Hacking a BFL Jalapeno to 32GH and beyond....(???)

[lightfoot]

True story! I had the bottom plate on mine as well... didn't want to have to clean and regrease it... too lazy to do it right the first time, so ended up doing it twice!

I'm using an IR station, but I concur... bottom heat is critical to a good chip installation.

Hee hee he... It's amazing how much use I am getting out of this dopey little pre-heater. Using it on big boards is much harder, but on the little single boards it does shine. And when I reflowed the chip the result was perfect (and now it's at a solid 60c, I think the heat sink was being lifted off other chips due to the high chip)

I thought about buying an IR station, but then I realized that the trick to using air is to lower the blow speed down so much it's almost like a little cushion of hot air over the end of the heating element. Then the air stays at the right temperature, doesn't go everywhere blowing things off the board, and acts more like what you see with IR. Using the thing as a blowtorch is not the idea.

Live and learn. Now I want to get a Monarch to take apart, but $2,000 is a bit rich for my taste. :-)

C

[Delarock]

Sending Lightfoot a 25 GH BFL single tomorrow with some PSU's. This unit is underperforming at ~23 GH, so I figured it would be best to get the expert under the hood to tinker and see what he can find, maybe add a few chips into it.

Post pics, tell us what you find. 

[lightfoot]

Sending Lightfoot a 25 GH BFL single tomorrow with some PSU's. This unit is underperforming at ~23 GH, so I figured it would be best to get the expert under the hood to tinker and see what he can find, maybe add a few chips into it.

Post pics, tell us what you find. 

Will do! I'm looking forward to it; one question to start with is do the singles contain the supporting hardware to add chips. If they do it's a matter of adding a heatsink (of which I have... several :-) and chips to boost performance.

Should be interesting.

C

[bcp19]

Sending Lightfoot a 25 GH BFL single tomorrow with some PSU's. This unit is underperforming at ~23 GH, so I figured it would be best to get the expert under the hood to tinker and see what he can find, maybe add a few chips into it.

Post pics, tell us what you find. 

Will do! I'm looking forward to it; one question to start with is do the singles contain the supporting hardware to add chips. If they do it's a matter of adding a heatsink (of which I have... several :-) and chips to boost performance.

Should be interesting.

C

Looking at my Little Single I would have to say no, there is no supporting hardware.

[Photon939]

BCP is correct. I have a little single with Rev A chips running at 31GH/s, but the other half of the board is almost completely unpopulated. The vcore regulator is unpopulated with the exception of some electrolytic capacitors. It's not entirely impossible to get it going, but you'll need to make a decent order from mouser/digikey to get anywhere.

[Delarock]

I figured this would be the case, but it's a good learning experiment. As far as I know, there is some issue with this unit where it was hashing at ~26 at one point. No clue how to fix it. As far as I'm concerned, what goes on in between the side plates is pure magic. That's why I put the screwdrivers away and went to the post office.

This is possibly the most fascinating thread on the boards, so I wanted to participate somehow.

[lightfoot]

Interesting. I had another unit in here, little single, with a complete load of everything but the hashing chips on the other side. Even had the 1 volt power regulators, no problems adding chips if we wanted too. I guess BFL made some units with half a load, and some with a full set of supporting chips.

C

[lightfoot]

And in the really interesting news, I had a unit come in with a power supply that went *bad*. Jalapeno, ran it with 3 chips, checked in and saw that it had a massive amount of errors going on. Shut it down, put it on the bench supply, worked fine. Tested the power supply and I see 60 volts between negative and AC neutral.

Glad I have an isolated power supply on my miner. I'm wondering: Does BFL want the old power supply back when they RMA, or will they just send a new one out? I would really like to "interrogate" this one and see if it has the same blown caps as the one from England.

Good news: No damage to the jalapeno. In fact it's now hashing at 14gh with a second added (client provided) chip, much better than the 5.4 it came in with (it has two lousy chips on it and two very good ones).

C

[bcp19]

And in the really interesting news, I had a unit come in with a power supply that went *bad*. Jalapeno, ran it with 3 chips, checked in and saw that it had a massive amount of errors going on. Shut it down, put it on the bench supply, worked fine. Tested the power supply and I see 60 volts between negative and AC neutral.

Glad I have an isolated power supply on my miner. I'm wondering: Does BFL want the old power supply back when they RMA, or will they just send a new one out? I would really like to "interrogate" this one and see if it has the same blown caps as the one from England.

Good news: No damage to the jalapeno. In fact it's now hashing at 14gh with a second added (client provided) chip, much better than the 5.4 it came in with (it has two lousy chips on it and two very good ones).

C

I've heard both... apparently the right and left hands aren't talking... so maybe it depends on if you are wearing matching socks and Mars is in the shadow of Venus...

I honestly can't see them needing the PSU back, it's not like you could fix it and reship it.

[lightfoot]

I honestly can't see them needing the PSU back, it's not like you could fix it and reship it.

They are total crap. Crap crap crap. Good for nothing but cutting the power cords off for connection to real power supplies.

I'll check on the BFL forums. Still do not have my PCIe power cable whips.....

C

[lightfoot]

On another side note I finished reviewing a blown board that was given to me which had a complete blow-out of the 1 volt FETs due to a short.

Micro-summary: It's destroyed. Not because of the FETs; I got those off easily and cleared the shorted 1 volt lines. It's the fact that the whole 3.3 volt line is shorted as well; I was trying to figure out where that went bad and found out there was a nice bubble in the board by the chips. Apparently the 1 volt line not only shorted but it burned inside the board, shorting the 3 volt supply to ground.

Not good :-) So if you blow the FETs, repairs would require board-level digging, not something I would do for a $200 part. I may fiddle with it some more, but for the moment I will classify this as "toast". :-)

C

[synapse]

On another side note I finished reviewing a blown board that was given to me which had a complete blow-out of the 1 volt FETs due to a short.

Micro-summary: It's destroyed. Not because of the FETs; I got those off easily and cleared the shorted 1 volt lines. It's the fact that the whole 3.3 volt line is shorted as well; I was trying to figure out where that went bad and found out there was a nice bubble in the board by the chips. Apparently the 1 volt line not only shorted but it burned inside the board, shorting the 3 volt supply to ground.

Not good :-) So if you blow the FETs, repairs would require board-level digging, not something I would do for a $200 part. I may fiddle with it some more, but for the moment I will classify this as "toast". :-)

C

To be honest I contribute a lot of that to the tightly packed nature of the board. If anything blows it has a high chance of taking something else close-by with it.
glad I didn't try and fix my 3.3v shorted one.

[freddyfarnsworth]

On another side note I finished reviewing a blown board that was given to me which had a complete blow-out of the 1 volt FETs due to a short.

Micro-summary: It's destroyed. Not because of the FETs; I got those off easily and cleared the shorted 1 volt lines. It's the fact that the whole 3.3 volt line is shorted as well; I was trying to figure out where that went bad and found out there was a nice bubble in the board by the chips. Apparently the 1 volt line not only shorted but it burned inside the board, shorting the 3 volt supply to ground.

Not good :-) So if you blow the FETs, repairs would require board-level digging, not something I would do for a $200 part. I may fiddle with it some more, but for the moment I will classify this as "toast". :-)

C

Getting to the real cause of all the problems, board traces not heavy - thick enuf for the extra amps even if the chips can deal with it,
I noticed alot of MBmakers went to heavy coppertraced boards a few years ago, when OC became the norm.

I wanted to know if you have made any progress on finding replacement chips at 80-100a or so (what ever would still have some headroom for what is on it and adding 8 chips.
But hell, if the boards are to thin....

Maybe they can sell you fresh PCB's ? or evenmake better ones ? They would not be interested - they must be tired of fixin em by now tho .
Or make you and friends a official repair center
Lots to ponder, thank you

freddy

[Drug5bitz]

And in the really interesting news, I had a unit come in with a power supply that went *bad*. Jalapeno, ran it with 3 chips, checked in and saw that it had a massive amount of errors going on. Shut it down, put it on the bench supply, worked fine. Tested the power supply and I see 60 volts between negative and AC neutral.

Glad I have an isolated power supply on my miner. I'm wondering: Does BFL want the old power supply back when they RMA, or will they just send a new one out? I would really like to "interrogate" this one and see if it has the same blown caps as the one from England.

Good news: No damage to the jalapeno. In fact it's now hashing at 14gh with a second added (client provided) chip, much better than the 5.4 it came in with (it has two lousy chips on it and two very good ones).

C

I've heard both... apparently the right and left hands aren't talking... so maybe it depends on if you are wearing matching socks and Mars is in the shadow of Venus...

I honestly can't see them needing the PSU back, it's not like you could fix it and reship it.

^^^ Yes, Thank you.

[godsdog]

I'm wondering: Does BFL want the old power supply back when they RMA, or will they just send a new one out?

C

Lightfoot client here. sent Rma pending approval no mention of return psu... BFL>"Once your RMA is approved, no further action is necessary on your part, and a new PSU will be sent to the shipping address in your order. "
just want to Thank lightfoot and promote his work

[lightfoot]

Cool. In that case I will take apart the power supply with my bench vise. They're pretty easy to crack open, I'm curious to see if it will match the UK blown up one (which had dents in the caps from manufacture, *wow*)

C

[lightfoot]

I wanted to know if you have made any progress on finding replacement chips at 80-100a or so (what ever would still have some headroom for what is on it and adding 8 chips.
But hell, if the boards are to thin....

Maybe they can sell you fresh PCB's ? or evenmake better ones ? They would not be interested - they must be tired of fixin em by now tho .
Or make you and friends a official repair center
Lots to ponder, thank you

freddy

To be honest, Mr. Teal did all this and more with the Chili design. That is without a doubt an amazing board, the BFL singles really aren't built for this kind of stuff.

Which is the whole point why BFL was late: Their chips used more power than they were expecting, forcing a complete redesign. Remember that the initial jalapeno design was to not use the big on-board 1 volt supply, but to use a little supply that was left on the board. It was all supposed to be powered from the USB port or somesuch. That... didn't... work... :-) Likewise the board simply wasn't built to handle the extra 30 or so amps the chips needed for a full 30gh single. They could have made 24 without blowing up the moon but that was it, we all "paid" for 30, and we had to wait months for it.

I wonder how many people would have been happy with 24gh in march as opposed to 30gh in October. But they would have all bitched about being "ripped off", so in the end they came out worse and BFL did the "right" thing. Interesting how that happens...

Anyway, at this point the next system to look at will be the monarch class systems. I could open a repair center for jallies and singles, but would the cost of doing it (let's say $100 an hour) be worth it? Perhaps it would, thoughts?

C

[CeeCee]

if anyone is interested i have some chips in hand that i would like to sell. escrow is of course possible.

[lightfoot]

Interrogated the power supply in the danger shed. Using the danger vise of course.

Turns out the caps in the power supply are nicely domed. Looks like the one doing the PFC is blown to AC, which is why there's 60v ac on the negative line. Same exact damn failure as the UK one.

In addition I reballed another two chips, and installed one in my five chip jally. Came up, now a 6 chip jally, no problems. Still the slowest in the bunch, but a solid 22gh speed.

Next up? Need to wait for the water cooling block to come in, then I'll take it to 7 chips and maybe make it the 8 chip jally. I still have two chips left here, and a pair of others to ball. Need another jalapeno :-)

C

[lightfoot]

Some reballing thoughts:

I'm using a $5.00 screen set and a $2.00 super-cheap-o screen holder that holds the chip to the screen with a small spring. Not as good as the $60 kits which allow you to ball then pull the screen off before heating. This has the screen on all the time and the chip can stick/rip off a ball. But I'm cheap-ass.

What I do is use slightly undersized balls (.45mm) and a .5mm screen. I flux the chip, use a soldering iron at 800f to remove the balls by going across the chip with the CLEAN tip, making little solder balls. Using wick is only good to get remaining mountains off the chip, I find just stroking works well. Not too long and always keep some tacky flux on it.

Then it's time to clean the chip in alcohol, brush it shiny, then mount it in the tool. Flux the chip with some sticky flux, then set the screen so the pads are centered, then pour on some balls. I press them in with my finger in a stroking motion, then use a tweezers to place missed pads. Press down on the balls, they have to contact the chip.

Then it's heat time. First heat is at 375c at high blow speeds. This is to heat the chip and the balls. Then I put some liquid flux on and do it again. Then I let it cool completely, then brush it with alcohol a *LOT* to get the chip to release. Be gentle, you can rip balls off.

With the balls off check it out with a 6x jewelers' loupe. The most important balls are the ones at the edge; they have the signals. Especially the ones opposite the dot; those are critical. You can miss a few in the middle, it's all 1 volts anyway.

Then mount and test. The slightly undersized balls make it less likely to stick to the screen, and make it come off the screen more easily. Once again the $60 kit is going to be better, but I'm doing this on a budget.

Done three chips so far, all good. Not bad.

C