Topic: [ANN] Spondoolies-Tech - carrier grade, data center ready mining rigs - page 74. (Read 1260421 times)

Personally, I am troubled by the obviously burnt PCIe power connector. Pulling 250+ Watts through a single connector when running near full speed means that all the pins must be a tight fit, all the crimps to the pins must be solid, and the cables themselves should be at least 16AWG (IMHO). That applies at both the SP20 end, and the power supply end (assuming it's modular). If things are bit off, then you'll get heating that will just spread to the other wires and pins in the same connector.

I wouldn't ever try and pull more than about 200W through a single connector, on a long term (i.e. days) basis.That's just me though.

I would also encourage you to use the the "Power Limit" values for each of the loops. The SP20 seems to pretty well adhere to the limits you specify.

I reduced power to the "burnt out" unit, and it's hashing with all loops running @ ~1.2TH/s, with quakefiend's replacement board.

I power cycled the second unit, with no changes to any voltage settings, and it came back up with all loops hashing. I'll keep an eye on it, and lower the voltages if/when the next time it decides to weird out on me.

TBH, I'm actually enjoying this troubleshooting Love how tweakable these Spondoolies rigs are.

( As I mentioned in an earlier post, I figured I'd keep an eye on posts in this thread given that there were mysterious page counter decreases.  The difference in posts between June 14th and June 28th (today) are all attributable to pickaxepete deciding to wipe all of their posts, everywhere, and editing them all to being blanked out ("Edited") in those threads where he could not.  In this thread that was a total of 11 posts, all on or before page 25 (end of March, 2014), with nothing remarkable about them content-wise. )

Did you try to reduce power to all chips, reduce fan and see if fourth loop revitalizes itself.
It did happen to me upon similar situation.
My settings:

0.642 starting voltage
0.647 maximum

Power: 165W per loop (sometimes adjust one or two loops down a little if it/they overheat(s)
Fan:10-20

produces ~1230GH at ~680W (at the wall)

This is exactly how one of my SP20's is now.

I am still able to get about 1th from the unit without too much trouble though.

Hah ! Just my luck. A second SP20 I have has started to give me grief. Will troubleshoot tomorrow.

I need some new mining gear. Badly

yeah ¾ alive is a good way to think of it.  

@ xian01 let us know what sp-tech tells you

BTW on the bright side diff looks to drop just a bit which makes the 3 boards a little better come this sunday.


I have seen that happen with my seasonic 1200 plat

the fault could be in a loose pcie connection  thus arcing.

I have had cablez make up some seasonic cables and it never happened again.

Replacement board came in and works well. Thanks quakefiend.

Now as for why the board went kaput to begin with... well... not sure why this happened as it was underclocked to 1.3GH/s Was using an AX1200 to power it.

No other obvious signs of damage on the board.

---

EDIT: After more tinkering, I'm starting to think it's the controller board. Was getting a "PLL A" on one of the loops after swapping in quakefiend's board, and troubleshooting with swapping the cables to each hashing unit. Managed to clean up the "burnt out" socket on the board I yanked, plugged it back in, and got the same "PLL A" issue on one of the loops.

The zany thing is if I power cycle it enough times, it'll eventually work with no "PLL A" on one of the loops (even with the "burnt out" board) and seems to be hashing fine ATM.

Anyone have an extra controller board they might be willing to sell me ? Would like to tinker around some more.

And just so I'm clear, is it safe to assume LOOP 0 and 1 is the board on the left, and LOOP 2 and LOOP 3 are the board on the right, when facing the ethernet and power jacks ?

BTW: Has anybody got a spare PSU for an SP10 for sale?

Purchased a replacement board from quakefiend420 for a reasonable price. Will be sure to look at the bad board once I've swapped it out and see what's what.

That's unfortunate. You can always look at it as 3/4 alive, rather than 1/4 dead. Should run cooler and use less power. 

I wonder if at this late in the SP20 life cycle, Spondoolies will accept just a blade, and not require shipment of the whole miner and the like.

Depending on what Spondoolies says, and how adventurous you are, you might pull the 1/2 broken blade and examine it for obviously blown components and the like.

Even with running very conservative numbers. Sorry to hear it!

Yep. Even swapped the PCIE cables for those loops She's dead, Jim.

The main problem is that the change would require a hard fork. A hard fork requires a consensus (otherwise it's suicide) which isn't easy to reach for a coin with a $3 billion dollar market cap and thousands of users. With that said, I still think if the block size increase proposal is refined enough eventually everyone will be on board.


AFAIK that's not true. I don't know of any altcoin that comes close to the tps of bitcoin. I'm almost positive that if any of those 60 second block time coins gained the amount of users bitcoin had, we would be seeing coin breaking issues with network propagation.

did you check the pcie wires carefully?

If you're interested in a replacement hashing board I have a few from machines with controller issues.  SPTech wanted too much to replace the controllers, but the boards are fine.  The warranty on the SP20 is only 90 days, so I bet you'll be in the same boat that I was in...

Well shit.



First SP20 that's given me issues. Just up and died on me after working faithfully for several months. Sent an email to SPTech via their website "Contact Us" link and hopefully I can get a new board or something...

The Great Block Size debate is one of the reasons I'm more into alts. Bitcoin has too many egos and not enough desire to be mainstream. It's obviously a problem, so what the HELL is the big deal? They could easily raise it incrementally. This is an engineering problem, not a political one. Personally, I'd be in favor of keeping the size as it is and decreasing the block time. Every solution involves a hard fork, and the current confirmation times are (to put it nicely) suboptimal. A great many alts have proven beyond a shadow of a doubt that fast block times decrease backlog and do NOT kill the system.

If Bitcoin does not or cannot grow and evolve, then it deserves to be beat out in the marketplace.

my issue was with the "process" if from what I gather they 'agreed' with the china position that with the 'great firewall of china' and poor internet that 20mb was TOO BIG
a jump..which seems to be the case now ..WHY WAS THIS NOT SETTLED IN PRIVATE information gathering on the process of dev and code improvement..no dev egos are
involved so as soon as someone thinks (ego wise) this shall be so...it hits twitter without enough sober research in the background ..public ...but working towards consensus

can you imagine the btc coin price now if this was discussed/and consensus was reach without all this FUD and press drama/twitter etc  for the last 6 weeks...and just announced here as a done..I mean debate open in all this is the norm...but I mean really ..this was so bush league....

The devs at least in the open source manner of public discourse I think at least get ALL THE INFO before saying "such shall be so" and  consensus is reached and then discuss it...a valid proof of working method towards problem solving...taking extreme positions in public seems dumb when it seems to me they never even had all the facts to even base this on
assuming the china info is correct and anything more then 8mb would be problematic

just not the way to state your position on code change..... very worrying imho...hopefully they have learned something about how to approach these dev questions as
a more adult process now in the more towards consensus less jumping on twitter and other things saying my way or the highway ALL OF THIS SHOULD BE PUBLIC but
hell would we have really paid all that attention to it ..if they at least had gathered all the facts first and then did it in a less inflamatory manner ...same result less drama imho

my 2c worth

I've posted the original link because of my strong disagreement with the process Gavin and Mike tried to force, e.g. Bitcoin-XT
I think that such a drastic change should be made in consensus by the core developers and then agreed upon by the big mining pools before implementing.
There should be a block size increase, but it shouldn't be enforced "from above".

Guy