Restored a 2013 Butterfly Labs Jalapeno and put it on a live solo-mining stream

[MissyMiningCat]

I got into mining back in the Butterfly Labs era, when the first ASICs were landing and nobody really knew what they were doing. Put my money into a Black Arrow Prospero, which, if you were around then, you know how that went. Late, hot, loud, and a long expensive lesson. It left a bad taste, and that wasn't just from the smoke when it tried to burn the house down. I walked away from mining for good. Or so I thought.

The Bitaxe generation pulled me back in. Open hardware that actually works, that you can tinker with and understand. And it got me wondering whether the old first-gen gear could still run. Not just power on, but run properly, clean and hashing at spec. So I started tracking the old machines down and restoring them, and I've put the first one on a 24/7 webcam with live stats: vintagehash.com

It's a Butterfly Labs Jalapeno, circa 2013, 5 to 8 GH/s, solo mining to my own node. Live hashrate, share log, temps, and the odds of it ever finding a block (spoiler: about 1 in 73 trillion). Figured people here would appreciate it more than most.

The restoration:

It turned up filthy. Years of dust packed into the heatsink, thermal paste gone to chalk, running hot. Stripped it right down, cleaned everything out, fresh paste and pads, new fan. Powered off a dedicated 12V 10A PSU on the barrel connector (USB for data only, these don't run off USB power). Then the usual fight getting an old cgminer build to actually talk to it, plus the Zadig/WinUSB driver dance. Anyone who's done this with BFL gear knows exactly what I mean.

Picking the runner: I've got a few Jalapenos, so I tried them to see which ran cleanest, going by cgminer's Device Hardware% (bad nonces over total, the number that actually matters, not raw HW vs accepted). The one on the stream came out best by a good margin, sitting around 0.1% once it had settled. Honestly surprised me how accurate a cool, properly re-pasted chip stays, even on hardware this old.

Everything on the stream is live and verifiable. Stats come straight off the miner, and the mining address is on screen so you can check it on the chain yourself. If it ever finds a block, it happens on camera.

And yeah, the irony isn't lost on me that the machines that burned me (nearly literally) are the ones I'm restoring for fun now. I've still got the two Black Arrow Prosperos. One of them actually runs again. There might be a reckoning on camera at some point.

Got more queued up as well, more Jalapenos, a BFL Single, a few Monarchs. Happy to go into any part of the process in more detail if anyone's interested.


Anyone else still crazy enough to be running first-gen gear from back then for fun, or am I on my own with this?


Come watch it not find a block, forever and ever and ever: vintagehash.com

[NotFuzzyWarm]

I too still have two jalas which were the first miners I bought - they are what got me hooked on mining....
When you had yours apart you should have seen several empty spots on the pcb. If you have some spare chips from a dead miner and are handy with hot air soldering you can mount more ASIC's in those locations. Think there are maybe 8 vacant spots? As I recall from old posts the Vcore regulator will handle at least 4 and up to six more. Of course you will need a bigger psu for it...

[MissyMiningCat]

Thanks NotFuzzyWarm, great to hear from someone who started on Jalas too.

I've seen the empty spots on the board. They look like the same footprint as the populated chips, which is what makes it so tempting.

Honest answer: my hot air rework isn't up to it yet, and I'd rather practise on junk boards than risk a working one. If I do get there, I'll document it on the stream. A super Jalapeno climbing past 7.8 GH/s on camera would be fun to watch.

[philipma1957]

best share 210,000
you need to be

132,770,000,000,000 only 132,770,000,000,000/210,000=

632,238,095.238 x 1 too small 

[MissyMiningCat]

best share 210,000
you need to be

132,770,000,000,000 only 132,770,000,000,000/210,000=

632,238,095.238 x 1 too small  


https://www.talkimg.com/images/2026/10/12/UxAnUv.png

So you’re saying there’s a chance....   

Had to fix my odds maths after that though, turns out I was being too pessimistic. Recrunched it: 1 in 121 billion per block now, not 73 trillion. So only 2.3 million years instead of 1.4 billion. Practically imminent. Get the champagne on ice.

[FP91G]

Here are the specifications for this device:
Hashrate: 5.3 GH/s. Some later revisions or slightly overclocked units could reach up to 7GH/s.
Power consumption: around 30 W.
link

My point is that modern solo mining devices have a hashrate 1,000 times higher, so the Butterfly Labs Jalapeno Bitcoin Miner is of historical interest only; using it for any kind of mining is unprofitable.

[MissyMiningCat]

Here are the specifications for this device:
Hashrate: 5.3 GH/s. Some later revisions or slightly overclocked units could reach up to 7GH/s.
Power consumption: around 30 W.
link

My point is that modern solo mining devices have a hashrate 1,000 times higher, so the Butterfly Labs Jalapeno Bitcoin Miner is of historical interest only; using it for any kind of mining is unprofitable.

Ha, that’s exactly it, it was never about profit. I know running a 2013 ASIC against a 2026 network is a lottery ticket that won’t win. The whole idea is seeing if 13-year-old silicon still runs clean, not the pay-out, and having a bit of fun while I’m at it. Modern gear can do the earning, this does the trying.

The way I see it, obsolete just means it can’t compete, not that it’s broken. The chip still does exactly what it was built to do, searching for a block, same as the day it shipped. The odds don’t matter to it. It just keeps trying.

I’ve got the champagne on ice for when it lands, roughly 2.3 million years from now. No rush. 😉

[FP91G]

Here are the specifications for this device:
Hashrate: 5.3 GH/s. Some later revisions or slightly overclocked units could reach up to 7GH/s.
Power consumption: around 30 W.
link

My point is that modern solo mining devices have a hashrate 1,000 times higher, so the Butterfly Labs Jalapeno Bitcoin Miner is of historical interest only; using it for any kind of mining is unprofitable.

Ha, that’s exactly it, it was never about profit. I know running a 2013 ASIC against a 2026 network is a lottery ticket that won’t win. The whole idea is seeing if 13-year-old silicon still runs clean, not the pay-out, and having a bit of fun while I’m at it. Modern gear can do the earning, this does the trying.

The way I see it, obsolete just means it can’t compete, not that it’s broken. The chip still does exactly what it was built to do, searching for a block, same as the day it shipped. The odds don’t matter to it. It just keeps trying.

I’ve got the champagne on ice for when it lands, roughly 2.3 million years from now. No rush. 😉

I have some old PCs at work, over 20 years old- that I use to test hard drives with MHDD. Why shouldn't old hardware work for Bitcoin mining if the algorithm hasn't changed?
If you can find a pool that accepts that kind of hashrate, then you can have some fun with it.

[MissyMiningCat]

It would be fun to see how a 2013 era CPU compares to a modern one.

Funny you mention it, as part of building a new getwork-to-stratum bridge for an old Block Erupter Cube, I needed something to test the bridge with while troubleshooting the real hardware. So I wrote a little CPU miner that pointed at my own Public Pool hosted on my node, and it worked fine, shares went through.

It was gloriously slow though, only about 620 kH/s, but that’s because I wrote it in pure Python purely to prove the bridge worked, not to mine efficiently and it was only using a single core, too. A proper optimised CPU miner (something like cpuminer-opt, using modern SHA hardware instructions) would be hundreds of times faster. I’ll have to dig it out and get a proper figure.

And you’re spot on, the algorithm hasn’t changed since day one, so there’s no reason old gear won’t still hash. It just hashes slowly. Really slowly.

[MissyMiningCat]

Here are the specifications for this device:
Hashrate: 5.3 GH/s. Some later revisions or slightly overclocked units could reach up to 7GH/s.
Power consumption: around 30 W.
link

My point is that modern solo mining devices have a hashrate 1,000 times higher, so the Butterfly Labs Jalapeno Bitcoin Miner is of historical interest only; using it for any kind of mining is unprofitable.

Ha, that’s exactly it, it was never about profit. I know running a 2013 ASIC against a 2026 network is a lottery ticket that won’t win. The whole idea is seeing if 13-year-old silicon still runs clean, not the pay-out, and having a bit of fun while I’m at it. Modern gear can do the earning, this does the trying.

The way I see it, obsolete just means it can’t compete, not that it’s broken. The chip still does exactly what it was built to do, searching for a block, same as the day it shipped. The odds don’t matter to it. It just keeps trying.

I’ve got the champagne on ice for when it lands, roughly 2.3 million years from now. No rush. 😉

I have some old PCs at work, over 20 years old- that I use to test hard drives with MHDD. Why shouldn't old hardware work for Bitcoin mining if the algorithm hasn't changed?
If you can find a pool that accepts that kind of hashrate, then you can have some fun with it.

Right, I dug the CPU miner out and got the proper figures.

Swapped my pure-Python tester (0.62 MH/s, single core) for cpuminer-opt with SHA extensions and all cores. The 16-thread Ryzen settled at about 295 MH/s on sha256d, 100% accepted, against my own Public Pool node. So the optimised build is ~480× the Python effort, hundreds of times faster, as expected.

The fun part: that ~295 MH/s, the whole CPU flat out, is about the same as one Block Erupter at 333 MH/s. And that’s all 16 threads combined, not one core, a single Ryzen core only does ~19 MH/s, so one Erupter beats one modern core by ~17×.

Against current difficulty the Ryzen’s looking at roughly 60 million years to a block. The Jalapeno (~8 GH/s) is 27× faster, so ~2.3 million years. The 2013 ASIC comfortably out-hashes the modern CPU, which is your point exactly: algorithm never changed, old gear still hashes, just slowly.

So yes, old hardware mines fine, and a 2013 Erupter will out-hash a modern CPU at it. That’s the charm.

I’ve got a few other machines here I want to run it on too, so if/when I get the chance I’ll put together a proper CPU comparison as a separate thread.

(Champagne still on ice. ~2.3 million years and counting.)

[vapourminer]

I have some old PCs at work, over 20 years old- that I use to test hard drives with MHDD. Why shouldn't old hardware work for Bitcoin mining if the algorithm hasn't changed?
If you can find a pool that accepts that kind of hashrate, then you can have some fun with it.

Right, I dug the CPU miner out and got the proper figures.

Swapped my pure-Python tester (0.62 MH/s, single core) for cpuminer-opt with SHA extensions and all cores. The 16-thread Ryzen settled at about 295 MH/s on sha256d, 100% accepted, against my own Public Pool node. So the optimised build is ~480× the Python effort, hundreds of times faster, as expected.

The fun part: that ~295 MH/s, the whole CPU flat out, is about the same as one Block Erupter at 333 MH/s. And that’s all 16 threads combined, not one core, a single Ryzen core only does ~19 MH/s, so one Erupter beats one modern core by ~17×.

~300 mh/s was around what a HD6850 / HD5830 video card could do. just for further comparison.

i still have my jalapeno, back in the day i ran it with no case and extra fans pointed at it.