Bitcoin Forum
June 28, 2024, 01:42:32 PM *
News: Latest Bitcoin Core release: 27.0 [Torrent]
 
   Home   Help Search Login Register More  
Pages: [1] 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 »  All
  Print  
Author Topic: Hacking a BFL Jalapeno to 32GH and beyond....(???)  (Read 54272 times)
lightfoot (OP)
Legendary
*
Offline Offline

Activity: 3122
Merit: 2242


I fix broken miners. And make holes in teeth :-)


View Profile
November 17, 2013, 06:05:31 PM
Last edit: September 23, 2014, 07:02:48 PM by lightfoot
 #1

Summary: We are going to turn a Jalapeno into a Jupiter. Or bust. Probably bust. But maybe not!

Being one of those people who spent $169 for a Jalapeno in January and got his... money's worth, I've been thinking of ways to keep it relevant as difficulty has been tottering up.

I'm going to start this thread to track what I have been doing, what's worked so far, and where we can go with this. So far I have managed to boost my poor little Jally from 4.5gh barely (it had crappy chips) to 7.3gh, then 12, then 16, and for the moment, 20. I've since taken one to 24gh, and will try for 28gh next.

Update: Yep, 28gh, 7 chips. Beyond insane. Heat issues get very funny at 7 chips, thinking about that before trying for 8.

Update: Yep, 32gh. A real, honest, 32gh from a fracking Jalapeno. Read the thread, think about if you really want. to. go. there.

(I have a house there)

I will talk about hardware hacks I am doing, software hacks from this forum, and thoughts on just how big I can make this poor little bastard of a device.

First, a word of serious thanks to the following people:

  • Butterfly Labs: For releasing the source code to the little singles and also releasing the schematics.
  • CK: Who compiled an elf file that allowed my POS jally to go to 7.3gh which is as much as I could get out of it. He showed me the way.
  • Danattacker, who had the sheer balls to be the first to put 2 chips on his JP and showed that it was possible
  • lentbt2, good seller on Ebay who has MORE CHIPS as of 12/19. http://www.ebay.com/itm/191006340017 is the link, note I have bought from him in the past and he's good.

    If you have a source or know of one please PM me. I'll take the risk of ordering one and if they look good I'll post it here.
  • My uncle, who inspired me to fix watches which gives me the fine tools to pull shit like this off.
  • Whoever runs this BBS/Forum, it's been a great place to read, chat, and think.

The parent thread for all this is at https://bitcointalk.org/index.php?topic=236875.500, I posted a lot there but want to put the summary instructions and my future travails on it's own thread (and because I am kind of bogarting that thread because it's more software hacking and I am doing some serious hardware fuckery here)

Anyway, she's now running at 20gh, I'll post more later this afternoon, but first a few pictures of it running.

Here is the unit in it's current... configuration.


Here is the back of it with the extra lights on.


And here is the BFG screen.


On with the show...

C
minternj
Sr. Member
****
Offline Offline

Activity: 434
Merit: 250


View Profile
November 17, 2013, 07:01:41 PM
 #2

Good stuff - are you using a reflow station to do this?

Warning about Nitrogensports.eu
https://bitcointalk.org/index.php?topic=709114.0
volosator
Sr. Member
****
Offline Offline

Activity: 272
Merit: 250



View Profile
November 17, 2013, 07:56:16 PM
 #3

Just added chips and flashed the FW?
Or something else was replaced on the board too?
lightfoot (OP)
Legendary
*
Offline Offline

Activity: 3122
Merit: 2242


I fix broken miners. And make holes in teeth :-)


View Profile
November 17, 2013, 08:12:17 PM
 #4

Good stuff - are you using a reflow station to do this?
Not quite. I'm using an Aoyue 968 rework station that I used to use for SMD rework on electric car drive trains. 50kw IGBT drivers blow out a bunch of parts when the gates float. I also have an Aoyue pre-heater that's nice to have, but I don't think it's quite mandatory for this work.

Anyway, that plus some liquid flux plus some time under the air flow seems to do the trick. I'll post the details in a write-up.

C
lightfoot (OP)
Legendary
*
Offline Offline

Activity: 3122
Merit: 2242


I fix broken miners. And make holes in teeth :-)


View Profile
November 17, 2013, 08:22:15 PM
 #5

Just added chips and flashed the FW?
Or something else was replaced on the board too?
Not yet. So far it is add chips and reflash the firmware. I'm running speed 7 with full error detection which is probably running me slower than I could by a gh or two, but with the benefit of lower power draws and temps.

You need to have the 1.2.7 or better software because the original chips were SMD chips on BGA carriers and couldn't run with engine zero. The newer chips are real BGA, but have engine 0 enabled by default (16 engines instead of max 15. cool). So grabbing the 1.2.9 software, changing the type from SINGLE to LITTLE_SINGLE (important) does the trick.

I think the limiting factor for going past 6 chips on the little single board is going to be the power converters for the 1 volt supplies. They are built to source "75 to 80 amps" and I'm guessing the original chips ran at less than 10 amps/watts each. However since each chip is boosting the power draw at the wall by 20 watts or so, I think they pull closer to 15 watts each. 15*5 is 75 amps, so I might be pushing the limit with 5 chips. But I'll try 6 for the hell of it, if it fails I'll disable or remove one of the crappy original chips.

Here is the diag output for my Jally as of now. The 1 volt rail is at .998; input voltage is 12.22 instead of 13.2 with 4, so I might be running up against the limits of the power supply or the 1 volt FETs.

DEVICE: BitFORCE SC
FIRMWARE: 1.2.9
IAR Executed: NO
CHIP PARALLELIZATION: YES @ 5
QUEUE DEPTH:40
PROCESSOR 0: 16 engines @ 267 MHz -- MAP: FFFF
PROCESSOR 3: 13 engines @ 262 MHz -- MAP: EFFC
PROCESSOR 5: 16 engines @ 245 MHz -- MAP: FFFF
PROCESSOR 6: 16 engines @ 259 MHz -- MAP: FFFF
PROCESSOR 7: 15 engines @ 240 MHz -- MAP: FFFE
THEORETICAL MAX: 19342 MH/s
ENGINES: 76
FREQUENCY: 274 MHz
CRITICAL TEMPERATURE: 0
TOTAL THERMAL CYCLES: 0
XLINK MODE: MASTER
XLINK PRESENT: NO
OK

I have an ATX power supply coming to provide a more solid 12 volt supply, might give it a go before committing to that next chip. Or I might just toss it on tomorrow.

C
jelin1984
Legendary
*
Offline Offline

Activity: 2408
Merit: 1004



View Profile
November 17, 2013, 08:30:44 PM
 #6

Do you install extra chips at jalapeño Board?
Or only firmware updated?
lightfoot (OP)
Legendary
*
Offline Offline

Activity: 3122
Merit: 2242


I fix broken miners. And make holes in teeth :-)


View Profile
November 17, 2013, 08:40:43 PM
 #7

Do you install extra chips at jalapeño Board?
Or only firmware updated?
Update the firmware and you can take the chips from 5 to 7.5 or 8 or so. That's pretty well documented on the forum, searches will find that. Costs $50 for a Dragon programmer and maybe $4 for a cable. And some time. And Atmel software.

You can add one chip and get to 12 and still be able to run it in it's case (barely). Two more will get you to 16 and is about as fast as you can go without adding bottom cooling. Cool the bottom of the unit and we're in different territory.

C
jelin1984
Legendary
*
Offline Offline

Activity: 2408
Merit: 1004



View Profile
November 17, 2013, 08:43:01 PM
 #8

Is it easy to install the extra chip at jalapeño board?
danattacker
Full Member
***
Offline Offline

Activity: 121
Merit: 100


View Profile
November 17, 2013, 09:06:13 PM
 #9

Nice work. I like the heatsink you put on the bottom. I might try that but I'm a bit hesitant to do any more work on my Jalapeno because it doesn't really seem to be worth it at this point. But, it might try it whenever I switch over to solo mining.

Also, with that one chip you removed, if you wanted to you could solder it back on without reballing. I had to do it myself since I messed up on the first chip I soldered. Just remove all the solder from the chip and use the solder balls on the PCB. The problem is that the chip will sit very low and might not be able to be effectively cooled since it isn't coplanar to the other chips. I had to put two layers of thermal padding to make contact with the heat sink.

  • Danattacker, who had the sheer balls to be the first to put 2 chips on his JP and showed that it was possible
Actually, I don't think I was the first. I recall somebody who was developing a BFL based board doing it before I did using sample BFL chips and also offering it as a service. But, I think I was one of the first to start reflashing the firmware on the Jalapeno. And, I was the one who went and told everyone to buy the AVR Dragon.
lightfoot (OP)
Legendary
*
Offline Offline

Activity: 3122
Merit: 2242


I fix broken miners. And make holes in teeth :-)


View Profile
November 17, 2013, 11:28:24 PM
 #10

Nice work. I like the heatsink you put on the bottom. I might try that but I'm a bit hesitant to do any more work on my Jalapeno because it doesn't really seem to be worth it at this point. But, it might try it whenever I switch over to solo mining.
Thanks! Yeah, the heat sink on the bottom removes a lot of the heat; I think that's the point of BGA: because you have a whole bunch of pads you can drop the heat out the bottom as well as the top.

I'll get some solder wick and start removing the balls this week; I was thinking of just reforming the balls and not worrying about the missing ones, but it looks like some of the outside edge ones are gone and that is where the control signals are. Not too worried about the lack of heat sink; oddly enough putting it closer to the board will just source more heat out the back.

From a purely money-making point of view it's not very worth it; I'll make a profit of about $50 a chip over the next few months if BC prices stay constant. About as much as a pair of watch cleanings. But from a fun standpoint, it can't be beat :-)

Thanks for all the help in the other thread. Jallies may have been the only devices from BFL to actually turn a profit on mining, most other devices did not do as well.

Oh and the heat sink at the bottom? It was reading 90 degrees F with the unit at 65C. To warm up a chunk of copper to 30 degrees above room temperature takes a fair bit of heat. Once I drill the holes in it and get longer screws I'll put the whole shooting match on it's side or something to boost the cooling rate.

C
ericdc30
Newbie
*
Offline Offline

Activity: 46
Merit: 0


View Profile
November 18, 2013, 04:52:22 AM
Last edit: December 17, 2013, 03:07:32 PM by ericdc30
 #11

I have 3 bfl chips for sale if anyone is interested pm me.

Sorry none left. Will be selling some used mining gear when my neptune comes in.
lightfoot (OP)
Legendary
*
Offline Offline

Activity: 3122
Merit: 2242


I fix broken miners. And make holes in teeth :-)


View Profile
November 19, 2013, 02:57:48 AM
 #12

Power supply came in, I'll fiddle with it later. Doing some calculations based on the Chili boards I think the BFL chips pull 12.5 watts each, so I should be able to go to six and 75 watts on the DC-DC power supplies. Worst case they should just crowbar and shut down.

I've ordered longer screws so I can screw the bottom heat sink to the top plate, and a set of small stick-on sinks for the hot FETs. If I get them by Wednesday I'll put the last chip on.

C
Bogart
Legendary
*
Offline Offline

Activity: 966
Merit: 1000


View Profile
November 19, 2013, 04:17:15 AM
 #13

That's pretty cool.  I have 2 jalapenos and 7 loose BFL chips on hand, but no hot air rework equipment.  I'm getting ideas involving my kitchen oven...

"All safe deposit boxes in banks or financial institutions have been sealed... and may only be opened in the presence of an agent of the I.R.S." - President F.D. Roosevelt, 1933
Inaba
Legendary
*
Offline Offline

Activity: 1260
Merit: 1000



View Profile WWW
November 19, 2013, 04:35:56 AM
 #14

We have test units in the shop that have 8 chips that run fine with a heatpipe heatsink.  Out of the 8 or so we have for test, to my knowledge only one has failed since about June... you should be able to run an 8 chip Jalapeno just fine with adequate cooling, at least for a few months.  Of course, it voids your warranty... however, as a side note, if it does fail, the chips will likely be ok and reballing those chips and putting them on another board would probably be just fine.

If you're searching these lines for a point, you've probably missed it.  There was never anything there in the first place.
lightfoot (OP)
Legendary
*
Offline Offline

Activity: 3122
Merit: 2242


I fix broken miners. And make holes in teeth :-)


View Profile
November 19, 2013, 12:27:09 PM
 #15

We have test units in the shop that have 8 chips that run fine with a heatpipe heatsink.  Out of the 8 or so we have for test, to my knowledge only one has failed since about June... you should be able to run an 8 chip Jalapeno just fine with adequate cooling, at least for a few months.  Of course, it voids your warranty... however, as a side note, if it does fail, the chips will likely be ok and reballing those chips and putting them on another board would probably be just fine.

Interesting. I wasn't sure if the 1 volt power supplies would hold past 6 chips. Looks like I'll be ordering two more from somewhere, then going the distance.

Could you share what type of heat sink with pipes you used? I've been looking around at sites like frozencpu.com, and there seem to be as many sinks as stars in the sky. Any that would allow this thing to run in it's original case?

Thank you!
Chris
daddyfatsax
Hero Member
*****
Offline Offline

Activity: 857
Merit: 1000


Anger is a gift.


View Profile
November 19, 2013, 01:15:07 PM
 #16

My question is how in the hell did you solder the chips on the board? I can do small soldering, but how hard is it really?
Inaba
Legendary
*
Offline Offline

Activity: 1260
Merit: 1000



View Profile WWW
November 19, 2013, 03:33:43 PM
 #17

The same heatsink that comes on 60 GH/s long boards are the ones I'm referring to.

If you're searching these lines for a point, you've probably missed it.  There was never anything there in the first place.
lightfoot (OP)
Legendary
*
Offline Offline

Activity: 3122
Merit: 2242


I fix broken miners. And make holes in teeth :-)


View Profile
November 19, 2013, 04:19:31 PM
 #18

My question is how in the hell did you solder the chips on the board? I can do small soldering, but how hard is it really?

Well....

What I used was an aoyue 968 rework station. It's basically a pretty nice soldering iron and an air wand that can put out air at a specific temperature and flow rate. Costs about $160 on Amazon. I bought it a few years ago to fix SMD components on 90's era electric car controllers, think 50kw IGBT drivers that explode and need to be rebuilt.

http://www.amazon.com/Updated-Aoyue-Digital-Soldering-absorber/dp/B006FA481G/ref=sr_1_fkmr1_2?ie=UTF8&qid=1384876991&sr=8-2-fkmr1&keywords=aoyue+962

Cripes, you could probably use this:

http://www.amazon.com/REWORK-SOLDERING-IRON-STATION-handles/dp/B004ZB9D4O/ref=sr_1_24?ie=UTF8&qid=1384877176&sr=8-24&keywords=aoyue

In addition I have an Aoyue pre-heater thing which basically puts out a fairly constant amount of heat and holds the board.

http://www.amazon.com/Aoyue-853A-Infrared-Preheating-Station/dp/B000PGPU7W/ref=sr_1_1?ie=UTF8&qid=1384877226&sr=8-1&keywords=aoyue+preheater

For $20 more you can get a probe. Wow! Modern technology.

To be honest, it's not too hard. You need to put liquid flux on the board (use the no-cleanup flux, try to get some tacky stuff) then put the board on the preheater and bolt it down. Now spend a *LOT* of time with an eye loupe (4x) fingers, and a set of tweezers to put the chip on the board so that:
  • The chip dot matches the dot slik-screened on the board. That's um. critical
  • The chip is perfectly centered within the silk-screened square. I mean perfect. As in you can just barely see all four outlines around it.

Then you turn on the heat and check the chip again. I brought my pre-heater to 300F because Aoyue assumed I was stupid and didn't use celsius on the preheater. Oh well, check to see if yours is C or F, there is a difference. :-)

Then you get your stopwatch. And time things. After 4 minutes or so fire up your 962. Bring it's temp to flow level 5 and temp of 402C. Yes, that is hot as hell. Yes, it bothers me too. Yes, that's life. 250-300C doesn't do crap.

Check the chip again. It probably moved. If not good. Note the preheat has activated the flux.

Now hit the chip with the heat. I use the round nozzle that is about 80% the diameter of the chip. Whatever. The square one sucks. Put the nozzle over the chip and bring it down over 10 seconds to the point where it is right over the chip. Start the watch.

After 30 seconds I start moving around the edges of the chip with the nozzle. Blow straight down, if the chip blows off you're fucked-ish.

Watch the chip. It should do a little move when the solder melts and it aligns itself. Keep the heat on it.

Do heat for 60-90 seconds. If the board catches fire you did it too long. I felt horrified I needed that much heat but I haven't blown a chip yet. Yet.

Remove the heat, then turn off the bottom preheater. Wait a minute or two for the solder to cool. Then take the board off the preheater, burn your fingers, put on gloves, and walk around with it to cool it down.

Look at the chip. If it's not square within the 4 lines on the board you really blew it. It probably is.

When room temp, plug it into the Jally that you have already flashed with 1.2.9 with the LITTLE_SINGLE definition and made sure it WORKED and see what happens.

If 3 LEDs pop on the back along with the power led, power down immediately. You win.

If not, reheat and try again. Go a bit longer with the heat.

If the front LED flashes, power down, you shorted the 1 volt power supply. Thank BFL for the crowbar detection and come back to this thread. Or re-heat the board again, it might not have soldered correctly.

Anyway if it lit 3 LEDs then put on the heat sink, fire it up and check out the performance. Should be 12gh. Then keep adding chips until something melts.

C
lightfoot (OP)
Legendary
*
Offline Offline

Activity: 3122
Merit: 2242


I fix broken miners. And make holes in teeth :-)


View Profile
November 19, 2013, 04:23:33 PM
 #19

The same heatsink that comes on 60 GH/s long boards are the ones I'm referring to.

Ah, ok, haven't seen one of those personally. Is there a part number or something I can order that sink from somewhere? Or something similar aftermarket?

I will say that checking it this morning that the copper heat sink I have on the bottom is contributing a *lot* to keeping the board cool-ish. Heat can go either way, and a good sink on the bottom seems to work well at 5.

As for the warrantly, yeah this is way beyond "I just opened it to peek". If it blows up it blows up. I (and anyone else trying this) can deal with that. :-)

Inaba
Legendary
*
Offline Offline

Activity: 1260
Merit: 1000



View Profile WWW
November 19, 2013, 06:07:31 PM
 #20

The same heatsink that comes on 60 GH/s long boards are the ones I'm referring to.

Ah, ok, haven't seen one of those personally. Is there a part number or something I can order that sink from somewhere? Or something similar aftermarket?

I will say that checking it this morning that the copper heat sink I have on the bottom is contributing a *lot* to keeping the board cool-ish. Heat can go either way, and a good sink on the bottom seems to work well at 5.

As for the warrantly, yeah this is way beyond "I just opened it to peek". If it blows up it blows up. I (and anyone else trying this) can deal with that. :-)



Yeah, we have a new heatsink that goes on the bottom of the units that are shipping with newer units that replaces the flat reinforcement backplate.

Email me one of your order numbers and I will see if I can get a heatsink sent out to you so you can give it a test.

If you're searching these lines for a point, you've probably missed it.  There was never anything there in the first place.
Pages: [1] 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 »  All
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2006-2009, Simple Machines Valid XHTML 1.0! Valid CSS!