They didn't all die at once, rather one at a time over a period of months. A few of them were already hashing at the much reduced rate of 30TH rather than their nominal 45TH, so they were already running at relatively tame current levels ... but a current overload due to brownout can't be discounted. I'll open up and do a close inspection of a few of those PSUs; I'm sure the hashing boards themselves are all fine, so I'm suspecting either the PSUs or controller boards at this point.

208V, which is on the low end. The current draw into the PSU would be higher than if I had access to a higher voltage (higher voltages, hence lower currents, are generally more desirable). Having all the units go down in what appears to be the same way, over a narrow period of time, suggests to me the PSUs may have been driven too hard. I haven't taken one of the PSUs apart yet, but I'll do that just to see if there is anything to see. Now that I don't have any more working M20's, I don't have a trusted PSU I could swap in to see if the miner works.

I had two of them apart and didn't notice anything obvious like charring or PCB/component discolouration. They were both quite dust-free, as their environment was clean. The fans were not seized; no rattling or loose components. I reseated all of the connectors I could find in case something vibrated loose or had some oxidization.

Yeah, I couldn't figure that drop from 45T to 30T out; the tuning process seemed to have completed successfully because there is a flag in the web interface that indicates when tuning completes. I let a few units cool down to ambient for 24h before powering them up again, thinking they might be running too hot; but my observation was once they took that initial drop from 45T to 30T they would never tune up to beyond 30T from a 'cold' start again; the 'factory reset' button didn't change that behaviour.

They were in an air conditioned datacenter since first arriving; ambient there is kept at 25-26C. Nothing different from when first powered up at 45T, to some weeks/months later when they would only power up to 30T. And now they won't power up at all. I can see a green light on the controller board itself, but none of the externally visible LEDs turn on, nor do the fans spin up.

I may have a bad batch; it happens. I've seen the usual posts here of some people having various Pangolin/MicroBT hardware issues, along with Bitmain hardware issues, etc. The usual expected issues. Nothing particularly out of the ordinary. Current mining hardware quality is generally known to be 'not great'. But I'm mostly disappointed by the lack of vendor support in this case. Butterfly Labs, Spondoolies, even Bitmain support was leagues beyond the silence from Pangolin. Not something I would have expected.

I'll just mention that I put 6 Pangolin M20 units into service this January 2020, rated 45TH/s each. They all ran at that speed for many weeks, then, one by one over the next few months, most had dropped to 30-35TH/s and would not go faster.

Finally, over the period May 28 to July 23, all 6 units eventually died. They no longer boot up. The last one died a few days ago. Early on I sent emails to PangolinMiner requesting technical support/service but still haven't heard anything after at least a month. I had a couple dozen Bitmain S9's before this, but they had their own hardware issues over time so I gave Pangolin a try. Unfortunately the experience has been even worse.

I recently tried Coinsquare's 'advanced trading' UI but find it very confusing. Let's say I want to place a limit order to SELL 0.2 BTC for $25,000 CAD/BTC.

On most exchanges, this is done simply by entering the amount (0.2 BTC) and the limit price ($25,000/BTC). Pretty simple and intuitive.

But on Coinsquare, my understanding is that this is actually entered as a *buy* order for CAD where you enter 0.000008 (0.2 / 25,000) in the 'Price (BTC)' field, and you enter 5000 (0.2 * 25,000) into the 'Quantity (CAD)' field. Essentially Coinsquare treats 'selling BTC' as 'buying CAD'. Is this correct? If so, it seems ridiculously convoluted.

I agree; S9's in general have been very problematic compared to the scores of S1's, S3's, S5's, and S7's I've purchased. A significant drop in quality overall, which is disappointing and frustrating to see. I've had to RMA a large percentage of S9 boards already, some of the replacements have failed as well, and still have out-of-warranty dead boards to deal with. I'm not purchasing additional S9's, even though I have capacity for them ... hoping that something more reliable will show up at some point.

This is interesting. I tried changing the password on that one miner but it always says 'invalid password' and doesn't accept the change. If I try to change the password to my other S9 miners it works fine. Very strange.

I have an interesting problem here with a Batch 3 S9:

1. I have set it for a static IP of 192.168.1.60
2. I have set up 3 mining pools
3. It operates as expected for a few hours, maybe up to a day.

Then, seemingly on its own, it changes its static IP address to something different (like 192.168.1.203) and changes just the first mining pool to f2pool or  peercoin pool (not my pools or credentials). Pools 2 and 3 remain untouched. I change the settings back to 192.168.1.60, and reset the first pool, but several hours later its back to some random IP address and the first pool has been changed again to something.

Has anyone else experienced something similar? I've had this happen 3 or 4 times now over the past few days. I keep changing the IP and pool information back, but in less then a day that miner is back on some random IP address and mining at some random pool. I've reflashed the firmware but that didn't help. I've reset the S9 to its factory settings by keeping the RESET button pressed for several seconds, then re-configuring, but that didn't seem to help either.

I don't know what could be causing this, or how to stop these changes from happening. I have several other S9 units on the same network and none of them exhibit this behavior.

I agree with this. Batch 8-15 are the least efficient variants of the S7 and require more power than any of the previous batches. I use EVGA 1300G2 on batches 2 and 4 (1160W stock). Batch 8 and above require another ~140W at stock which is beyond what most 1300G2's can reliably sustain due to the inherent power variance. Sure, you might get lucky but that is the exception. Use 2x 850W or a single 1600W supply with these.

The reality is that an S7 is expected to find a block every 4.7 years under the current difficulty. Think about that for a second. 4.7 years.

So sadly, no, your S7 is unlikely to hit a block here (or anywhere). And this is a good illustration of why most miners, especially the smaller ones, prefer pools to solo; a small solo miner can hash for months or years without getting any payout at all.

Wow, didn't notice that before ... thanks, goxed.

No worries, just a dead fan. I'll likely just source something locally.

Anyone have an email address for them?

The 1300G2 will work fine with an S7, but you have to manage the current flow in the PCIe connectors. The 1300G2 has 6 PCIe connectors. Use 5x single PCIe cables, and 1x double PCIe cable. Connect 2 cables to each blade of the S7 (the S7 does not need all 3 blade connectors to be populated; it runs fine with just 2 connectors per blade). One output of the double connector goes into an S7 blade; the other output goes into the S7 controller. This works because the controller requires very little power. So you end up with 6 cables from the 1300G2 going into 7 S7 connectors.

Each of the 3 S7 blades uses about 400 watts (~1200W total for an S7). Splitting that between 2 PCIe cables means 200W per cable. This is well within spec for the 16-gauge wires that the 1300G2 comes with. Note that each S7 blade has 3 PCIe connectors, but only 2 are required.

That's unfortunate, though 10 is all you need for an S7 (they actually work fine with only 7). My most recent order for these came about 3 weeks ago and those units still had 12 PCIe connectors.

Not sure what you're asking. This page: https://bitmaintech.com/productDetail.htm?pid=000201505040743496917U7kGsCm0694 pretty much explains it all, and has photos of all the PSU connectors. The AC input is IEC320-C13 (no cable is included; you have to provide your own); the 12V output is 12 PCI-E 6-pin cables.

Bitmain's price on these will (have to) drop significantly over the next few months; both the USD price and the BTC price. There're only 8 months left before the block reward halving and the value of an S7 continuously drops as the weeks pass. We already know that pcfli is ordering S7's for $1150USD for January 20 delivery which is a good clue where to expect the near-term public price to head.

With the recent run up on BTC exchange, the current Batch 8 pricing is unattractive, compounded by it being one of the less energy efficient S7 batches.

I haven't run any S3 units for a while now for that very reason, but I don't think anyone will argue that the more recent generation of Ants (S5, S5+, S7) have become less 'friendly' to home mining if only for the sole reason that they are LOUD. The S3 is whisper quiet compared to the later generations.

The S3/S3+ is still profitable depending on your local electricity costs, and even more so in the northern hemisphere during the winter months because it offsets heating costs. An S3+ can be found on eBay for under $150USD which still makes it attractive enough in certain conditions. Certainly an S5 will be more efficient, but efficiency isn't the only criteria for everyone.

S3's are cheap, they are quiet, and they will remain profitable during the winter in many regions of the world.

I'm not defending the S5+ or the S7 ... but it is worth keeping in mind that most folks who ordered the S5+ already have them mining, 6 weeks (aka 3 difficulty adjustments) before most anyone will be mining with an S7.

At the time the S5+ buying window was open, bitcoin was higher than it is now; a 7.7TH S5+ cost 9 BTC, which is quite a bit more cost effective than the 4.8TH S7 for 8 BTC. Unless your local power rates are outrageous, an S5+ mining today will break even sooner than an S7 starting mid-October will.