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

[lightfoot]

I believe you've seen the molex to barrel adapter that takes 5 minutes to make. It's a lot safer power if you have the PSU for it. Yellow 12V line to the inside shielded wire, and the two center grounds on the molex to the outside braided ground.  I wired up 4 of these just to get my miners up while waiting on cablez. Not that he took long, 2 days maybe on his end and an amazing looking product!! 

Yes, but I am just too damn lame to order a molex socket from somewhere. So what I have been doing is picking up old D type connectors from Radio crap, wiring those to power whips from these crap supplies, then plugging them into my power supplies. Problem is I have two of the 20gh class systems using the one string of D subs, so I finally gave up and clipped the cable on the supply so I could use one of the SATA lines. (They are all on the same rail, but one should still balance the wires since those have resistance as well)

So in the end you're going to hit 8 chips right? Have we thought about cooling the FTDI chip with a simple sink? If you need any help with water cooling let me know, I love the stuff. I didn't go that route because of ROI. Just be super careful if you drill into the water block not to permeate the chamber. Don't mix metals between sinks and radiators ETC. ionization will happen and slowly deteriorate everything. It's usually suggested to run the system for hours to leak test. The thermal conductivity of water exceeds air something like 10x so this makes me giddy.  Air vs water let my processor go from 4.0 Ghz to 5.2 Ghz in the end it is usually efficient enough that the next hindrance is hardware related as there should be plenty of overhead for heat. 

Well, I have 4 chips coming, two old style chips in the box, and three chips that need to be reballed (one old and two new). The problem now with the old style chips is that fitting them on a board could be tough; since they are tall the newer chips all have to have heat pads on them to even up. What I might do is try putting the first reballed tall one on the "danger board" (the one that some poor guy nuked to try and put chips on it) and if that works put the three on that board and call it "done" (5 of the chip pads are seriously damaged. Sad.)

The problem with the FTDI IMO is not that it generates heat, but since it's on the ground plane it has to deal with the plane's heat load from the FETs and chips. Add to that the genius idea of blowing 50-60c air right on it (with the fans down) and you have a recipe for fail. My next cooler is going to be the corsair big one, I'm thinking of just strapping it to the bottom of the board so it can pull the heat off the bottom. In theory I should be able to get away with only a little AL sink on the top, which will have to be the bottom since I bet you can't run a water block upside down.

The corsair and 120 that I am using now are sealed units, so corrosion inhibitors are probably in there. What I really *should* do is take them outside, then immerse the radiator in a 5 gallon bucket of water in the shed with a garden pump circulating the water to the 60 gallon rain barrel outside. Even better would be to immerse the radiator straight into the barrel, but in either case if the water froze solid I'd lose cooling power. Hm.... That is crazy actually...

C

[lightfoot]

What I will probably do is put two of the new chips on one board, one on the other, and run 7 chips on one water cooled, six on the other with air cooling. Then hold the last chip in reserve to see if I want to go to 8.... :-)

C

[HellDiverUK]

I believe you've seen the molex to barrel adapter that takes 5 minutes to make. It's a lot safer power if you have the PSU for it. Yellow 12V line to the inside shielded wire, and the two center grounds on the molex to the outside braided ground. 

The two Jally PSUs I've butchered have had a red and black wire inside, no shielding.  Red is centre positive.  I just solder on to an old molex fan cable (I've heaps of them from years ago), bit of heat-shrink, job jobbed.

[lightfoot]

I believe you've seen the molex to barrel adapter that takes 5 minutes to make. It's a lot safer power if you have the PSU for it. Yellow 12V line to the inside shielded wire, and the two center grounds on the molex to the outside braided ground. 

The two Jally PSUs I've butchered have had a red and black wire inside, no shielding.  Red is centre positive.  I just solder on to an old molex fan cable (I've heaps of them from years ago), bit of heat-shrink, job jobbed.

True, if you crack the case you will find that. However oddly enough I have a use for old jally supplies, one is powering a beautiful halogen lamp that had it's power supply die years ago (4 amps at 12 volts needs a big power supply). The other is for a peltier cooler based fridge that had it's power supply (12 volt 4 amp) die as well. So oddly enough how can you screw up these things if the jally supply decides to explode :-)

They would also make good 12 volt battery chargers oddly enough.

C

[Drug5bitz]

Radiator in the barrel maybe not, outside better idea. Mine currently heat my apartment, only a 1 bedroom but still. Water cooling is amazing because it works in any orientation. I wasn't aware you're using one of those self contained units. I'm not a huge fan but I have owned one, they're extremely simple to use. Never worry about adding coolant or bleeding the system proper, as for actual life mine was fine for the year I owned it. The only reason I wish I was 100% water cooled is the 10 fans running for my miners. Do a google search and you can find what I assume is the water system they used in the minirigs. A lot more coverage, kinda wished this was an aftermarket option.

[lightfoot]

Ok. Given the water is almost frozen, it would be cool to run a chili with frozen chips.

But I should think about the summer that's coming up: One interesting thought here is to mount a heat exchanger to an old 5000 btu air conditioner, and water cool from that directly. I wonder how hard it would be to build a water to air conditioner exchanger or just put the AC condenser in a big fish tank and put the water radiators right there. Or just plumb the water straight into the tank. Wrap it wih insulation and your cooling problems would be solved.

Hm. Well, that's the summer when this is all going to be running in the shed on batteries and solar panels. Speaking of which I should order a new battery bank tomorrow so I can take the solar tax credit. Hm.

C

[freddyfarnsworth]

Ok. Given the water is almost frozen, it would be cool to run a chili with frozen chips.

But I should think about the summer that's coming up: One interesting thought here is to mount a heat exchanger to an old 5000 btu air conditioner, and water cool from that directly. I wonder how hard it would be to build a water to air conditioner exchanger or just put the AC condenser in a big fish tank and put the water radiators right there. Or just plumb the water straight into the tank. Wrap it wih insulation and your cooling problems would be solved.

Hm. Well, that's the summer when this is all going to be running in the shed on batteries and solar panels. Speaking of which I should order a new battery bank tomorrow so I can take the solar tax credit. Hm.

C

Since you are all on water to cool the Custom boards you guys are creating, Here is some automotive tech you may or may not know. Ethylene glycol pure has to be mixed 50-50 most comes that way now. It will raise the boiling point to 252degrees with a 16lb pressure cap.. It will also conduct heat far better than plain water, and give corrosion protection till it gets contaminated with sulfides from burning fuel and small amount of gasses leeching into cooling system on a engine.. no need to worry on that. Should last forever. Also would never freeze in your conditions.
You can drink it, tastes sweet (rain Barrel) but you will regret it later. So repurpose that barrel.
Thought these chips run best at 46 to 50 c tho, so you do not want them to cool.

[HellDiverUK]

Anyone 'fixed' the 1.2.9 firmware to run a bit quicker?  My 5GH Jalapeno is the 'rev 2' board, and the 1.2.5ck firmware doesn't work on it.

My 8GH Jalapeno has 2.9.2 on it - is there any way of pulling it's firmware and putting it on the 5GH unit?


Or does anyone have a list of firmwares and what they work on and what they do?

[lightfoot]

So last night I had my first fail in awhile; I tried to put an old style chip on a completely burned up board. Found a pad that was mostly flat (repeated attempts by previous owner to add chips were spectacular fails) centered the chip, and hit the heat.

Shorted the 1v line. Odd. So I pulled the chip; the solder under it globbed together in one place. I had flux on it, I think the problem is that without the balls under the chip, the chip solder will tend to flow. Also the old style chips seem to have much smaller solder pads on them, the balls might have just run as a matter of course.

Not the end of the earth, just means I have 3 chips now to practice reballing on. I'm going to try one of them on this board again later; need to figure out how to get these chips to go on a board with no solder for the heck of it.

But no more experimenting with good chips. Have 4 more showing up today, I think this will be the last of the chips I will be installing for awhile.

C

[lightfoot]

Ok, we're up to six chips, but it wasn't easy and this is where it gets more real.

In a nutshell, once you go to six chips odd things will start to happen. First you will hash at 24gh. Whee. Second your custom fan assembly with heat pipes will start feeling the heat as board temps hit 70c. Remember they were in the 40's earlier, now they are in the 70's. It's a bit warmer up here.



Note the larger heat sinks on the power FET chips. They read in at 120 f right now, without them I think the FETs would be on fire. Likewise the fet driver chip is getting hot, and the ground plane of course is at 70c which is close to 160f. I put the Al heat sink on the bottom there for a reason; the plate on the bottom with heat sink compound was getting to 120f+.

In a nutshell, the issue is one of heat load: The more you put on the board the harder the board has to dissipate it. I thought about going to seven, but I think I am going to wait for the water cooling block before I try for that.

In addition I screwed up a chip; put too much flux on it and the flux bubbled out from under the chip while under heat. Very very bad, it pushed one of the balls into another on the outside and created a logic short so I had to pull the chip. It's ok, just needs the official reballing treatment. Another chip in "reserve".

When the water cooler comes I'll try putting chip 6 and 7 on the other jally; this one will not be able to go past 7 at this point due to the chip coming off the board so I'm kinda stuck. I'll leave it at six for now.

Update: Slight miscount, the other jally has only four chips, so I took it to five with FET heat sinks. Running fine. Next up will be to take it to six when I get the water cooler. At that point I am running low on chips here....

C

[dentldir]

When doing large BGA chips, there is usually a flux activation time during the heating cycle.  You just heat at the chip at the activation temp of the flux which is well below the soldier temp.  Once its done doing its thing (bubbling and stablizing), you ramp up the temp from there to do the soldiering.  At that point, the flux and the soldier mask are working together to keep everything in place for you.

I did manage to fix a cold soldier job I did using the info in this thread.  So I've got a 4 chip Jally now just under 15Gh/s.  Haven't been brave enough to push it past that yet.

[elaramus]

Update on BFL Black Friday Jalapeno purchase:

I bought 4, and haven't had a chance to really look at them until today. On arrival I had someone unbox them and plug them in and the 4 Jallys put out an aggregate hash rate of 32 Gh/s. 2 Were hashing above 9 and the others bounced between 6 and 8.

I've only opened two so far. One of the 9Gh/s models was running extremely hot (75c) so this was my first victim/patient. I opened it and found 3 chips installed (new version), firmware 2.92 label, a test date of 12/21 (so much for "in stock for immediate delivery"), and the fan blowing down onto the chips. I flipped the fan, replaced the thermal pad with artic silver alumina, put a layer of alumina between the bottom of the board and the aluminum plate and it is now running at 9+ Gh/s at a much more reasonable 61c.

The second Jally I opened, had the old school chips on raised PCB, it also had 2.92 firmware. I figured I'd try to add a chip to it while it was open. It worked! I was able to add a chip and have it recognized without flashing the Jally (The Dragon was standing by just in case).

I am working on the 2nd 9Gh/s jally now. If it has the new rev of chip, I will try to take it from 3 chips to 4 (again without flashing firmware) then I'll move onto the last Jally which I suspect has only 2 chips and will add one to it.

Of note, while inspecting the casings, one of my Jally's has 8 pinholes in the case to allow you to see each respective LED for its associated chip. Only 1 of the 4 had this the remaining ones only had two holes to view the chips..

More to follow..

I'll have some spare chips available for sale when I'm done. If interested, PM me. I expect to have at least 6 or 7 available.

[lightfoot]

The safe max seems to be five chips, with that you can run at 20gh, and still use the stock heat sink with the fan blowing straight down full speed. Going to six causes all sorts of weird things.

So lard em up with chips.

C

[lightfoot]

In the meantime I've been bbqing a test board here trying to resolder chips that have already failed install. The results have been far less than spectacular.

The big problem is the chips will not align, and due to the board differences in height between the remaining solder balls it's impossible to place the chip. So the result is a chip that causes solder shorts, blowing the 1 volt line. I've gotten to the point where I can just sit there and measure the resistance of the board's 1 volt line. Anything lower than 1 ohm means boom.

Oh well, I think I have to wait for the um.. stencil. Once I have that I should be able to do some real work by removing all the solder from a board thing, then putting on a perfectly balled (hah) chip. I have 4 chips that have been screwed, so I should be able to get something out of this.

Eventually my stuff will come from China. Just. need. to. wait. :-)

C

[lightfoot]

Side note: That lentbt guy on Ebay is running out of chips. If you want to do this, I'd suggest checking in with him. I'm going to wait till my reballing stencils come in from China and work with that for awhile. :-)

C

[lightfoot]

Well, in the "lessons cost bitcoin" category I finally figured out why I have been failing in mounting this chip on the danger board.

This is one of the "early model" chips that are on a tiny carrier and do not have a logo. Got three of them for .12btc a few weeks ago. And one of them I tried to mount on the danger board. But each time I tried to clean it up and put it down, the chip would short the 1 volt supply line. Every time. Over and over. I thought I was a complete loser.

Then I tried to put another one down on a totally good board. Pre-heated, aligned the balls perfectly, applied the heat to the top, after 30 seconds looked in and saw a tiny solder ball on the chip carrier.

CRUD! Now I know what has been happening: Instead of bonding the chip to the carrier board with high temperature solder, they used normal solder. Which means the heat from my air gun would heat the chip and not the board under it and as a result the solder on the chip itself would short under it on the carrier.

CRUD AGAIN! That's why this other chip was shorting. And why this chip I was putting on didn't even mount to the board, just made a total mess of things. So I removed the solder on the board, swore for awhile, and put the chip away.

I'll sell all three of them for .04 btc, as either souvenirs or if you think you can get them going. It's just the cost of doing research, but I would *HIGHLY* recommend that you do not buy old style chips unless you have a full reflow oven and pace placement tool.

Lesson learned. One good chip left, two chips in need of reballing. So I should be able to put this last chip on the 5 chip board, making both 6 chip boards, then if I can get these other chips reflowed I'll try putting them on the danger board. :-)

C

[freddyfarnsworth]

Just to let you know, watching and reading with great interest, you are cutting edge on info on homebrew projects like this, your pain - my gain

Found my old toaster oven in storage and my industrial heat gun (paint removal- looks like a hair dryer), also my old laser temperature tool. May give it a whirl on some stuff.

Reballing seems to be mandatory from your experience tho, was hoping to avoid all that.

[lightfoot]

Feel.free to drop some bitcoins in the research bowl or send repairs my way. However one may be able to mount a chip without reballing. My error was working with these older chips.

We learn by doing. And people have donated time coin and chips so it all works out...

[technocoma]

Interesting read.

[elaramus]

BlackFriday Jally Update 2:

So finished up my splicing and dicing on New Years Eve. I decided to make 4 3-chip Jally's so I could salvage their resale value:
Here's a screen shot of the final output. The two doing 11+ are the 2 chip Jallys that I added a chip to.


Of the 4 received, only 1 had the original old style chips (notice the one running @ 69C, that's the one). The remainder having a level plane all made good contact with the stock heat sink. I decided to go with arctic silver alumina for thermal compound and after flipping the fan and adding thermal compound beneath the bottom of the board and the aluminum/zinc back plate, all of my temps are good.

Here's a shot of the 8 hole plate I mentioned on the first post:


It was nice to see the boards with the newer style chips. I don't see any immediate advantage of the 292 firmware other than you can add chips without flashing. Here's a shot of one of the three chip boards I received, I received two 3 chip and 2 two chip models.




And I didn't go crazy so I have 13 chips left.. Look for a thread under Computer Hardware if you want them.

https://bitcointalk.org/index.php?topic=395129.new#new

PS: I'd ordered two PCI-E to 3 Plug custom power cables when I placed my order. They were not in the box. Logged back into BFL's site and checked my invoice, and it clearly shows I paid $29.98 for the cables and also indicates that they were shipped. I wrote an inquiry to office@butterflylabs.com, but to date have not received a reply.