Topic: Hacking a BFL Jalapeno to 32GH and beyond....(???) - page 20. (Read 54329 times)

No question. I forked this thread off because I really felt like I was bogarting the chip flashing thread. I'd recommend reading all of it; there is amazing stuff there.

And the answer to your question: 99.9% of the changes can be done in std_defs.h. That's the key file to mod.

C

This thread is really built on the back of this thread https://bitcointalksearch.org/topic/experimenting-with-jalapeno-firmware-236875 that's where all of this really got rolling... LOTS of good reading there.

Ok I've looked through some different source code lately. Which file reads like a regular config file? Is it when you actually go to flash you can edit it or prior as I would think? Could use a little guidance on the flashing as I have never flashed a jalapeno. I've read the thread 10 times atleast but it doesn't specify where you edit the code or on what file. Thanks guys I'm loving every second of this shit, and if you find a new vendor let me know I'll clean them out.

Lol... it's worse than that! I bought nine... installed 4 (2 on each jally) then started on the 3rd one the next day...
Long story short installed 2, removed them, blew out one of the original chips through my own stupidity, removed that chip, and installed my remaining 3 unused ones...

So 3rd jally had 4, 1 original (rev 1 chip) and 3 rev 2 chips... working great until this morning... one light flashing crazy... unplugged power then plugged back in... now running on 3 chips 1 original and 2 rev 2 chips

Another sidenote... I compiled my own 1.2.9... I didn't turn off error checking just changed it from 10 to 1... and obviously changed to "little single"...

As for the chips, that's a problem, didn't realize that BFL was not selling them anymore :-) Drat, kind of sucks to come all this way and have things go poof. Unless of course we can source some more chips. My guess is the chili people sucked them up.

Note: It's also possible they stopped selling them because they anticipate an orgy of sales tonight. We shall see.

C

Naah, I don't think anything we did here points out a negative light towards BFL. In fact, I'd say it's pretty damn positive. They did a lot in order to give us what we paid for, I'm betting that they could have gone with the original design at 4gh but instead they gave us a more expensive board, capability to go to 7-8gh, and a much better case. Same price.

Fact is they delivered when a number of other vendors didn't. Avalon has their own issues, there are at least two other groups that just folded (and took people's bitcoins), BFL seems to have forged through the problems to deliver.

As to the number of Jally failures, I really am wondering if they are really hardware fails, or if people are installing the 1.2.9 code for single boards and not recompiling to little_single. That will do exactly what people have been reporting, in which case it's not BFL's failut it's simple ignorance.

Oddly enough this is why companies pot their equipment and then refuse to divulge source. Because people who screw it up then try to tarnish the business' name. Better to be like Sony or Apple and lock the whole thing away from you...

Such is life.

C

I like what he's saying with the cross pattern. Sounds like better distribution of heat and sink pressure across all the chips. Less of a chance of chip next to chip uneven height.

I think that's how I'm going to do it to not stress any components to their limit. Mostly I would want a lot more chips in the future. Problem is BFL isn't selling anymore chips!!!!!  I think they want to pinch this in the butt, as just adding chips looks like a simple mod. It also brings to light a lot of their blunders. (WTF did they expect with a bunch of hackers, crackers, and exploiters getting their shit.)  Good luck finding 65nm chips guys, these things are going to be worth more than gold soon.

*wow* Well, they were kind of odd ducks, but *wow*. Out of curiosity did it have a lot of errors on it and a lot of engines shut down? Just wondering if the dead engines put out full heat at idle, and leaving them running blew the chip.

So in a nutshell you had six chips, modded 3 jallies, two are running with 4 cpu's one with three including a replacement for one of the crappy onboard ones. You're still doing very well in terms of odds.

C

The infamous 3rd jally has 4 chips mounted in a cross pattern, aka, the easy pads to use... all 3 of my jallys are modded this way. The light that was flashing crazily this morning was the light in the 7th position away from the power jack, 2 away from the usb jack.

I'll try to get some pics from my adventure up tonight or tomorrow... I'm going to try to reball the 2 chips I have... the original chip I blew up had the chip actually separate from the bga base...

Try reflowing all the chips; I can't believe that they're not seated. And re-viewing my data, I see that even blasting them with 450c was not raising the temps much past 425F, which is what the temp should be on the bottom of the board to verify proper chip flow.

Side question: How many chips are on the board now, and can you post some pics?

C

Just an FYI...

Came down to check on everything this morning...

The last jally I did ended up at around 14.9gh after all the fun times (read above)... cgminer was reporting values between 30gh and 120gh...obviously not correct and one hashing chip light on the back (by the barrel connector) was flashing like crazy... don't get me wrong I wish it was actually hashing at that rate

I reset that jally and it did disable that chip on power up. Hashing at 11.2gh now...

Later tonight I'll reflow that chip and report back... hopefully it's just a bad connection

I'm doing some experimenting on chips with the air tools while waiting for my next batch to come in. I'm wondering if I am going too hot here for some reason.

Test bed right now is a bunch of 1gb DDR, 2006 vintage memory modules. Granted they don't have the ground plane, but I can float them off the board with no problem with about 10-15 seconds of heat at 350c. At 300c they do not budge no matter how long I leave the flow on them. At 350 they come off but only after applying a fair bit of heat. At 375 they come off even quicker. Low air flow rates seem to be fine.

400 should be plenty. I'm not sure where my problem is.

Old DRAM chips (1990 vintage) come loose in 15 seconds or so at 375.

Putting a dead BFL chip on it's back, putting flux on it, and reflowing it at 350c at low airflow rates makes the solder liquify without much time.

Then again these *may* be non-ROHS equipment. I'm thinking right now of going to Staples and getting some piece of electronic crap and floating those chips. That might be the problem, according to Freescale documents you need to have 350c for lead solder, 400-500c for ROHS in your ovens. In which case I might need even *more* heat.

Another data point: Blasting the BFL chip at 450c air temps for minutes showed up as a temperature of 440F on the device's top side. Which is only 232c, which is barely the BJT (ball junction temp). Another reason temps might need to be higher than I expect.

I need some late model garbage. Hm, what can I buy at Staples that is ROHS, loaded with chips, and cheap as hell (yes, I know, an X Box :-)

Nice. I see BFL had something similar for about $18 bucks. I was going to make my own cable but a real one would be nicer.

Let us know how fast it comes in, looks interesting. Since I'm buying another Jally, it would be nice to power both on this ATX supply sitting here in its box.

C

https://bitcointalksearch.org/topic/wts-atx-psu-to-25-or-2155mm-12v-coaxial-plug-power-cables-new-product-74397

^ I'm planning on 2 adapters from this guy for powering the units.   Chips I would love a group buy but it would prolly be months......

Whoah, wicked. My protocol for programming is to plug the jally in last, have ATMel all ready to clear the memory, and as soon as I plug in the jally I hit the "erase memory" button so it clears the program. Then the chips are in idle mode.

As a backup, I have the board fan pointing down. Have not lost a chip yet, but your experience points to having *something* on the chips. Can you post a photo of the blown up chip?

As for duds, I can't believe that the solder will melt with only 120c coming down from the top. Granted hot air sucks, but the chips really are mounting to the board, was there any damage to the chip pads when you floated off?

I doubt the chips are bad.

Yep. That board is extremely efficient on the heat end; how high are you running the bottom heat BTW?

There are; I think the latest run of chilis and all the people boosting their jallies are sucking down supply. Damn bitcoin has such a shallow market that when I write about boosting units everyone buys all the chips :-)

That said, there is arbitage: Chili purchases usually leave a lot of left over chips. We might want to have someone ask the chili vendor if they have left overs from their sales. Or go in on a big buy from BFL, but at $75 a chip they are really not worth that much. Well, sort of. Maybe they are.

C

What are you using to power your Jalankenstein?

Well fun night all around! Blew up one original chip on my jally Thought I would get cute and try to flash the firmware without a heatsink real quick... didn't notice it error out and reset... well got it unplugged quickly enough that the other original chip was good, but the magic smoke leaked out of one Put 3 chips on... 2 ended up either duds or heated too long. Floated those chips off, cleaned up the pads on the board, and put my last 2 chips on...

Couple things I found out tonight...

This board soaks heat very, very well! High bottom heat really helps installing chips, cleaning pads, and removing chips. Cleaning the pads was quite the chore until I got the bottom heater high enough...

End of story jally 3 is at 14.9gh... I'm gonna go mourn the loss of my 3 chips now

Looks like there aren't any more chips floating around... 

Oh you're going to have a *FUN* time. This is getting more interesting every day.

I just noticed they're doing another Chili board run. $350 or so for the board, you add the chips and the heat sink. Sounds like a "good" deal but I wonder if anyone noticed two things:

1) This thread points out that Jallies can jump from 2 chips to at least six.
2) Butterfly Labs is running a sale on Jallies with delivery in December for $200 on Friday.

You get the board, fan, sink, case, even the power supply. I am really. Really. Really tempted to pick another one up and stick a pair of chips on it. In fact I'm in on at least one; anyone else?

I've got 10 chips in the mail, should arrive before the reflow tool and the jalapenos. I just dumped part of my BTC savings and some cash totaling almost a grand. Here we come 60Ghs, new heat pipes ordered too. I'm as giddy as a kid for christmas. Plus the BTC prices are going nuts and all my FTC and XPM is worth a lot of cash now. Didn't think I would see the day FTC made me a couple hundred bucks.

Looks like a pretty good thing to have if you need to do reballing. I'm split between something like that and a full-bore IR rework station. I can do things with the air tools, but that ability to focus heat down on a single component could be very helpful when I am replacing transistors as small as the little capacitors on the BFL boards.

How many chips do you have to do?

C