GekkoScience has a new stickminer that does 300+GH

[kano]

You should also know that when running older Compacs along with the 'F', those older ones should be ran under a separate instance of cgminer because of radical differences in performance between them and the 'F'. Can still all be on one hub, just need the different instance of cgminer for all of them to be happy.

... and in case it wasn't obvious there's options (documented in the README) to select the different USB miners when you run cgminer:

--gekko-compacf-detect

--gekko-r606-detect

--gekko-newpac-detect
--gekko-2pac-detect
--gekko-compac-detect
--gekko-terminus-detect

So in each cgminer you run you can add one or more of the above to only mine on those USB miners

[1714138194]

1714138194

[1714138194]

1714138194

[CochnocherCrypto]

Hey everyone! Quick question that maybe someone can help with. I’ve been running a handful of Compac F’s for a few months - love them and how easy it was to get them up and running having used Newpacs in the past. My issue is this - I bought another one as I had 11 and one open high powered port on my second gekko hub. I got it yesterday - powered down the PSU that powers the 2 hubs (750W HP PSU with a Parallel Miner X11 breakout board), put the new stick in (didn’t touch the potentiometer - have never really had to running them at 400mhz…), powered up the PSU and started CGMiner on my Raspberry Pi. The new stick (serial number ends in 2807) which I got from ASICPuppy (I mention this because I’ve bought 4 other CompacF’s from him so I know he’s not selling bad hardware and in my conversations with him he seems knowledgeable). The stick will not run with any stability at any frequency. CGMiner wound up tuning it down to 160MHZ and it would’ve continued to drop if I didn’t stop it to investigate. I put the stick on my USB Multimeter to compare the current draw against another stick and it doesn’t seem too too far off. I figured it may be an issue with power so I tried powering it alone in one of my gekko hubs - same thing… when CGMiner is starting up it shows the current clock frequency (200MHZ), the target frequency (T) as 400, then (P) which normally ramps up quickly on my other sticks to the target I’ve set seems to get stuck at 154 and after several minutes the stick just throttles itself. It has yet to achieve more than 55 GH/S. Is there any sort of fundamental difference in the Compc 8.2’s vs the 8.1’s that would require
I do anything differently? Current hardware and software I’ve tried to troubleshoot this with is as follows:

- Gekkoscience USB Hubs (tried the stick in several ports on both hubs)
- RaspberryPi 4 Model B (tried using both the 32bit and 64bit versions of Raspberry Pi OS and tried running it on a Pi with Ubuntu)
- software I’m using is CGMiner 4.12.0
- tried running it in my old Sipolar hub (each port has a maximum power output of 2.4a according to the specs)
- tried notching the core voltage potentiometer in very small increments which didn’t change anything)

Has anyone else had any experiences like this? If yes - any ideas as far as what the next steps should be? I reached out to ASICPuppy but have yet to hear back from him and in the event I do t hear from him today, would like to try to figure this out if there’s anything I may be doing incorrectly.

Thanks!

[n0nce]

Has anyone else had any experiences like this?

In my experience, every chip needs a bit more or less voltage; that's due to the 'silicon lottery' effect.

Since you have USB multimeters: can you post how many watts / amps your 'working' sticks are pulling at a given clock speed and how much the new one is pulling at the same clock speed?

- tried notching the core voltage potentiometer in very small increments which didn’t change anything)

How much did you increment it? How many watts did you reach? As long as it's under 3A, you still have some headroom. Also make sure that it's cooled properly, but I assume it is, since the other sticks seem to be getting enough cooling as well.

[sidehack]

You might also tweak the starting frequency. It could be having trouble at the bottom end and needs to be forced over the knee past where the chip stabilizes internally and starts operating close enough to expected parameters that it ramps up on its own. When we test at the factory, the start freq is 300 and target is 400. If we don't see at least 220GH sustained for 24 hours with those settings, it doesn't go out the door.

[CochnocherCrypto]

Hey guys - thanks so much for your input!

n0nce - I was only able to compare the initial current as it would start ramping up - which was the same as all of my other sticks. Trouble was I couldn't get it to achieve the clock speed I was setting so I wound up comparing it to another one of my sticks at 200mhz which was 4.96V / 1.671A (so approx. 8.25-8.3 watts).

Sidehack - Initially at the end of the cg miner command line I was typing I was adding "--gekko-mine2 --gekko-compacf-freq 400". I tried adding the starting frequency (300) but it made it display the 300mhz frequency, T 400 P 0 for about 25 - 45 seconds and then the word ZOMBIE would show up for a few seconds and it would reassign the next consecutive # before GSF (ie. if it was 7: GSF 10052807, it would then change to 12: GSF 10052807, and the loop would continue). Was chatting with a coworker earlier who is a home-mining enthusiast also and he suggested I take out the "--gekko-mine2" in the text I was inputting and *knock on wood* for the first time, it's working! Running steadily for the past 45 minutes. I hadn't thought to start it at a higher frequency - didn't realize it was possible until I was combing this thread at work today lol.

Really appreciate the suggestions - love my stickminers <3

Cheers!

[sidehack]

I'll be completely honest, I'm not really sure what kano's -mine2 algorithm does. I use an older (test build) version of his driver on the production testers because it defaults to spitting out some extra data I like to have for diagnosing.

Glad you got everything working.

[n0nce]

*knock on wood* for the first time, it's working! Running steadily for the past 45 minutes. I hadn't thought to start it at a higher frequency - didn't realize it was possible until I was combing this thread at work today lol.

That's great to hear! I just quickly looked up my config file and I have these settings, just for the record:

Code:

"gekko-compacf-freq"   : "500",
"gekko-compacf-detect" : true,
"gekko-start-freq"     : "400",
"gekko-mine2"          : true,
"gekko-tune2"          : "60",
"gekko-tune-up"        : "70"

So I actually have to start at 400MHz - but as I said, this is a pretty bitchy stick; I expect normal chips to work with a lower start frequency.

n0nce - I was only able to compare the initial current as it would start ramping up - which was the same as all of my other sticks. Trouble was I couldn't get it to achieve the clock speed I was setting so I wound up comparing it to another one of my sticks at 200mhz which was 4.96V / 1.671A (so approx. 8.25-8.3 watts).

Alright; I believe that's a normal number - at full speed, I'm around 2.8A but when ramping up from 200MHz it can definitely be a bit over half of that I guess. If there's instability when running full speed, keep an eye on the wattage - that's what helped me solve my issues.

[CochnocherCrypto]

Really appreciate you taking time to post some suggestions! That’s interesting - I currently have mine running at 425mhz and this morning when I was on my way out the door for work I took a quick peak - they’re all stable and have been hashing away for over 24 hours now!. I’ve been toying with the thought of running them higher but wasn’t sure what sort of temps are safe to sustain for extended periods of time. If you don’t mind me asking - what sort of cooling equipment are you using? I can’t figure out how to get any sort of thermal info from CGMiner so I brought my IR Thermometer home from work and just measured the back side of the PCB directly behind the chip. Looking at the pin out information for the BM1397 it does appear to have temperature monitoring capabilities.

[hispace]

Hello,

I am looking for the immersion in oil to refrigerate the compacf, my question is the following:

Would it be possible or if someone has tried it to put the whole hub with the Gekko's, inside the pool with the oil to avoid the noise of the fans and lower the temperature, would it work?

Thanks.

[n0nce]

Looking at the pin out information for the BM1397 it does appear to have temperature monitoring capabilities.

It does, but it's not implemented, so the chip neither throttles itself when getting too hot, nor does it report the temps to CGMiner, which is a real pity.
I personally use a 120mm PC fan, because that's what I had lying around, but I still plan to get some small fans to put directly onto the heatsinks. With this type of heatsink (relatively low surface area), I believe the added static pressure will be better for the sticks.

A little 3D printed adapter like this will be needed, though.

Would it be possible or if someone has tried it to put the whole hub with the Gekko's, inside the pool with the oil to avoid the noise of the fans and lower the temperature, would it work?

I would personally just turn the hub around and fix it in a way that it doesn't dip into the liquid, just to be on the safe side and since it's easy to do. Would look nicer if the whole thing was submerged, though, for sure.

[sidehack]

What the chip actually does is use some pins to interface to an external temperature sensor and then the comm protocol has some stuff built in to set up temperature reads via register access. Sensors aren't built into the chip, but it'd be awful handy if they were.

[bigdaddymccarron]

Hello,

I am looking for the immersion in oil to refrigerate the compacf, my question is the following:

Would it be possible or if someone has tried it to put the whole hub with the Gekko's, inside the pool with the oil to avoid the noise of the fans and lower the temperature, would it work?

Thanks.

I have mine water cooled, see this post.

[bigdaddymccarron]

sidehack,

Can/do you repair the USB hubs?

[sidehack]

Service questions should start with the reseller.

[CochnocherCrypto]

n0nce - so I actually used something VERY similar to that - I bought them from Bitcoinmerch. However, they ran way, way hotter using those as opposed to the setup pictured in the link (measured the rear of the PCB directly behind the chip).

I don't have a jr rank so you'll have to go to this site to see my current setup. https://imgur.com/a/OsYdU8M

Running my sticks with the BitcoinMerch "fan upgrade" I measured the rear of the PCB at about 53 Celsius (420mhz). With my current setup (this was prior to getting the industrial noctua fans - was using Arctic F12's) I measured the same spot at about 46 Celsius, which I would certainly consider substantial. I would, conservatively, assume the chip itself is probably 20% hotter than the rear of the PCB.

[nullama]

n0nce - so I actually used something VERY similar to that - I bought them from Bitcoinmerch. However, they ran way, way hotter using those as opposed to the setup pictured in the link (measured the rear of the PCB directly behind the chip).

I don't have a jr rank so you'll have to go to this site to see my current setup.





Running my sticks with the BitcoinMerch "fan upgrade" I measured the rear of the PCB at about 53 Celsius (420mhz). With my current setup (this was prior to getting the industrial noctua fans - was using Arctic F12's) I measured the same spot at about 46 Celsius, which I would certainly consider substantial. I would, conservatively, assume the chip itself is probably 20% hotter than the rear of the PCB.

Quoted for photo visibility.

Also wanted to ask, how are you measuring the temperature of the sticks?, I'm thinking about using a thermal camera to do it, but I'm not sure they're actually very accurate. At least the cheaper USB ones.

A drop from 53 to 46 degs is considerable. Those noctua fans are really great at keeping silent miners cool.

[CochnocherCrypto]

What the chip actually does is use some pins to interface to an external temperature sensor and then the comm protocol has some stuff built in to set up temperature reads via register access. Sensors aren't built into the chip, but it'd be awful handy if they were.

BUMMER! Looking at the CGMiner README I saw some examples in which a temperature was displayed with different miners. I was doing some research to see why I couldn't figure out how to get mine to display that information so this makes a lot of sense. I contemplated using the temperature probes similar to the ones that come with fan controllers and just putting adhering them to the heatsink right between the heatsink and the PCB - but none of the fan controllers I've looked at can monitor 12 temp probes. I know some ASICs are designed to run H O T for very long periods of time with little to no downtime - just can't find a ton of info on these chips in particular. Bitmain seems very secretive lmao.

n0nce - so I actually used something VERY similar to that - I bought them from Bitcoinmerch. However, they ran way, way hotter using those as opposed to the setup pictured in the link (measured the rear of the PCB directly behind the chip).

I don't have a jr rank so you'll have to go to this site to see my current setup.





Running my sticks with the BitcoinMerch "fan upgrade" I measured the rear of the PCB at about 53 Celsius (420mhz). With my current setup (this was prior to getting the industrial noctua fans - was using Arctic F12's) I measured the same spot at about 46 Celsius, which I would certainly consider substantial. I would, conservatively, assume the chip itself is probably 20% hotter than the rear of the PCB.

Quoted for photo visibility.

Also wanted to ask, how are you measuring the temperature of the sticks?, I'm thinking about using a thermal camera to do it, but I'm not sure they're actually very accurate. At least the cheaper USB ones.

A drop from 53 to 46 degs is considerable. Those noctua fans are really great at keeping silent miners cool.

I'm a service technician at a major Fiat-Chrysler franchise - brought my snap-on IR thermometer (very costly, but very accurate) home to take some measurements. And yes - as you can tell by how I outfitted my Pi, I'm a big fan of Noctua products lol.

[hispace]

n0nce: I would personally just turn the hub around and fix it in a way that it doesn't dip into the liquid, just to be on the safe side and since it's easy to do. Would look nicer if the whole thing was submerged, though, for sure. 

Thanks n0nce, I had my doubts but I think I'm going to try something simple to test. 


bigdaddymccarron: I have mine water cooled, see this post.

Thanks bigdaddymccarron, excellent work, I take ideas! 

Thanks guys

[sidehack]

I was looking at integrating a temp sensor, but we were already pushing two months behind schedule getting to market with this guy and I didn't expect they'd sell out as fast as they did, prompting "regular product" status with manufacture in the several thousands. It also didn't help that the parts were low stock everywhere, and I was already fighting stock issues on a couple other things. We're about to start manufacture on a new version again, since some parts we've found stock of for years are drying up overnight right now.

I'm not sure if BM1397 are built to run hot for a long time, but I do know the die is much larger than, say, the BM1387 we're used to, and the top plating of the package is copper rather than plastic or ceramic, all of which means it handles higher power draws much better, stays cool easier.

[n0nce]

n0nce - so I actually used something VERY similar to that - I bought them from Bitcoinmerch. However, they ran way, way hotter using those as opposed to the setup pictured in the link (measured the rear of the PCB directly behind the chip).

I don't have a jr rank so you'll have to go to this site to see my current setup. https://imgur.com/a/OsYdU8M

Running my sticks with the BitcoinMerch "fan upgrade" I measured the rear of the PCB at about 53 Celsius (420mhz). With my current setup (this was prior to getting the industrial noctua fans - was using Arctic F12's) I measured the same spot at about 46 Celsius, which I would certainly consider substantial. I would, conservatively, assume the chip itself is probably 20% hotter than the rear of the PCB.

That's very good to know, thank you so much! I just assumed fans 'directly' on the heatsinks would be better, but then I'll stick to my PC fan setup (similar to yours).
What RPM are you running them at?

What the chip actually does is use some pins to interface to an external temperature sensor and then the comm protocol has some stuff built in to set up temperature reads via register access. Sensors aren't built into the chip, but it'd be awful handy if they were.

BUMMER! Looking at the CGMiner README I saw some examples in which a temperature was displayed with different miners. I was doing some research to see why I couldn't figure out how to get mine to display that information so this makes a lot of sense. I contemplated using the temperature probes similar to the ones that come with fan controllers and just putting adhering them to the heatsink right between the heatsink and the PCB - but none of the fan controllers I've looked at can monitor 12 temp probes. I know some ASICs are designed to run H O T for very long periods of time with little to no downtime - just can't find a ton of info on these chips in particular. Bitmain seems very secretive lmao.

The examples probably stem from different ASICs, not Compac F. Even if you were to connect a probe to the correct pins, the firmware doesn't implement forwarding that information to CGMiner, so I don't think you can add this functionality after the fact.

Of course Bitmain is secretive. It's all only available under NDA, if at all.