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

[ManeBjorn]

I use photobucket.
I don't now though if it will do what you need or not though but it may be worth a look.

Ok, I'm building a thread for the single. With a speed of 8 I have it running at a solid 28gh, and with a resistor in line with the big fan I have the noise level down to quiet with a temp of 65 on the board. Not too bad. The trick is to keep the fan over the chips running at full speed; the end fans just need to get the air out of the box and around the heat sinks.

https://bitcointalk.org/index.php?topic=424604.msg4624265#msg4624265

On a side note it looks like Flikr is being an ass and doesn't want to give me the raw embed URLs for my photos. Anyone got a better site to share photos?

C

[Drug5bitz]

Love imgur.com 

It is my love for photos and memes. Great community or just great photo hosting.  Share your stuff to the world or keep it relatively close to yourself.

[lightfoot]

Sounds good. I just posted the pics on the other thread. I think this might be possible to do, at least to 4 more chips. This is going to be fun; I'm going to screw up a perfectly working single.

Anyone from BFL still read this; if so are there any schematics for the bigger single boards around?

C

[lightfoot]

Lightfoot, I just got 4 jalapenos and I have gotten the hackit-to-go-faster bug.  Already changed out the 7k resistor with a larger one to get more voltage to the

ASICs.  This reduced the errors by 1% and increased the speed by 500Mhs.  Thinking about putting maybe two more chips in each unit.  Do you know where the early type

Speaking of which, what resistor are you swapping out to boost the power, and do you have a digikey/mouser part number? My 8 chip jally is actually humming at 32gh/45c due to the water block and now I'm thinking..... Just a bit.

C

[MrTeal]

Lightfoot, I just got 4 jalapenos and I have gotten the hackit-to-go-faster bug.  Already changed out the 7k resistor with a larger one to get more voltage to the

ASICs.  This reduced the errors by 1% and increased the speed by 500Mhs.  Thinking about putting maybe two more chips in each unit.  Do you know where the early type

Speaking of which, what resistor are you swapping out to boost the power, and do you have a digikey/mouser part number? My 8 chip jally is actually humming at 32gh/45c due to the water block and now I'm thinking..... Just a bit.

C

I don't know if there's been a new update that's changed the designators, but on the old boards R11 and R12 controlled to voltage. The formula is Rtop(R11, 7k) = Rbottom (R12, 10k) * [(Vout - 0.6V) / 0.6V], or Vout = (Rtop/Rbottom+1)*0.6V
Currently it's set to 1.02V, so if if you want to change it to 1.1V change R11 to 10k*[(1.1-0.6)/0.6] = 8.33k. Not a standard value, but 8.2k would work well enough.

[lightfoot]

Lightfoot, I just got 4 jalapenos and I have gotten the hackit-to-go-faster bug.  Already changed out the 7k resistor with a larger one to get more voltage to the

ASICs.  This reduced the errors by 1% and increased the speed by 500Mhs.  Thinking about putting maybe two more chips in each unit.  Do you know where the early type

Speaking of which, what resistor are you swapping out to boost the power, and do you have a digikey/mouser part number? My 8 chip jally is actually humming at 32gh/45c due to the water block and now I'm thinking..... Just a bit.

C

I don't know if there's been a new update that's changed the designators, but on the old boards R11 and R12 controlled to voltage. The formula is Rtop(R11, 7k) = Rbottom (R12, 10k) * [(Vout - 0.6V) / 0.6V], or Vout = (Rtop/Rbottom+1)*0.6V
Currently it's set to 1.02V, so if if you want to change it to 1.1V change R11 to 10k*[(1.1-0.6)/0.6] = 8.33k. Not a standard value, but 8.2k would work well enough.

Hm. I may try this on my 5 chip jally (the one I just can't seem to add chips to). What types of resistors are those little things (soic or something); I feel weird wiring a 1/4 watt axial resistor in there. :-)

C

[MrTeal]

Lightfoot, I just got 4 jalapenos and I have gotten the hackit-to-go-faster bug.  Already changed out the 7k resistor with a larger one to get more voltage to the

ASICs.  This reduced the errors by 1% and increased the speed by 500Mhs.  Thinking about putting maybe two more chips in each unit.  Do you know where the early type

Speaking of which, what resistor are you swapping out to boost the power, and do you have a digikey/mouser part number? My 8 chip jally is actually humming at 32gh/45c due to the water block and now I'm thinking..... Just a bit.

C

I don't know if there's been a new update that's changed the designators, but on the old boards R11 and R12 controlled to voltage. The formula is Rtop(R11, 7k) = Rbottom (R12, 10k) * [(Vout - 0.6V) / 0.6V], or Vout = (Rtop/Rbottom+1)*0.6V
Currently it's set to 1.02V, so if if you want to change it to 1.1V change R11 to 10k*[(1.1-0.6)/0.6] = 8.33k. Not a standard value, but 8.2k would work well enough.

Hm. I may try this on my 5 chip jally (the one I just can't seem to add chips to). What types of resistors are those little things (soic or something); I feel weird wiring a 1/4 watt axial resistor in there. :-)

C

They're just 0603 resistors according to the Altium files.
http://www.digikey.com/product-detail/en/RC0603JR-078K2L/311-8.2KGRCT-ND/729777
http://www.digikey.com/product-detail/en/RC0603FR-078K45L/311-8.45KHRCT-ND/730339

or just get a kit

http://www.ebay.com/itm/New-SMD-SMT-0603-1-144-values-resistor-kit-144-X-100pc-14400-filled-BOX-ALL-/170869001211?pt=LH_DefaultDomain_0&hash=item27c8960bfb

[helipotte]

Lightfoot, I just got 4 jalapenos and I have gotten the hackit-to-go-faster bug.  Already changed out the 7k resistor with a larger one to get more voltage to the

ASICs.  This reduced the errors by 1% and increased the speed by 500Mhs.  Thinking about putting maybe two more chips in each unit.  Do you know where the early type

Speaking of which, what resistor are you swapping out to boost the power, and do you have a digikey/mouser part number? My 8 chip jally is actually humming at 32gh/45c due to the water block and now I'm thinking..... Just a bit.

C

I don't know if there's been a new update that's changed the designators, but on the old boards R11 and R12 controlled to voltage. The formula is Rtop(R11, 7k) = Rbottom (R12, 10k) * [(Vout - 0.6V) / 0.6V], or Vout = (Rtop/Rbottom+1)*0.6V
Currently it's set to 1.02V, so if if you want to change it to 1.1V change R11 to 10k*[(1.1-0.6)/0.6] = 8.33k. Not a standard value, but 8.2k would work well enough.

That's it.  8.2K is what I have been using.  I get at least another 1Ghs out of a jalapeno.

Unrelated, but I have a Single with a chip that shows all zeros in bfgminer.  Chip 0d. Been that way since I got it.  is there a way to turn it on?  Is it just bad?  Can the chips be disabled in hardware?  This unit already hashes at 69Ghs but having that chip online would be nice.  Any ideas?

I have actually "looked" inside and counted 16 chips so I'm sure one is just turned off/dead.  Thanks for any insight.

[lightfoot]

Wrecked another old-style chip in testing; even with a completely perfect reballing and placement, the board was shorted out, and the problem was the chip. I could see a tiny amount of solder around the top of the chip. So I broke the chip open.

These chips are way, way different from what I thought. I was expecting a simple interface, but there are tiny little wires in there going to a chip assembly that is even *smaller* than the BGA. It's like a pico-BGA or something.

I have no idea anymore what BFL was thinking. But a warning: It is not possible to drop the chips into place by using heat from the top; I will have to explore using heat from the bottom if even possible.

Very interesting.

C

[MrTeal]

Wrecked another old-style chip in testing; even with a completely perfect reballing and placement, the board was shorted out, and the problem was the chip. I could see a tiny amount of solder around the top of the chip. So I broke the chip open.

These chips are way, way different from what I thought. I was expecting a simple interface, but there are tiny little wires in there going to a chip assembly that is even *smaller* than the BGA. It's like a pico-BGA or something.

I have no idea anymore what BFL was thinking. But a warning: It is not possible to drop the chips into place by using heat from the top; I will have to explore using heat from the bottom if even possible.

Very interesting.

C

I'm not sure what you mean there. The shiny part is the actual silicon die, and it has little solder bumps that attach it to the substrate which is just a tiny PCB. That's the bottom of the BGA. On another note, the new packages and substrates that BFL uses are amazing. The old style (below) is a pretty typical BGA substrate


The new package is altogether different, and looks incredible. I'd really love to see actual test results, showing the performance differences between them.

[lightfoot]

Sort of. Here is one of the chips with the top... Removed.



And here is a closeup with a 6x loupe in front of the camera:



Note that the curled up thing on the top is the chip itself, with little tiny solder dots on it. Also note the mask that connects it to the BGA and the smaller than hair *wires* that go from the chip to the BGA board.

What's happening here is that heat from the top does not flow through the chip carrier; it gets stuck in the top chip. And heats the solder balls under it to the point where they expand and explode out the sides of the chip (seriously I had a little ball of solder on top of the chip). I can't believe that most BGA chips have lots of little wires in there; maybe they do.

The new model seems to be the chip right on top of the BGA bottom. Heat flows through, there doesn't seem to be solder inside, and they never give me a problem.

I think I need to do this with a home-made reflow oven and not air heat. I can only get the bottom up to 375 or so; one option would be to direct the 400c air heat to the bottom of the board under the chip and try to bring that region to 400+. Or something. But man is it a problem to work with these chips.

C

[MrTeal]

Sort of. Here is one of the chips with the top... Removed.


Note that the curled up thing on the top is the chip itself, with little tiny solder dots on it. Also note the mask that connects it to the BGA and the smaller than hair *wires* that go from the chip to the BGA board.

What's happening here is that heat from the top does not flow through the chip carrier; it gets stuck in the top chip. And heats the solder balls under it to the point where they expand and explode out the sides of the chip (seriously I had a little ball of solder on top of the chip). I can't believe that most BGA chips have lots of little wires in there; maybe they do.

The new model seems to be the chip right on top of the BGA bottom. Heat flows through, there doesn't seem to be solder inside, and they never give me a problem.

I think I need to do this with a home-made reflow oven and not air heat. I can only get the bottom up to 375 or so; one option would be to direct the 400c air heat to the bottom of the board under the chip and try to bring that region to 400+. Or something. But man is it a problem to work with these chips.

C

I'm reasonably certain those wires are the top traces of the substrate, and they just lifted off when you overheated it.

Edit:

[lightfoot]

Possibly, however I have never had anything like this happen with the new chips (I have literally baked them at 450c and they still work); these can't take a few seconds before going "pop" and spitting solder out the side. Annoying.

Need to think about another method; I think I'm going to try just preheat for 10 minutes (I'm using lead solder on the reballs) and if that doesn't stick go to the microwave oven. Ok, the toaster oven. :-)

C

[lightfoot]

On the other hand I have figured out how to mount them to the board after reballing:

1) Set the heat under the unit to as high as possible (370f)
2) Walk away for awhile
3) Turn off heat.

The bottom heat is enough to melt lead solder, slowly. However I would have to just touch the chip while it was cooling and squish a bit of solder out, which of course means it's shorted again and needs to be reballed.

I. Hate. This. Junk. Board.

(This is the "danger board" that I am trying to get working. Was a completely baked nightmare, I have been trying out junker chips with mixed success)

[Brassguy]

Just looking for opinions...

I have 3 6-chip jallys

cgminer stats...
#1@ 23.6gh @38-42C 1.3%HW
#2@ 23gh @42-47C 2%HW
#3@ 22gh @45-50C 3.5%HW

My hottest mosfets are at ~105F. They have heatsinks on top and bottom. Bottom of boards are actively cooled with cpu heatsinks (and fans).

I have 8.2k 0603 resistors and 2 new style chips that need to be re-balled and may or may not work...

Any opinions on what I can expect with installing the resistors?

Fire, smoke, damnation for all eternity?

[Drug5bitz]

I agree I would like to see some more results on this resistor. Post up if you get any good data on those. All air cooled I think that's my final step before I start selling a miner or two. These are slowly becoming ever less profitable.

On that thought Alpha-t.net the first scrypt asic soon to be on the market. Batch 1 preorders are closed, but I'm seeing them being a lot better than BFL ever was.

[lightfoot]

Just looking for opinions...

I have 3 6-chip jallys

cgminer stats...
#1@ 23.6gh @38-42C 1.3%HW
#2@ 23gh @42-47C 2%HW
#3@ 22gh @45-50C 3.5%HW

My hottest mosfets are at ~105F. They have heatsinks on top and bottom. Bottom of boards are actively cooled with cpu heatsinks (and fans).

I have 8.2k 0603 resistors and 2 new style chips that need to be re-balled and may or may not work...

Any opinions on what I can expect with installing the resistors?

Fire, smoke, damnation for all eternity?

You have so much cooling I would worry about a snowstorm. Go for it, see how it works.

C

[lightfoot]

Wow. Something is different. Just got my BGE bill, I pulled 873kwhr in December on my top floor instead of the normal 200kwhr.

Ooops. Guess these rigs are pulling more power than I thought. Might explain why I haven't had to run heat up there. But it does mean I need to start selling some bitcoin to pay the power bill.

Welcome to the big time.

[Brassguy]

Just looking for opinions...

I have 3 6-chip jallys

cgminer stats...
#1@ 23.6gh @38-42C 1.3%HW
#2@ 23gh @42-47C 2%HW
#3@ 22gh @45-50C 3.5%HW

My hottest mosfets are at ~105F. They have heatsinks on top and bottom. Bottom of boards are actively cooled with cpu heatsinks (and fans).

I have 8.2k 0603 resistors and 2 new style chips that need to be re-balled and may or may not work...

Any opinions on what I can expect with installing the resistors?

Fire, smoke, damnation for all eternity?

Preliminary results are in... I decided to mod #3 (my worst one) for testing...installed 8.2k resistor in R11... old voltage (reported by cgminer) .98v-1v... new voltage 1.05v-1.07v

new stats (after about an hour)

#3@ 24.6 @55-61C 3.6%HW

So there was a pretty big improvement to the hashrate... I didn't have any appreciable change in HW error rate.

BE WARNED My mosfets went from about 105F to about 130F... they are now being actively cooled down to around 110-115F...

Hope this helps

[lightfoot]

I'll try this on my five chip unit, need to place a small mouser order anyway. Maybe I'll give it a go on the 7 chip water cooled unit, we'll see.

C