I had the same issue a couple of days ago after an overnight failure. I rebooted the machine, inserted the 2Pac, ran Zadiq as Administrator and had it reinstall the driver. Upon completion I unplugged the stick, replugged it, and ran cgminer. It detected and initialised it and began hashing.

I think you'll find they were all built with the same components, but electronic components have tolerances with which they are allowed to vary. The ASICs themselves will have some variation too, because at the sharp edge of processor implementation and design there are differences in the behaviour of individual processor dies cut from the same wafer.

Consider Intel for a moment. They fab a full wafer of Xeon chips intended for the high profit margin server sector. They slice the wafer and set each die in a substrate and wire it up, and drop the resulting chip in a test assembly. The test assembly runs each die at a variety of clock speeds and voltages, and the results of this testing identify a few dies as outright duds, and they are turfed. Some cores in a multicore die don't perform well, so Intel disables them and releases the chip as a hex core instead of an oct core. Some chips perform well on all cores, Depending on the clock speed and voltage each die will tolerate happily, the individual chips are "binned" as various frequencies. Some of the chips will run stably at crazy clock speeds, and these are binned as engineering samples or Extreme Edition CPUs, for which a huge profit margin is commanded.

The Bitmain ASIC chips will be the same. Some of them will go gangbusters and put up with ridiculous clock speeds without going wonky. Others will operate at more pedestrian speeds without failure. I seem to have a pair that will run at 500 MHz without complaint if I give them enough voltage and keep them properly cooled. I'm currently running them at 300 MHz because that produces the most work for my setup.

Philosophical waffling from me aside, if you have one 2Pac that has hardware errors at the same speed as another, and you have adequate cooling, turn the voltage up a little bit on the error-prone stick so it stops erroring. It's simpler than multiple cgminer instances if that's even a thing that works.

Good work. That suggests your trimpot is still in the default position as configured by sidehack during the initial test. At that voltage it should reliably run at 100 MHz, and possibly a bit higher.


Good choice.


Yes, the trimpot configures the voltage regulator stage on the PCB, and the voltage regulator controls the voltage supplied to each ASIC. You move the pot, the voltage changes, and this happens irrespective of the clockspeed you're running at.

At a given voltage, increasing the clock speed will increase the heat output of the chip, and put more load on your cooling solution. My understanding is that each time a transistor in a high density integrated circuit like a CPU switches from one state to the other, it loses or wastes a bit of electrical energy as heat. The more times per second you have the switches oscillating (ie. the higher the clock speed) the more times per second the transistors have to lose a unit of heat energy. So, running at a higher voltage but low clock speed will put less strain on the heat sink than running at that same voltage and a higher clockspeed.
Running at low clockspeed and low voltage will produce less waste heat than running at low speed with high voltage, so we choose the former condition if we can get reliable operation that way.

Generally speaking, CPUs require a lower supply voltage to run at a lower clockspeed. Modern PCs have power saving methods that automatically ramp down your computer's clock speed and voltage when the OS detects that the CPU isn't doing much. A bit like your car engine idling at traffic lights; no point running at redline if you're not doing any work with the engine.

When the workload increases, your PC will ramp up the CPU voltage and frequency, producing more waste heat, and consuming more power. The voltage is increased because operating at a higher clock speed needs a higher voltage supply to ensure reliable switching of the transistors in the CPU. I think this is due to inherent resistance and capacitance in the chip's die, but it's been a long time since I studied this and I am a barely educated buffoon in this discipline.

Your 2Pac stick does not have space age fancy automated voltage control systems which would increase cost and complexity. It does have a trimpot that lets you alter the ASIC voltage between a safe minimum and maximum range, letting you tune each stick for the minimum voltage at which it will stably run at your chosen clock speed. You need to make sure you can keep the chips cool at the voltage and clock speed you've chosen, because a hot CPU wears out more quickly, and a too hot CPU will just fry its delicate electronic circuitry.

Armed with your multimeter, a thermometer, and cgminer logs, you can determine the minimum stable voltage that your 2Pac will continuously and stably run at. Any higher voltage and you're wasting power and have to deal with more heat. Any lower voltage and your stick won't run continuously without errors.

It's possible to run a stick, and dial down the voltage until cgminer starts reporting hardware errors. (HW:0 in the cgminer status line will start increasing to HW:1 and above.) When you see this, you increase the voltage a bit until the HW errors stop incrementing. Then run it for a few hours and ensure you still have no additional HW errors, and you'll probably have found your stable voltage for that clock speed.

I haven't been able to restart a zombie without a replug. I intend to build a relay board controlled by the Pi3's GPIO pins so I can have a script detect a zombie and replug it automatically.


Not with a single instance of cgminer. cgminer takes a single --gekko-2pac-freq parameter and applies it to all detected sticks. It might be possible with multiple instances of cgminer and a cgminer.conf for each to limit which 2pacs a particular instance of cgminer controls, but with the cgminer conf Write function currently broken I haven't played with this. Oh, and I also only have one 2Pac at present.

Apologies if off-site images are bad etiquette.


Set your multimeter to DC Voltage. This multimeter autodetects AC and DC voltage. Many multimeters require you to select AC or DC.

https://i.imgur.com/WRLMglT.jpg

Touch the multimeter probes on each side of the capacitor.

https://i.imgur.com/YBAf1Cq.jpg

Read the voltage on the multimeter's readout. ASIC N1 reads 0.7643 volts.

https://i.imgur.com/O8geaqV.jpg

Touch the probes on the second ASIC's capacitor.

https://i.imgur.com/EFw7opj.jpg

ASIC N0 reads 0.7490 volts.

https://i.imgur.com/vFdvI52.jpg

These measurements were taken with the trimpot in this position.

https://i.imgur.com/U1DY1ax.jpg

Turning the trimpot anticlockwise lowers the ASIC voltage.

https://i.imgur.com/ae02w0g.jpg

Lower voltage measured on ASIC N1.

https://i.imgur.com/pydVlRi.jpg


To reduce PCB flex and improve heat transfer from the ASICs to the heatsink I machined a small aluminium piece and attached heat transfer rubber insulator where the clamp presses against the back side of the PCB where the ASIC chips are located.

https://i.imgur.com/ECVY7Wz.jpg

Side view of the back side clamp, PCB, and heatsink.

https://i.imgur.com/nxPhgS2.jpg

cgminer running.

https://i.imgur.com/4IMz2PE.jpg

I can't offer anything on the unusual LED state, but what happens when you unplug and replug the nonworking unit with cgminer running? Does it detect the new stick?

Troubleshooting could involve dropping back down to 100 MHz to see if both sticks initialise. I found when running at speeds over 200 MHz I might need to replug several times to get the stick to fire up. At 500 MHz it can take several (up to six) before the stick starts hashing. I've also found that when cgminer detects the stick but fails to hash, I get a single blip of the white LED (looks white to my eyes) but no further activity, and the ASIC voltage measured across the cap is about 0.265 volts.

A successful replug at inadvisable clockspeeds sees the initilisation flash, and then a few seconds later the little white LED blinks happily away and the ASIC voltage is whatever I've set it to with the trimpot.

I've also noticed a weird trend with my stick. 325 MHz is its sweet spot, but when I first run cgminer it'll spend about 2 hours hashing at 21-23 avg GH/s, and then climb up to the 35-36 Gh/s that it tootles along at when running at 325 MHz.


Poor little ASIC. You shouldna done that; here were jist a boy.

I'll take some pictures tomorrow. If you can post a picture of your multimeter (and I have a cheapie in my desk drawer at work, because you don't leave home without a multimeter) I can tell you how to set it up.

Regarding the heatsink, I drilled and tapped and old Xeon server heatsink because I planned to do impolite things to my first 2pac. I'll get some pics of that, too. Any heatsink can be drilled and tapped to fit the 2pac, but if you're uncomfortable tapping threads you can simply drill through and use screws and nuts.

Ah, cool. I have a few of those on order because they're neat, but I'm comfortable with a multimeter and oscilloscope. For reference, your stick, or USB device is running at about 9-10 watts. The ASICs are the two BM1384 chips that lie under the heatsink. While your stick is being supplied with about 5 volts, there's a nifty voltage regulator on the stick that supplies a much lower voltage to the ASIC chips. The ASIC chips run from about 0.650 volts up to (in my experience) 0.8050 volts, and this supply voltage is determined by the setting of the variable resistor (or trimpot) near the USB connector.


That's easily fixed, and I can walk you through it if you like. This picture shows you where to measure the ASIC voltage.
https://ip.bitcointalk.org/?u=http%3A%2F%2Fi.imgur.com%2F3k4llxx.jpg&t=582&c=uImuQQqhyMCxIQ

You set you multimeter to voltage DC range. Then you apply one probe to one side a capacitor, and the other probe to the other side of the capacitor. Depending on the position of the trimpot, you will get a voltage reading between 0.650 volts and ~0.8 volts. This is the voltage being supplied to the ASIC chip. Because there are two ASIC chips, and each has a capacitor, you can measure the ASIC chip voltages independently.

sidehack has built a very tolerant device, and I have been pushing it to absurd limits beyond its design spec, and it has not died on me yet. At first you'll feel nervous putting multimeter probes on a delicate device, but if my particular brand of idiocy hasn't fried the 2pac, then I can reassure you that you don't need to be too worried about hurting it with your multimeter.

Each 2pac stick will have slightly different characteristics. With a 25 degree ambient temperature, my 2pac stick will run with an ASIC voltage of 0.732 volts at 325 MHz. When the ambient temp goes above 26°C I start to have issues.

The idea is to run each ASIC at the minimum voltage at which the stick doesn't fail (and report ZOMBIE) or generate hardware errors. The lower the voltage, the less heat you have to shed with the heatsink. However you can only go so low with the voltage before the stick will begin to experience hardware errors.

I have attached an enormous heatsink to my 2pac, which lets me run at very high voltage without the chip getting too hot, but right now I'm experimenting with just how low a voltage I can run and still achieve stable operation.



Good. From the current your stick is consuming and what sidehack has said, I think you can run your ASIC chips at a lower voltage. But before you try this you need to get comfortable with measuring each ASIC chip's supply voltage with your multimeter. Your 2pac stick is being supplied with about 5 volts and 2 amps, but each ASIC chip on the PCB is supplied with a lower voltage. In your position I'd make it my job to get comfortable with measuring and recording the individual ASIC voltage. From there you'll be able to twiddle the voltage down to reduce the power consumption, and reduce temperatures without sacrificing stability.


It wouldn't hurt, but you can just write down the ambient temp when you notice a failure.



I hope I've explained this above. If I have been clear enough, let me know and I'll take some pictures that show measuring the ASIC voltage. The 2Pac stick needs a 5 volt supply, but it has a really nice voltage regulator that supplies each ASIC chip with a lower voltage.


I use stick-on NTC thermistors, but I hear good things about the banggood and HobbyKing IR point and shoot temp guns. Between 15 and 20 bucks ought to do it. https://hobbyking.com/en_us/catalogsearch/result/?erp_category=&q=thermometer

How're you measuring the current consumption of each stick?
Can you measure the voltage across the capacitor for each ASIC chip?
Finally, do you have a fan on the heatsinks?

In the time I've been playing with a 2Pac, the only zombie events I've had were when the cooling wasn't sufficient. Trying for minimum heatsink fan speed mine ran happily all day but then zombied about 3 hours after the building's aircon shut down for the evening. Running exactly the same setup with a higher fan speed and it's just run happily overnight @ 325 MHz with ASIC voltage of 0.73V. I feel it's worth monitoring both heatsink temp and ASIC temp on the back side of the PCB to get a feel for how effective your heat transfer is. Playing with tightness of the heatsink screws will let you find the optimum clamp force for heat transfer. Too  tight and the PCB flexes and transfer is less good.

I'm not an expert, but it appears so. Searching for "cgminer" and the name of any coin in that list I've searched for so far has returned results and examples, and if you can mine it with vh's build of cgminer then you can mine it with a Gekkoscience 2Pac, simply by changing the pool URL in your cgminer command line.

The 2Pac will mine sha256 coins. The ASIC (the chip that does the hashing work) is specifically built for performing sha256 hashes. It will not do other work. xhomerx10 posted a link to a page listing the types of coins the 2Pac can mine: https://www.coinwarz.com/miningprofitability/sha-256

The USB commands menu in cgminer will let you list all devices (press the L key) and hitting S and entering the ID (zero if you have only one stick) will show the device settings and stats. At present the temperature is always reported as 0, and a casual visual inspection doesn't show anythign that appears to be a thermistor. Unless the BT chip itself reports temperature, an external thermometer is your best best, because either the hardware doesn't report temperature, or cgminer as yet doesn't have instructions for querying the temperature sensor.

A few observations and a question from a new 2Pac owner. Will try to limit eye-rolling numbskullery in my first post. I've read perhaps 60% of this thread and will be embarrassed if I repeat an answered question.

I have a single 2Pac bought off ebay because I wanted to play with a single unit. Mining on slushpool for giggles, and have been running at various speeds and difficulties for a few days. Network latency has a noticeable effect on scoring hashrate for me, and the best I've managed to date it a ~5 hour sustained period of scoring hashrate of ~26 GH/s.

It's nice hardware, and I hope sidehack's proud.

Run overnight with minimum difficult 512 sustained ~20.5 GH/s scoring hashrate from my current high-latency network location. @375 MHz.

Bumping up to 425 MHz and difficulty 128 it's heading back up to ~24 GH/s At any clockspeed It's difficult sustaining WU/m > 320, but 375-425 MHz gets the highest scoring hashrate on this stick. (One half hour chunk of time @ 29 MH/s scoring hashrate. Think it'd be higher still if slushpool accepted difficulty < 128 and I was running in the low latency network.)

Running a USB hub with a hacked in buck converter I've limited to to 3 amps, and attached to a dirty great copper heatsink from an old Xeon server, I'm sitting at 35°C @ 425 MHz.

I have quite a few heatsinks from retired servers. The current style would fit 6 or 8 2Pacs and still be well under the TDP capacity of the heatsink. I have a small machining shop and will probably muck about with back side heatsink clamps to increase evenness and clamp force. Leftover watercooling and phase change will probably come into play at some point too, not because it's cost effective, but because it's amusing.

My particular stick has a 0.258 V difference between the two ASICs. N0 always has the lower voltage across the cap. Inrush current can be an issue when starting cgminer, and I'm contemplating some PCB mods for grossly oversized low ESR tantalum caps to see if this helps.

Questions for the delightful curmudgeon, or anyone else who's had at these with a soldering iron:

1) Is the voltage difference between the N0 and N1 ASIC something I can remedy with careful installation of a precision multiturn trimpot in parallel (or series, to avoid R=0 smokeletting) with an existing resistor? I appreciate the risk inherent in this.

2) I'm in Australia and local purchasing options are borked. I'd like to buy 10-20 2pac sticks, whatever quantity is worth sidehack's time. Never mind, found the email address in the [FOR SALE] post.
If I successfully purchase 20 I'll probably shift up to 10 units to desirous AU residents at slightly above total cost to me. Because I think it's funny and am curious about the imminent pod miner, I'll drop any small profit back in sidehack's development fund.

I can probably offer a few ex-server heatsinks drilled and tapped for 2pac to the local market as well, but I'm in this for fun and education. Oh, and recycling.


EDIT: Had it running at 500 MHz for a couple of hours with 0 HW err, but hash rate topped at 21 GH/s. 325 MHz seems to be the most productive speed @ 0.731V  36 GH/s.