Bitcoin Forum

So I have 7 blades running in the standard back plane with the standard power supply.....So I blew a fuse (house earlier today). I removed the reset jumped so it would stay off and flipped the breaker.  I waited a few hours and when I put the reset jumper back in position I hear a SNAP which sounded like it came from a spark. 

Now I have two blades that do not power on or blink. One of them smells like burned electronics and clearly has a burnt component. The other blade doesn't smell but wont turn on.


What the hell would cause this? Are these things seriously going to fry like this in the standard factory setup?


I will post pictures in a few mintues... Thanks for any help!

Here is the one that smells with the visible burnt component (middle right side)
http://imageshack.us/a/img27/9685/kum6.jpg



Here is the one that will not turn on just like the previous one. But I cannot find any visible defects.

http://imageshack.us/a/img23/9409/ih83.jpg

Your first board has a fried buck regulator IC.  It's easy enough to replace, but most likely the board still won't work.  I had one of those same ICs fry when I mis-soldered a replacement feedback resistor for one of the VRMs, which resulted in infinite resistance leading to an overvoltage condition that somehow cascaded to something the buck regulator couldn't handle.  The good news is that your ASICs are most likely still fine.  The bad news is I don't know anything more that could help you other than the above, as additional context.  I ended up selling my failed one for parts.

I also have one that looks like your bottom one.  It had been hashing away just fine and now gives all xxxx's for the ASICs no matter which oscillator I use.  I also can't find an obviously-burned component or any causal event (there was a brown-out somewhere around that time, but I'm not sure it was coincident).

I'll be watching this thread to see if you get any more info.  Good luck!

My other two are doing just fine with 16MHz oscillators, though they run hot and are sensitive to ambient temps.  So far I've got 50/50 luck with these, unfortunately.

I've been messing with blades a lot the past few days tracking down issues. On the one with no power and no roast, first check the fuse to see if it's blown. If that's good, check the 3.3V output and see if that reads close to proper. Also check the VRM outputs and see if any of them are reading oddly high.

I think what happens when people are missing resistors in overclocks that blow the 3.3V regulator, is the output from the buck converter basically drops to maximum mode because the feedback reference pin on the regulator is tied to ground. This drives 12V (or very near it) into the ASICs. The ASICs also are tied into the 3.3V line, probably for logic-level interfacing with its paired buffer chip tied to clock and I/O. Likely what happens is the 12V destroys the ASIC and either backfeeds 12V up the 3.3V line or dead-shorts the 3.3V line through the melted ASIC which overcurrents the supply. So on ones with that problem, replacing the 3.3V regulator may only be half the battle - getting 3.3V back on the line will require removing the failed ASICs that are pulling it low. Even then you might need to pull any failed buffer chips that might be interfering with the IO bus.

What do the fuses look like when they fry? I cannot see into them so for all I know it might be fried.

With what SH said, I take back my guess that the ASICs might be alright.  Ack!

As for the fuse, the only way to tell is to use a multimeter and make sure there's essentially no resistance from one side to the other.  If it's blown, it will have essentially infinite resistance (open circuit).

Ok cool. Thanks for all the responses, I really appreciate it.

I just want to repeat though that Im really confused why this would happen and I'm actually afraid right now. I didn't move or touch the blades, I just used the reset jumper on the backplane! I've got $2000+ of these blades running and now I'm afraid more will fry for no reason like these did.

Have no fear or doubt, these boards are all going bad, over the past month alone i have seen about 50 threads pop up like this one, posted a few myself, am no holding about a dozen bad blades out of only 120 of them to start with. Every one of them has died in the past month.

Truth is, these blades were nothing more than a second rendition of a new technology at the time, and one without 10 million dollars to spend in stress testing and other types of duty and load testing beyond the standard manufacturers testing. They were great for starters, they had a few design flaws, and yet many of us ROI'ed pretty quick on them. Many who paid in excess of $500 for one blade will never see that investment back without bitcoin rising to $4,000 a coin, and over the next few months i expect to see at least 20-30% of the blades in existence suffer the same fate that hundreds of us are suffering now.

Good news is, there are thousands of new comers to bitcoin mining who invested in USB sapphire miners. The big fault with those is that the ASIC fries out. and the day after you find it on ebay and can usually pick it up for $10, i have come across a few dozen, just pull the chip off your board and solider away. 5-10 minutes later, you have 333 MH/s

We have all been searching for a PCB design for the AM chip to make something useful out of our dead blades, and across dozens of threads, nobody has found one, which is suffice enough to say that one does not exist. Even a small 2 chip USB board would be nice, so we could at least get some use out of our chips from dead boards. If nothing more then to give all of us something to tinker with. Hell i would even chip in on a bounty for someone to design one for us.

In the mean time, i cut my losses and broke down and ordered a few 200 GH/s prospector mining rigs today to keep my hashing power up as blades die and wait for pre-orders to arrive. i think the prospector is probably one of the few rigs you can order and receive within a week these days. a little over priced at almost $6,000 but to have it now is golden, i think it's safe to say, $6,000 hashing away in a few days now vs. what it will get you after all the march/april pre-orders arrive, the results will be exactly the same.

Just my 2 uBTC worth, troll away if you must.

If they'd done anything at all to improve thermal coupling between the heatsink and board would really have helped longevity, pretty sure.

I guess nobody's taken a logic analyzer to these things to reverse-engineer the IO protocol for the chips? Wish I had the time and equipment.

Well InfiniteGrim, i dont know about you, but myself, repairguy, and dozens more have had enough and have finally decided to do something about it as a community. Check it out https://bitcointalk.org/index.php?topic=382964.0 let me know what you think. as a community project, it would be a very viable cheap solution to our problems. If it dosn't take as a community project. i will fund it all myself foot the bill on manufacturing the PCB's then start charging $20- $30 a piece for them, If it takes as a community project, after talking with a friend of mine who works at / co-owns a PCB manufacturing company in Minnesota that said we could get the boards for less than $5 each with an order of at least 1,000 or more. that is of course once they are designed.

Quick question: Is it possible to just backfeed your 1 volt (or whatever it is the chips need) supply line from an external source? I was kicking around designing a board like that for blown BFL 50/60's, but increasing difficulty made it silly to invest in.

C

Should be possible. If you pull the FETs and inductor you basically fully isolate the output from the 12V line and you can do whatever you want. Each bank's VRM outputs are separate though, so you'd need to tap the board at 8 points to fully power it from an external supply.