to their proper setting?

I tried using http://172.16.4.155/cgi-bin/minerAdvanced.cgi to set the starting speed to 550M, hoping that would kick start the process.  The 2nd time I tried that, it appeared to work, but all cards are at the fixed frequency I specified.

Guessing I just need to flash one of the auto-freq binaries, but wanted to check and make sure before I potentially trash a $60 (plus ridiculous shipping) card.

Secondary question:  Can all the controllers support both fixed and auto-freq binaries?  If so, is there any reason to run fixed on any of my older miners that came with it?  Related, is it then just the mix of hash cards that determines if a miner is 13.5TH or 14TH?

Final question:  Anybody know of a trustworthy US based repair shop?  I don't mind sending in individual cards for repair, but no way I'm I going to send something like that miner with a few bad chips off to China for a few months of travel, losing all that hash the whole time.  I've previously used BitmainWarrenty (now MyRig), but am not 100% happy with them due to what seems like excessive repair time (several months).