Looks like it's time to sell my asic miners on Ebay. Fair enough; I doubt there is anyone here stupid enough to pay $40 bucks for one with bitcoin ($44-10% Ebay premium.) That would be .05btc; they are not going to mine that amount period.

Ah well. If anyone wants one for .05btc, drop me a line.

C

People are stupid?

People cannot do math?

Hype?

Hyperbole?

Basically everyone thinks that bitcoin will go to infinity so they all want to "mine". But in this case it would be better to buy coin than mine, so people will buy coin.

Which is good for those of us who have miners because there is less competition, and because the prices of our product go up.

This is why I tell people why I was smarter to mine than buy bitcoin. Which is exactly the advice to give people now.

Weird.

Hm. Should go a bit higher than that. Try CK's 1.2.5 pre-roll; it's further back in this thread. Tarkin's is nice but the problem is he turns off error checking on the cores, and since our cores have more errors, you wind up with dead cores being submitted as work which slows down and wastes power.


Going to 4-5 was not a problem, however I seem to have stalled. Either the last two chips I got were duds or something really went weird. I'll see if I can get another chip to try a different tack.

Try Ck's elf.

C

Yep. Pool rate. With CK's software I was averaging well under 1% errors, with others as high as 20%. My chips really were... 4.5gh tops chips :-)

I'm pulling between 19 and 20. Wish I could go higher. Drat.

Google says:

https://forums.butterflylabs.com/jalapeno-single-sc-support/4564-diagnostic-tool-bfl-com-port-scanner.html

And I don't think it does.

C

Interesting. I've got five chips running on my Jally right now and it's holding at 50c ever since I put a big copper heat sink on the bottom and used cheap-o Radio shack thermal grease on the AL plate to board interface+plate to copper sink interface. A lot of heat can be dumped from the bottom of the board; that I think is why BFL stuck heat sinks on the bottom as well as the top and put that side fan in.

I wish I could get my unit up to full blast, need a few more chips.

C

Nice job fixing it. Question since I am tormenting a poor little jally right now: Do you remember what the temps were, and do you have one heat sink or two in your unit? Did it have all 8 chips on there?

Odd. A co-worker had a 25gh unit and same thing happened. I recommended putting it in the freezer for 15 minutes, he did it and it came right up.

Given that I have been abusing the hell out of my 20gh jalapeno here by basically dumping a volcano's worth of heat on the board, I wonder what is up...

One way to find out, there is a utility over on the BFL forums, the handy dandy bfl com port scanner that works right with FTDI drivers and will dump out what the unit is thinking. Try grabbing it and giving it a shot.

Another thought on those lines, grab another computer, load up bfgminer, and see what it says. Do the D for display, then M to view all the CPUs and see what's up. Maybe it's conflicting with something else in your mining pile.

C

I try to explain that to people about the stock market: "It's not down 20%, everything simply is on a 20% off sale.". Unless the world melts down (which oddly enough I really doubt is going to happen) things will not be as bad as people think.

Probably. If you're compiling your own code it's over in std_defs.h,

 __ASIC_FREQUENCY_ACTUAL_INDEX   7

Actually ck's 1.2.5 firmware is my favorite. Right balance of tests and performance, got me 7.3gh on my jally.

C

Morning theory: Maybe the problem isn't in these new chips I am putting down, but with one of the existing chips I *have* put down.

Theory:
The BFL system seems to use three lines to control what address a chip has. Each chip socket is wired in a specific binary pattern, the system then selects a chip by putting an address on the bus. Maybe one of the chips I have put down and is working is masking that bus somehow.

Facts:
LEDs on back:
LED, status, engine number, engine ID from BFL
7 , on, engine 1, 0A 000
8 , off, engine 2, 0B 001
9 , off, engine 3, 0C 010
10, on, engine 4, 0D 011
11, on, engine 8, 0H 111
12, on, engine 7, 0G 110
13, on, engine 6, 0F 101
14, off, engine 5, 0E 100

The problem is I don't know how to map sockets to chips. Is this right?
Theory:
Pad, LED, binary, chip in place and working?
U3  engine 1 000 yes
U5  engine 2 001 no
U6  engine 3 010 no, does not work
U8  engine 4 011 yes
U7  engine 8 111 yes
U9  engine 5 101 no, does not work
U13 engine 6 101 yes
U14 engine 7 110 yes

The sockets which don't seem to work are U6 and U9. I have never tried U5

Maybe one of my "working" chips has an open address line and is masking these other chips. Not sure.

Counterpoint: If this was the case, then I doubt a "race" condition would always pick the same processor every time. Either I would see other lights coming on, or I would see a reduction in chips when I add the one that would overlap.

What's bugging me though is that my own reports show chip 5 didn't work, then more heat got it to work "sometimes", then running it at super heat got it online. Maybe I just was too timid about chip 6 there, it did come online for *one* power cycle, but did not repeat. So I never got to see how many cores it had. It was sinking power and generating heat, which is what the other chip I failed on (the one I didn't re-ball and is now probably welded to the board) didn't do.

Ok, 2:30am again. Took the board to 300F, and ran the wand to 460c, as high as it would go.

Hit the chip for 2-3 minutes at full temp. Cooled down, tried the chip on the board. Nothing. Tried removing the chip from the board, could not do it with the board on the side. Chip is not getting hot on startup, so it's probably 100% dead.

Oh well. I needed a space-holder there.

Next up is to perhaps find a new chip, and in my spare time see if I can re-ball one of these with a few torn pads. The other one is being sold to a bitcoin collector for .01btc. Till then I'll take a break, anybody got any BFL chips for sale?

C

Ok. Tomorrow I'll give it a "what's the worst that could happen" run up with 300F on the infrared pre-heater and all-out on the chip. And yes, the board has a very impressive amount of thermal conductivity, this is different from the boards of the 90's which were more along the lines of laminated fiberglass. It's pretty easy to overheat those and either delaminate the pad or burn the board beneath it.

Interesting. On the older boards one of the failures of component burnout was that the board itself would get hot and liquify inside, causing electrical shorts through the planes. This might not be a problem here.

In the meantime I went to the Ebay china crap store and bought a bag of lead solder balls. I'll give hand-reballing a shot as well, lead will melt at a way lower temp...

And BTW thanks for the advice and help. This is an experiment, if it turns out these chips are still usable I will be pretty tickled.

C

Hm. Well, let's go at this from the other end: At what point will these boards literally catch fire? :-)

I was just running 450c out of the unit at max speed for 3 minutes at that non-balled chip. Didn't work, it's on the board but not perfect. Damn damn damn. :-)

Yes, I hate ROHS solder in ways that would make a normal person go "hm". To be honest one way to deal with it on modern boards is to use Chip-quik to get the component off, then clean it up and use leaded solder to put the new parts on.

But what it sounds like is that the problem is not I have burned the chips, but that the damn thing just needs that much heat. One truth here is the more chips I have on the board, the more difficult it is to get the heat on the edges, that might be it as well. Or something like it, but man....

I'm going to re-group for a bit here and think: I tried setting the parameters to very... high... levels and got zilch. I have broken a few of the pads on two chips (including at least one of the critical communication pads), and a third doesn't seem to want to work even when low to the board.

I've also re-read the code, and I don't see anything in the chip init routines that would prevent the chip from firing up due to lower 12 volt voltage, so that's not it. I've tried running without heavy diags, and with disable engine zero on. So it's not a code issue.

Hm. I think I have two options:

1) Pre-heat the board to a solid 300 degrees F to warm it up, then use the wand for 120 seconds at 450c at that low chip. This is about as much as I can throw, and at that level I am running a serious risk of blowing capacitors off the board.

2) Pull the chip, and re-ball it at a much lower temp with 64 lead solder balls. Not much fun, but at least I will be working with a lower temp solder.

In the meantime I need to find another single chip to work on. So far I have knocked three out of commission, this is starting to get a bit expensive. :-) But that's part of learning.

Josh, if you were to use an IR pre-heater under it and a hot air gun what temps at the nozzle and the board would you use? What temps would fuse the chips internally.

C

I know. This is insane: I took the balls off the other chip (I didn't even overheat it) checked it, cleaned everything, checked with 6x loupe, then put the chip (sans balls, ok) on the board and applied heat.

375c for 60s, not enough to stick.
405c for 90s, stuck, good.

Plugged in, five chips. DAMNIT.

Ok, either one of four things is happening here:

1) God hates me today
2) I have managed to somehow either cold solder or overheat two chips in a row. Not possible; I have put five chips on with no problems like this.
3) I managed to get two bad unopened factory chips in a row
4) Something. else. is. going. on.

I wonder if the unit is checking the chips, checking voltage, finding the voltage drop below 12v when it enables chip 6, and it doesn't enable it. Time to check into the code, but either I completely suck as a human being today, or something else is on.

Note: I am still using the original power supply. Something's odd. Inaba there should be no way I am overheating these chips, and they're on the board right. Something else is going on.

Hm..... Meantime back to hashing at 20gh. I still have chip 6 the un-balled one on the board. Something else is going on.

C

Darn, still nothing, even at 450c for 1.5 minutes. Chip must either be dead or seriously confused.

Mining again, will leave it on and alone for now. Dead chip still on board, will watch temps and power draw to see if anything changes.

Oh well, you have the good days and the bad... :-).

C

Hm, darn another setback.

Got the two chips today and have tried putting them on. Chip 6 is pulling power but is not coming online; normally that means it hasn't completely flowed but I have re-heated it several times, each time with between 390 and 420c from the wand for 1.5 minutes. That should be more than enough, and was enough to get the last chip going.

And to add annoyance to injury after one heating all six lights *did* come on, then on the next reboot the new one went off. DAMN that means the chip is good, but not communicating. I'll have to think for a bit.

Crud. The chip is square, I'm going to have to think about this one for awhile before going further. Now it's hashing at 19gh, but pulling 150 watts total (computer plus rig) so the chip is drawing power. Drat.

Chip #7 went on the board, but floated down so it was off center. Had to take it off, and clean up the board from several solder fails. Drat again.

Need to pull the sink again to install the fan; maybe I'll try one more time to heat chip 6, then call it a day for today. Not sure why it isn't working, darn it :-)

Well, I know why, one of the control balls is not coming down. I wonder which one...

C

Yup. Sometimes you're lucky and it just nuked the FET. However since the FET gates on low voltage stuff like this are rarely driven by optoisolators and isolated DC-DC's, it usually takes out the drive circuits as well. IGBTs in high voltage systems are typically driven that way, which firewalls the damage to the IGBT, the resistors between gate and collector, and the um... circuit that snoops collector to emitter voltage and sends the HOLY FUCK EC voltage is going through the roof! WE HAVE A SHORT CROWBAR GATE NOW! thing.

The latter is usually what saves that 50kw fire that results when one side of your IGBT shorts. The crowbar sees the voltage from C-E screaming up (which means high current draws), forces the gate down to ground and lights an optoisolator to light up the "you're fucked" light on the dash.

Then if you're a super-special person, you say "No I'm not", bypass the crowbar circuit, fire up, and send me a box that looks like someone used a flamethrower in it. Been there, dealt with that. :-)

Indeed. looking at the specs for these FETs I think they have a solder pad under them that is where the heat's supposed to go. Hot air is probably the best way (400c is what seems to work with my Aoyue unit) to float these off. Not hard, just need the right tools. But they're small, so a pair of soldering irons one on each end might work as well.

Never dull.

Check out my other thread at https://bitcointalk.org/index.php?topic=336782.0

Note I'm tracking a possible problem with running a full set of 8. I would recommend not going past 6 without cooling those FETs, 7 tip tops.

C