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

[Drug5bitz]

Here you see the hole I blew in the voltage reg..... and the FTDI chip which controls the USB FT232HQ about $5.

In this pic you can see 6 mosfets on this side.


Careful guys apparently Bitcoin mining is as dangerous as gold mining.

[lightfoot]

Well, in the meantime I wired my Jally into my rather inexpensive (ie: cheap) ATX power supply.

I didn't want to play with the 12 volt 4 pin molex, so I took three of the yellow 12v wires (18 gauge each, blah!), cut them, spliced them together to the BFL cable (16 gauge, nicer, but blah), soldered, used shrink wrap to cover, and fired it up.

The power supply is reading 12 volts on the unloaded lines, but the Jally was reading 11.5 volts at idle. Which means a pretty hefty voltage drop, but most of it is in the BFL cable. I'll deal.

Up and hashing. I'll take a look inside my power supply in a few days to see what it can tell me. Man you do need a pretty beefy cable BTW.

Update: Ooops. Hit 80c, apparently lower voltage=lower noctura fan speed. Crud, I have the external fan back on it, we'll see what happens. Might have to take the sides off again, drat.

C

[lightfoot]

Wait, that chip you blew up is a voltage regulator? I thought the six FETs were it, or is that some sort of LM317 for the fans?

And is that big box thing damaged as well?

C

[lightfoot]

Oh and um, that cap damage took more than 72 watts to do. I think it shorted the mains to the output. Check the other side, see if any of the caps there are blown.

I just opened my supply, normal and somewhat stupid, nothing unusual there. I'll tear the one I have coming apart; are you in the US BTW?

C

[Drug5bitz]

Wait, that chip you blew up is a voltage regulator? I thought the six FETs were it, or is that some sort of LM317 for the fans?

And is that big box thing damaged as well?

C

Yes, notice the hole in the chip. The number on it was ST1S10, on mouser it comes up as a voltage reg. This particular jala is the one that took out the same power supply. When I used a good power supply it blew that chip.... doesn't necessarily mean that's the problem though....

The blown caps are on the blocks DC output side, everything else looks fine.  I'm not sure what you mean by big box thing? If you mean the USB controller chip, I ordered a couple. I have 2 jala's that wont register in windows... trust me I researched it on two PC's and 3 different miner programs. Again everything is just to the best of my knowledge, I'm out on a string with some of this stuff.

The 2 mouser chips I got ordered are

ST1S10 (Voltage reg)
FT232HQ (FTDI driver chip for USB)

Yes I'm in the USA unfortunately.

[lightfoot]

Wait, that chip you blew up is a voltage regulator? I thought the six FETs were it, or is that some sort of LM317 for the fans?

And is that big box thing damaged as well?

C

Yes, notice the hole in the chip. The number on it was ST1S10, on mouser it comes up as a voltage reg. This particular jala is the one that took out the same power supply. When I used a good power supply it blew that chip.... doesn't necessarily mean that's the problem though....

The blown caps are on the blocks DC output side, everything else looks fine.  I'm not sure what you mean by big box thing? If you mean the USB controller chip, I ordered a couple. I have 2 jala's that wont register in windows... trust me I researched it on two PC's and 3 different miner programs. Again everything is just to the best of my knowledge, I'm out on a string with some of this stuff.

The 2 mouser chips I got ordered are

ST1S10 (Voltage reg)
FT232HQ (FTDI driver chip for USB)

Yes I'm in the USA unfortunately.

What the hell is that? U15 is the power supply from 12 volts to 3.3 volts. That's the control bus power for the chips (the hash engines run on the 1 volt supply). Wow.

You had a cool failure. Swap the parts out, that chip first. And use a fuse inline with your power supply to keep blowing things to a min. Maybe a 4 amp fuse or something.

C

[lightfoot]

Here you go lightfoot...

https://dl.dropboxusercontent.com/u/37765158/image_01.png

I'm pretty sure this is exact... please let me know if you find any discrepancy!

I just ordered these from newegg http://www.newegg.com/Product/Product.aspx?Item=N82E16835708009 for cooling the vrms...

That does not match my layout at all. Weird.

C

[glendall]

Wow good work on the hacking guys.  a 18+GH Jalapeno is a sweet overclock. Nice work...

[lightfoot]

It's a start. If I get my new jallies I will try for a bit more. My current one seems to be confused.

C

[Brassguy]

It's a start. If I get my new jallies I will try for a bit more. My current one seems to be confused.

C

Just an fyi light.... when I had the one chip not working on my jally the reported hashrate was around 11gh after reflow about 15gh... simple enough. However, the temperature hasn't changed one bit with an addition of 4gh.... and... when it was hashing at reported rate of 11gh it was actually the same amount of shares ~100 as before any chips were added at 8.3gh. I ran it for 6 days that way btw... weird.

[lightfoot]

It's a start. If I get my new jallies I will try for a bit more. My current one seems to be confused.

C

Just an fyi light.... when I had the one chip not working on my jally the reported hashrate was around 11gh after reflow about 15gh... simple enough. However, the temperature hasn't changed one bit with an addition of 4gh.... and... when it was hashing at reported rate of 11gh it was actually the same amount of shares ~100 as before any chips were added at 8.3gh. I ran it for 6 days that way btw... weird.

Well, the rubber meet road question is how many bitcoin are you making? I am apparently doing about a penny a day now with the latest difficulty with 20gh, so as long as you're making the money it's all good.

C

[Brassguy]

https://db.tt/WuN8CdkW

Just some temp readings from various points...

[MagnaGen]

https://db.tt/WuN8CdkW

Just some temp readings from various points...

Someone should make a wiki with all of this information, I can host it and set it up if someone else is willing to gather all the information and instructions.

Also lightfoot, PM me I would like to donate some chips to your project

thanks

[lightfoot]

Hm, that's an interesting thought. I keep thinking about setting up a Wiki system on my home servers but just never get around to it.

I'd be happy to transfer the key points from this thread over so we can get this info out.

Thanks!
C

[elaramus]

So I got a little crazy this weekend and took both of my Jalapenos from 3 chips each to 6 chips each.



The results are below
http://imgur.com/a/FzKPM

In the linked images, BFL3 and BFL4 are my Jalapenos... the others are Chilis

The stock 6A adapter gave out after the 4th chip, so I cut the plug off and wired it to a PCI-E cable right off my computer's power supply. No issues with power after the mod. Adding 4 chips had both Jalapenos up to over 60C. The 5th chip needed additional cooling (up to 78C) and the 6th chip put the temps off the map.

I found applying thermal compound between the aluminum heatsink plate and the Jalapeno board then placing an active cooler on top (running the Jalapeno upside down) provided exceptional results... Both are hashing over 23Gh/s

Now I'm contemplating chip 7.... 

P.S Temps used: preheat the area with the wand for 10 seconds, then 90 seconds @ 460C with the wand 1/4 inch above the chip. The chip will float as the solder balls melt (DON"T FREAK OUT) as the remaining balls melt and adhere to one another the chip will realign itself. Patience was the key to my success. Sometimes it took a second heat blast to get the chip to recognize. Altogether I've placed 10 chips on Jalapenos this weekend

[Inaba]

Please, please, please make sure you actively cool the 1850 & the MOSFETs if you decide to do this.  The thermal plane on the board is really effecient and you will overheat both if you are running that many chips unless you have active cooling on both the top and bottom of the board for both the chips, MOSFETs and especially the 1850, which will cause intermittent and strange lock ups if it overheats.

[elaramus]

Please, please, please make sure you actively cool the 1850 & the MOSFETs if you decide to do this.  The thermal plane on the board is really effecient and you will overheat both if you are running that many chips unless you have active cooling on both the top and bottom of the board for both the chips, MOSFETs and especially the 1850, which will cause intermittent and strange lock ups if it overheats.

How different is the Little Single heatsink from the Jalapeno's? Any chance we could procure some on the open market? What cooling solution is used on the Singles/Little Singles for the MOSFETS and the 1850's?

 I'd absolutely love to take one, two or more of these guys from a two chip solution all the way to a full blown little single. I have 4 more Jalapenos on the way from the black friday sale and 20 chips left to play with.

[Photon939]

Please, please, please make sure you actively cool the 1850 & the MOSFETs if you decide to do this.  The thermal plane on the board is really effecient and you will overheat both if you are running that many chips unless you have active cooling on both the top and bottom of the board for both the chips, MOSFETs and especially the 1850, which will cause intermittent and strange lock ups if it overheats.

How different is the Little Single heatsink from the Jalapeno's? Any chance we could procure some on the open market? What cooling solution is used on the Singles/Little Singles for the MOSFETS and the 1850's?

 I'd absolutely love to take one, two or more of these guys from a two chip solution all the way to a full blown little single. I have 4 more Jalapenos on the way from the black friday sale and 20 chips left to play with.

The little singles and Jalapenos are now shipping with the plain aluminum heatsink, same one. They probably ran out of the heatpipe sinks and didn't want to order more. The hole spacing is rectangular so I'd have to measure the spacing to have an answer to the second question. My early revision Little Single came with heatsinks on the mosfets, nothing on the 1850. My new version Single came with no heatsinks other than the two main ASIC group sinks.

[lightfoot]

The stock 6A adapter gave out after the 4th chip, so I cut the plug off and wired it to a PCI-E cable right off my computer's power supply. No issues with power after the mod. Adding 4 chips had both Jalapenos up to over 60C. The 5th chip needed additional cooling (up to 78C) and the 6th chip put the temps off the map.

Yup. Some people have reported it blowing up at 4 chips, I ran mine with 5 for over a month before I got concerned about the ground plane faulting in it. A cheap-o PCX power supply (the $20 one on Ebay) with 18 gauge wires is a bit light for a 5 chip jally; I had to wire 3 of the 12 volt lines in series to make it work. Better supply should not have a problem.

I found applying thermal compound between the aluminum heatsink plate and the Jalapeno board then placing an active cooler on top (running the Jalapeno upside down) provided exceptional results... Both are hashing over 23Gh/s

Yes. I literally was getting equal cooling running with the Al heatsink on the bottom, then the board upside-down, then a copper heat sink thermal pasted to the board, then the fan on that. Even now in the case, putting thermal compound between base and plate, then plate to board turns the whole unit into a heat sink.

P.S Temps used: preheat the area with the wand for 10 seconds, then 90 seconds @ 460C with the wand 1/4 inch above the chip. The chip will float as the solder balls melt (DON"T FREAK OUT) as the remaining balls melt and adhere to one another the chip will realign itself. Patience was the key to my success. Sometimes it took a second heat blast to get the chip to recognize. Altogether I've placed 10 chips on Jalapenos this weekend

This was my biggest mistake initially, and cost me three chips total. You need a *lot* of heat, you need to keep it going, and the balls have to melt completely. I kept thinking I was going to burn the board, that led to pulling the heat early and winding up with an off-center chip by a micrometer. Then not enough heat again causes the chip pads to rip out on removal).

I need to keep reminding myself "Insane heat", the chips can tolerate a max temp of 500F for 120 seconds or so. That's chip temp, not air temp; as an experiment I sat there for 2 minutes with heat at 450c and the chip was only at 400 or so F immediately after heat removal.

Ah well, that's why we call it "learning" :-)

[lightfoot]

Please, please, please make sure you actively cool the 1850 & the MOSFETs if you decide to do this.  The thermal plane on the board is really effecient and you will overheat both if you are running that many chips unless you have active cooling on both the top and bottom of the board for both the chips, MOSFETs and especially the 1850, which will cause intermittent and strange lock ups if it overheats.

And the other yup: The MOSFETs *MUST HAVE GOOD HEAT SINKS* once you go above 4 chips. This is going to be what kills these boards BTW; and is a good reason to wire a fuse in series with your massive power supply (since otherwise you will have the smoking/flaming FETs).

Speaking of which, plug time: If you're really doing this put a simple 3ag fuse in series with your new power supply. Calculate it by the number of watts your unit draws (20 watts+15 per chip) and divide by 12. So a 5 chip unit is 20+75=95 watts/12=8 amp fuse.

That way you won't feed more current than that into a short.

C