Topic: [CLOSED] ASICMiner Prisma 1.4th/s - 1.47 BTC - page 9. (Read 49674 times)

The one thing I did notice is that these are not enclosed like the Antminers S3, I have an idea for those of you running Prisma to add a layer of safety. Perhaps you could find an old screen and screen in your miner. This way the air could flow, yet any exploding caps would be contained. I have a screen door I will use above the unit, while I continue to test my miner.

Well it is a lot of heat and current in a small spot. I mentioned arcing since the metal plate look like it had a welding burn.

Fact is it is an open design with a downward board using a lot of power.  be careful guys.

I am getting a solid 1440 on the site it is mining on = mmpool.org

using 1244 watts

I designed a better bottom cooling fan. the psu  has shorting protection and is on a gfci circuit.

shout out to the cap poppers what were the psu's on your gear.

I agree, 12v is unlikely to arc over this distance. I just measured where the miner was and the burnt area was about 1.7" above the metal plate. For the record, the metal plate was a last minute decision purely because of seeing the cowboy miner incident.

This is the one place that a arc may have occurred. It's the backside of the board but I can't tell if it was caused by the fire or if it caused the fire.


I did tighten all screws before putting the miner into service and it was running (unstable on a be controller) for a few days before the incident. I also didn't see any missing caps before installation but I cannot find the one missing cap anywhere. From the looks of the board it either fell off between tightening the screws and installing the unit or came off with the heat and disappeared somewhere. All that is visible now is it's solder pads. CrazyGuy and AM are taking care of me warranty wise, no complaints there.

If anyone's interested in the chain of events, I found 15 GPU's offline and no prisma boards showing up on the be controller which wasn't all that uncommon. I went out to see what happened and could smell the all too familiar burnt electronics smell however, it was somewhat faint due to the high airflow of the area. I found a 240v breaker popped, couldn't figure out where the smell was from so I reset it hoping for just another well done GPU and checked to see what was still working. That's when I took the screenshots of the one board with 0 hash rate. After that I powered down both prismas and found the damage hiding underneath one.

Also for those interested CrazyGuys minera image works great on a B+ RPi.

Arcing is not possible in the senerio.  The voltages are WAY to low. You would need hundreds of volts or even thousands to jump even a small amount.

http://en.wikipedia.org/wiki/Paschen's_law

This was current runaway.  These chips are most likely a "chained" design used to obtain maximum efficiency.  If one chip in the series "chain" shorts internally, it can

cause a great increase in current through the other chips in the chain thus causing the temperature to rise suddenly and dramatically.  This would literally set them on fire!

Somewhere on here I have seen a picture of a heavily burned "one string miner" that used Bitfury chips.  These use 14 chips and run directly off the 12Vdc feed in series.  

So if one ship shorts, it starts an overcurrent condition that very quickly pops all of the others.

"Mister InvalidSnack" ...lol. 

At least you sent me to bed with a smile on my face, Mr. CrazyGuy.  (Or do you prefer Dr CrazyGuy, PhD?) 

And as for the 5 MW facility --  Phillip you read my mind, I've literally spent the last few days reading about that and looking at the pictures you referenced and going..."well, I only managed to burn up 2/3 prismas, and the antminers were safe...so it LITERALLY could be much much worse."  But this wasn't supposed to be a pity party for me -- I was (perhaps prematurely) trying to alert people to a possible dangerous flaw ... and as I told Mr. CrazyGuy, it's entirely possible I just saw that picture of the other users' "blast pattern" and thought "oh fuck, you've gone and done it now...you never mentioned this and you know his post is going to say something about 'losing seven family members in the blaze,'"  -- in other words , guilt might have turned me into Chicken Little.  It's too late for me to really tell -- need to sleep on this.


But regardless, I would still strongly suggest people "keep an eye" on these miners.  Mine worked fine for 3 days.  No irregular temperature readings anywhere.  (Well, it was probably "irregular" about the time the fire was eating everything...but you get my point.)  I followed Dogie's guide so every screw was tightened, checked every cap to see if it was wobbly -- so , just, "be mindful of that middle area, and keep checking it for more than a few days before you start assuming everything's fine." 

There, those are my adjusted, far more reasonable "suggestions" for users.  Off to sleep now -- if I wake up and realize I'm a moron and embarrassed myself for no reason, I will go back and remove the more "excitable" language from my previous posts...because I never meant to cause a panic or fuss.  G'Night folks.

I did contact Mr InvalidSnack to get more details on what happened to his Prisma in an effort to prevent this from happening to anyone else. The response was very long, and I have yet to parse it all, but one thing I did gather is that the units were stacked on top of each other with a bumper in between each handle. Here are some tips I gave him to follow upon receipt of his RMA unit from Canary/AM.

1) Don't stack the units. I think it's possible the arc may have been board to board in Mr Snack's case.
2) Lay the units out on a non-conductive material with at least half a foot in between each unit
3) Make sure your units are in a cool environment.

As always, tighten heatsink screws upon receipt and visually inspect each board for damage during shipment. Use an IR gun to measure board temperature and notify your seller if you are seeing anything unacceptable.

I apologize for not being clear.

I purchased sheet metal from Home Depot 3 days ago, in anticipation of the NEW RMA REPLACEMENT Prismas -- because I wanted to make sure there was *something* fireproof and projectile-proof beneath the miner this time.  (If metal is a bad choice, then I made a bad choice.  But that's not the point.)

The fires do *not* have this in common -- because my fire was 2 weeks ago, not 2 days ago.  When my unit caught on fire, it was sitting on a piece of wire shelving, which was in kind of a "sandwich" -- two pieces of wire shelving with some "slices" of thick PVC pipe between them.  In other words, the Prismas were on NON-CONDUCTIVE, plastic-coated wire shelving -- suspended a good 3 inches above the MDF/whatever shelf below.

The fire was *not* caused by an arc -- there is no metal near any of my miners -- it's actulally all stacked next to the wall, because I was hoping to try to use it to PROTECT in the future against this kind of thing -- and if that's stupid (which apparently it is) then so be it, but please understand the fire I discussed involved *zero* chance for any "arcing" of any sort -- unless it occured within/between mining boards.  No metal was present and more than adequate airflow was available to the bottom unit.

I will try to post pictures tomorrow--apologies, its' late in my timezone and I need to be up early.  But I am eager to "make amends" as I told Crazyguy in a series of questions he just asked over PM, which I gave him full permission to quote or paraphrase as he likes....whatever needs to happen in order to make sure nobody else is potentially put in danger.

Again, and I apologize for beating a dead horse but I regret sending the wrong message in my previous post: The first guy's post, where he shows the metal plate beneath his miner... whether it's a "best practice" to use metal or not, the fact remains:  that metal plate saved his house from burning down, assuming the flames happened when he was asleep or not at home.  It deflected / blocked the exploding capacitors and other components that (in my case) simply bounced through the giant holes in the wire shelving and were hot enough to set a slice of PVC pipe on fire...along with parts of the shelf itself.

It is not a "metal arcing issue" that caused this damage -- though that might be a valid concern, I don't claim to know otherwise.  It's just *not* the reason this particular "blast pattern" happened.  By all means, use ceramic tiles to protect your shelving and not metal -- I couldn't find any light (and not super expensive) tiles myself, and I didn't want to triple the weight of my shelves with the cheaper heavy tiles...  And other devices (like the Antminers) have a metal exterior shell, so I thought it would be possible to "box in" one of the Prismas to keep other things safe.  Maybe this was a stupid idea.  Unfortunately, stupid ideas made *after* the fire don't really serve to explain why the fire happened in the first place.

I would write more, but like I said I just wrote it all to Crazyguy who in theory was asking so he could share the info with you guys.  -- oh, and please don't misunderstand the "tone of voice" in my message here, I'm trying to type quickly and sometimes I come across like an ass or argumentative, and I assure phillip, you I don't disagree with your statement that metal might not be the best firepoof material to use -- I just wanted to clear up the misunderstanding that led you to believe there was any metal, anywhere near the fire that I experienced last month.

I hope that makes more sense this time?  Crazyguy can you help me explain this better?  (I'm tired as fuck and possibly not making any sense to anybody...hopefully CG can swoop in for the rescue and put this into normal human words that people can understand ...sorry for the confusing first post.)

hey is it possible to get 1600gh with this? never mind it dropped to 1500  3.5% error 40 C due to garage door opened a bit. watts are 1244 at freq and 1500 gh or .8293 watts a gh at the plug





watts at 1244

top of the line psu evga 1600 watts platinum




opened garage back door let in some cool fall air.  fan brings in cold air



a second fan a delta ebay 9 bucks .   and I spaced the long tube since we don't want our long hot tubes popping from the heat do we.





one fan for the middle burnout issue



let the hash crank up

Yes, ssh

You can do it straight from the user interface by clicking on system->open terminal. Or use putty or some other ssh client.

That is solid advice. Please remember, you are running a 1200 watt device with exposed circuit boards .Two things I would highly reccomend not doing.:
Touching the exposed circuit boards, or putting them in close proximity to conductive material.


Also, I have just paid out compensation to batch 1-5 for the late shipment from ASICMiner. Thank you to all that participated in this group buy.

it is running pretty well  .

  is shell in the same as ssh.  I am more of a gear head then a coder.  I would be very likely to kill my sd card.

as for the second person talking about burning up a tube ..   both had the same thing in common close to a solid metal plate.    a  power surge then an arc to the plate then pop.    as a point of safety do no rest your tube miners near metal plates.  use other safer ways to stack them
  it is common sense an open set of ciruits .25 inches away from a metal plate = danger.

look at the tube design .  I told more then one person do not use it on metal. I used wood.  now on this the long tube I rest it on

stone.



it will not have any metal close for it to arc to



runs very cool with second fan

I had to modify minera to report the correct total hash rate for prismas. This modification was made after your unit shipped.

To get the latest code perform the following steps

shell in and change directory to /var/www/minera

git fetch --all
git reset --hard origin/master

Wow, okay now I'm a bit concerned.

I had the exact same damage -- in the exact same area your picture shows.  In my case, it actually started a fire which took out a second Prisma (blew all the caps off one side , popping like popcorn kernels before I could hit hit the power and grab the extinguisher).  But I didn't want to panic people and told ASICMiner and Canary I was assuming it was a fluke, and was focused mostly on getting my RMA completed.   I'm really glad you're safe and didn't burn your house down, man -- because in the back of my mind I've been thinking "what if the problem *ISN'T* a fluke and hits someone else, when he's not home?"  

I told Canary (who presumably told FC) that I wasn't really interested in trying to "publicly embarrass AM" and was happy to keep things between us -- so long as two conditions were met: (1) they replace the damaged Prismas in a timely fashion (which didn't happen unless you consider weeks to be "timely") and (2) "unless it appears to be a widespread problem and not a random component failure."

So both #1 and #2 have hit at this point;  when i saw your pictures in my browser, I swear I thought "Wow, FC is posting the images of the damage that happened?  That's really honest of him..."   I mean, it's like looking at a duplicate of mine ... though you seemed to get things under control faster than I did (the fire charred the entire bottom of mine, and there were no capacitors "unpopped" on the Prisma where this damage occurred).

So let me be clear since I've been silent (and OBVIOUSLY should not have been)  -- I can't calculate the odds of *two* random manufacturing defects causing the *exact* same damage in *exactly* the same spot (center -top of one board, or center-bottom depending on how you look at it, and causing enough damage to affect the adjacent corner board as well).  Maybe the odds aren't so bad.  But I'm inclined to think that there is something amiss with whatever component blew in both of our Prismas -- is it a vreg?  

All I can say for certain is that if I had not been home (just arrived back from errands about 5 minutes before I heard the alarm) and if I hadn't put a smoke alarm on my mining "shelf" (the main one in the hallway took a good 10 minutes or so to finally go off) then there's no doubt in my mind that the fire caused by a Prisma just like yours, would have done an *enormous* amount of damage.  Even putting it out as fast as I could, it took practically all night with all the windows open and like 5 box fans trying to pump all the smoke out of my apartment and it took all night sifting through charred parts of the wire rack that it *melted* before I could finally get enough of the debris out of my office ...but then I took a shower (and I guess washed the smell out of my nose) because it was sure as hell still there...  Took *days* to finally stop smelling like melted plastic and...

Anyway, I'm not going to lie here -- I'm a mixture of embarrassed and sincerely horrified to see another customer affected by the *exact* same damage, which I could have sounded the alarm about ~2 weeks ago -- and I'm so sorry to you, and so very glad that you had a metal plate under your Prisma to contain the damage and not lose things in a fire.  What's more, you're doing *exactly* what I should have done (and didn't) -- I sent my pictures in and asked for an RMA, and *you* posted the pictures here and warned people to help avoid more damage/injuries/etc.  

I've never met you before in my life, but you are the better man (or woman) and I am pretty certain I'm a selfish piece of shit right now.  I apologize to you, and to everyone else for not saying something sooner -- and I hope others will *please* take this threat seriously and put some kind of metal protection around/under their Prismas...


...and to be honest,  I feel like someone just punched me in the stomach -- I know I should probably post more and add details and etc, but I need a drink or something.  I don't know how I would have lived with myself if someone had been injured, or worse...  I don't know what the fuck I was thinking assuming it was a "fluke" -- that's not my job to determine, I have no qualifications to assume that, I should have done like you just did....  Fuck, I ... will come back and post more if needed, but I'm literally floored by this and -- I'm just VERY glad that nobody was hurt by my failure to report this, and I fear I'm just repeating the same shit over and over...so again, let me come back to this and post a bit less freaked out a little later, but I'm truly sorry and very ashamed of my inaction; my main goal in writing this (admittedly hastily and shittily written) message ASAP was just to make sure I got that message across as soon as I saw this had happened, but I apologize for not being able to cobble together a slightly better post than this.  

Friends/fellow miners, please take this danger seriously. While I don't have any way to know how many Prismas are affected by this, it seems hard to imagine that *only* our two are capable of this very specific type of failure -- so it seems clear to me at this point that it's not an isolated problem.  (Well, it's isolated in that it apparently happens in the same spot ... but not isolated in terms of people, or Batches, or etc.)  And for the record, I never overclocked *anything* -- everything was running at stock speeds, and running well, for about 3 days, with zero warning signs before it happened to me.  I can't believe I didn't say something sooner.



PS: The exploding capacitors were so hot, they actually *embedded* themselves into areas all over my office.  When I was finally getting the mining rigs (other than the Prismas) up and running the next day, I found several of my S3s were seemingly dead ... but it turns out they just all had dead network cables.  I swapped them out and didn't notice until later that day that somewhere on each of the 3 ethernet cables was either a cap (melted halfway through cable, through the exterior protective sheath, embedded right where all the individual data connections are) or a small "cap-sized hole" melted far enough into the cable to leave a recognizable impression (and a clear view to the data connections that it had severed).  So this probably is the most obvious statement in the world but: "The exploding capacitors can and will start fires or damage other equipment," it's vital people take serious care in terms of *where* they choose to store Prismas, because they will want something protecting other equipment from damage if it's all on the same shelf, for example.

okay running 2 boards   poolside reports 690gh

minera has a low number


 two fans running handles out two boards  no worries. 

No, it adds 1 for some reason. --bet-clk 22 is 230, 23 is 240 and so on.

Edit: https://github.com/ckolivas/cgminer/blob/master/ASIC-README

I would guess --bet-clk 22 is freq 220 correct?

Just update the startup parameters on the settings page.

Default is:
--bet-clk 23

changed it .  do I ssh this to change freq rates?