ANTMINER U3 Discussion and Support Thread

[Mikestang]

So here's another mystery to add to the U3 story.  I've been running 1 U3 for several weeks now and it has been behaving "normally".  Today I turned my S3+ back on and, for no other apparent reason, the U3 now runs at a 13%-15% error rate.  If I turn off the S3+ the errors drop back down to "normal" (for this unit normal HW error is ~ 3%).  They are both powered by the same PSU (800W gold rated, plenty of juice for them both), but that is the only common thread.  Isn't that peculiar?  Anyone else had a similar experience?

[notlist3d]

So here's another mystery to add to the U3 story.  I've been running 1 U3 for several weeks now and it has been behaving "normally".  Today I turned my S3+ back on and, for no other apparent reason, the U3 now runs at a 13%-15% error rate.  If I turn off the S3+ the errors drop back down to "normal" (for this unit normal HW error is ~ 3%).  They are both powered by the same PSU (800W gold rated, plenty of juice for them both), but that is the only common thread.  Isn't that peculiar?  Anyone else had a similar experience?

My life has been amazing on U3's last day and a half.  I bought a R1 from crazy guy and he did a custom firmware.   I have just stock  power supplies that come with them, and 3 U3's.

I normally at least at morning or night had one to reset, possibly some during day.  But with his custom firmware I had not had to touch them for almost a day and a half.  He has been able to fix zombies through his custom firmware it seems like.  Still testing it out by very very promising results.

If I get a few more day's of this I will be suggesting everyone to get one from him with his fimware.

[Jake36]

So here's another mystery to add to the U3 story.  I've been running 1 U3 for several weeks now and it has been behaving "normally".  Today I turned my S3+ back on and, for no other apparent reason, the U3 now runs at a 13%-15% error rate.  If I turn off the S3+ the errors drop back down to "normal" (for this unit normal HW error is ~ 3%).  They are both powered by the same PSU (800W gold rated, plenty of juice for them both), but that is the only common thread.  Isn't that peculiar?  Anyone else had a similar experience?

Kind of similar, have 2 U3's (had some more, but they died) running along with a S3+ on a 1000W EVGA. Was getting zombies like normal with both U3's on same pcie, put each U3 on it's own pcie and haven't had a zombie after more than 48 hours (YET). They have reset from au3 0 and au3 1 to au3 5 and au3 6, but there still going without me unplugging and replugging. Also for 48 hrs I've got 531HW total right now, where I would be expecting ~ 3-4 times that many in 48 hrs.

So I'm wondering if bitman either: A) way under estimated the power requirements for the U3's or B) used some really cheap power component, that it takes the power of a single pcie per U3 to over come the power loss of the component and stop zombies? Second thought, both A and B combined.

Which (A) would probably explain a lot of the reports of the supplied power bricks melting (along with the bricks being cheap).

[notlist3d]

So here's another mystery to add to the U3 story.  I've been running 1 U3 for several weeks now and it has been behaving "normally".  Today I turned my S3+ back on and, for no other apparent reason, the U3 now runs at a 13%-15% error rate.  If I turn off the S3+ the errors drop back down to "normal" (for this unit normal HW error is ~ 3%).  They are both powered by the same PSU (800W gold rated, plenty of juice for them both), but that is the only common thread.  Isn't that peculiar?  Anyone else had a similar experience?

Kind of similar, have 2 U3's (had some more, but they died) running along with a S3+ on a 1000W EVGA. Was getting zombies like normal with both U3's on same pcie, put each U3 on it's own pcie and haven't had a zombie after more than 48 hours (YET). They have reset from au3 0 and au3 1 to au3 5 and au3 6, but there still going without me unplugging and replugging. Also for 48 hrs I've got 531HW total right now, where I would be expecting ~ 3-4 times that many in 48 hrs.

So I'm wondering if bitman either: A) way under estimated the power requirements for the U3's or B) used some really cheap power component, that it takes the power of a single pcie per U3 to over come the power loss of the component and stop zombies? Second thought, both A and B combined.

Which (A) would probably explain a lot of the reports of the supplied power bricks melting (along with the bricks being cheap).

I'm telling you if you use a R1 with crazy guy's custom firmware it's like night/day compared to normal runnning.  He added some code that do many things in trying to make a zombie come back.  And it appears to be working.

I did use 3 on a Windows 8 machine and had to fix zombies a decent amount.  So far after custom firmware I have not had to touch them for 2 day's. I will keep testing but I have been blown away by it.

[Mikestang]

Kind of similar, have 2 U3's (had some more, but they died) running along with a S3+ on a 1000W EVGA. Was getting zombies like normal with both U3's on same pcie, put each U3 on it's own pcie and haven't had a zombie after more than 48 hours (YET). They have reset from au3 0 and au3 1 to au3 5 and au3 6, but there still going without me unplugging and replugging. Also for 48 hrs I've got 531HW total right now, where I would be expecting ~ 3-4 times that many in 48 hrs.

One thing I found that made a big difference to my U3 (and I specify "my" because they all do different things) was when I switched it from an instance of cgminer shared with a few other usb sticks to its own cgminer instance.  That change alone made 50% of my zombies go away, and it only very, very rarely changes from AU3 0 to AU3 1 (only happened 3 times in last month).

As soon as I get paid I'm ordering an R1 from crazy guy.  Discounting the wireless and other functions of the R1, I look at that device like a 5GH stability upgrade to the U3. 

[notlist3d]

Kind of similar, have 2 U3's (had some more, but they died) running along with a S3+ on a 1000W EVGA. Was getting zombies like normal with both U3's on same pcie, put each U3 on it's own pcie and haven't had a zombie after more than 48 hours (YET). They have reset from au3 0 and au3 1 to au3 5 and au3 6, but there still going without me unplugging and replugging. Also for 48 hrs I've got 531HW total right now, where I would be expecting ~ 3-4 times that many in 48 hrs.

One thing I found that made a big difference to my U3 (and I specify "my" because they all do different things) was when I switched it from an instance of cgminer shared with a few other usb sticks to its own cgminer instance.  That change alone made 50% of my zombies go away, and it only very, very rarely changes from AU3 0 to AU3 1 (only happened 3 times in last month).

As soon as I get paid I'm ordering an R1 from crazy guy.  Discounting the wireless and other functions of the R1, I look at that device like a 5GH stability upgrade to the U3. 

Since I upgraded I have not touched those 3 U3's.   His firmware has been able to fix any problem without any interaction from me.  Which is awesome. 

So far 2 day's of not messing with a single zombie unplug/replug .. I could get use to this  

[-ck]

Hypothesis. Sounds like the problem then is power supply noise. Presumably each device adds noise back upstream to the PSU which gets carried to the other devices. In which case a high quality linear power supply or generous filtering on the power source going to each device should make them more reliable.

[Mikestang]

Hypothesis. Sounds like the problem then is power supply noise. Presumably each device adds noise back upstream to the PSU which gets carried to the other devices. In which case a high quality linear power supply or generous filtering on the power source going to each device should make them more reliable.

That makes a lot of sense.  I sent one of my U3s to a forum member several months ago, he was a college student and into hardware and was studying different inexpensive miners.  I wished that I had saved what he wrote me (any way to recover deleted PMs?), but in his diagnostics he found a flaw/shortcut with something to do with the power stuff (excuse my lack of hardware techno-speak) and was pretty sure that was why these things have so many issues.

[Jake36]

Hypothesis. Sounds like the problem then is power supply noise. Presumably each device adds noise back upstream to the PSU which gets carried to the other devices. In which case a high quality linear power supply or generous filtering on the power source going to each device should make them more reliable.

Very good possibility.

I wonder if that would explain why in video .anto. posted, the PSU led was flashing in rhythm with the squeak and the miner led?  

[zOU]

I have 3 U3.

All of them have a solid red LED beside the USB port.

2 of them have a solid red LED beside the flashing red LED (the one indicating hashing)
the 3rd one didn't have this LED on for several month although it was hashing correctly.

And now (about 1h ago) it has decided to turn itself ON.

Note that this particular U3 had never suffered from the zombie syndrom... let's hope this doesn't start

[notlist3d]

I have 3 U3.

All of them have a solid red LED beside the USB port.

2 of them have a solid red LED beside the flashing red LED (the one indicating hashing)
the 3rd one didn't have this LED on for several month although it was hashing correctly.

And now (about 1h ago) it has decided to turn itself ON.

Note that this particular U3 had never suffered from the zombie syndrom... let's hope this doesn't start

That sounds like a zombie.  Sometimes they recover via software but not always when I was using windows.

I am on day 3 without touching U3's with the R1 and crazy guy's firmware.  It's working great.

[zOU]

That sounds like a zombie.  Sometimes they recover via software but not always when I was using windows.

I am on day 3 without touching U3's with the R1 and crazy guy's firmware.  It's working great.

I know when they're zombie because the front led goes from ON to OFF.
This one never had this LED on.
Never ever.

And the blinking LED (quite hard to see as it flashes rapidely and is dim) was always ok.

As I said, this one has never been a zombie in several month despite having the "status" LED off.

[spazzdla]

This zombie issue killed the U3... :S...

[notlist3d]

This zombie issue killed the U3... :S...

With crazy guy's custom firmware I have not touched my U3's in 3 days.  It seems he realy did fix them via his coding on R1 custom firmware.  Which is pretty awsome.

I have checked pool side and it's had a constant of all 3 of my U3's mining.   If you own a U3 its honestly worth getting a R1 from him to not unplug/replug multiple times a day.

[Meuh6879]

Just a capture from P2Pool + CGMiner (and Zadig driver) + Antminer U3 (no tuning).

To connect CGminer to P2Pool, you need to write this in the command line :

Code:

-o stratum+tcp://localhost:9332

[zOU]

U3 mining wirelessly ?

[JaredKaragen]

Hypothesis. Sounds like the problem then is power supply noise. Presumably each device adds noise back upstream to the PSU which gets carried to the other devices. In which case a high quality linear power supply or generous filtering on the power source going to each device should make them more reliable.

I agree.

I switched my U3 from a 2.5A power supply (one it shipped with) with a 5.0A power supply and my results were a little baffling.


With my U3 on the 2.5A supply it shipped with my settings were:
--bmsc-voltage 0830  --bmsc-freq 0982
Hashrate:  58-67Gh/s

Switching to the 5.0A with the same settings netted me 43.0Gh/s at most at any time.
With the new power supply I tried countless combinations and found this to be the best:
--bmsc-voltage 0805  --bmsc-freq 1306
Hashrate: 53-60Gh/s

I am 100% certain the power supply to the antminer is the direct cause of the issue.  Its a 5.0A that came with a USB hub I bought;  and I dont use the hub to feed juice, so the amperage is pretty much wasted there.  But;  the power supply is noisy as hell on a logic sense.   When I finally get enough room cleared here to bring home my oscilliscope;  I'd be able to do more of an in-depth analysis.

Nothing changed the presence of HW errors.   They happen and they don't happen.  These I would throw a big high bet at the power control/regulation circuitry.   I am assuming the ripple in my lower amperage supply is the most "harmonically stable" compared to my other for my unit running like it does.   

[.anto.]

Hypothesis. Sounds like the problem then is power supply noise. Presumably each device adds noise back upstream to the PSU which gets carried to the other devices. In which case a high quality linear power supply or generous filtering on the power source going to each device should make them more reliable.

I agree.

I switched my U3 from a 2.5A power supply (one it shipped with) with a 5.0A power supply and my results were a little baffling.


With my U3 on the 2.5A supply it shipped with my settings were:
--bmsc-voltage 0830  --bmsc-freq 0982
Hashrate:  58-67Gh/s

Switching to the 5.0A with the same settings netted me 43.0Gh/s at most at any time.
With the new power supply I tried countless combinations and found this to be the best:
--bmsc-voltage 0805  --bmsc-freq 1306
Hashrate: 53-60Gh/s

I am 100% certain the power supply to the antminer is the direct cause of the issue.  Its a 5.0A that came with a USB hub I bought;  and I dont use the hub to feed juice, so the amperage is pretty much wasted there.  But;  the power supply is noisy as hell on a logic sense.   When I finally get enough room cleared here to bring home my oscilliscope;  I'd be able to do more of an in-depth analysis.

Nothing changed the presence of HW errors.   They happen and they don't happen.  These I would throw a big high bet at the power control/regulation circuitry.   I am assuming the ripple in my lower amperage supply is the most "harmonically stable" compared to my other for my unit running like it does.  

For 250 MHz (0x0982), I believe you need above 0.83V and that needs at least 6A power supply. I even use 12V/12.5A power supply and running at 200 MHz/0.83V, and my 2 Antminer U3' are still very unstable or dead/zombie within 20 hours of mining everyday. So it might really depend on your luck.

[Mikestang]

I tried countless combinations and found this to be the best:
--bmsc-voltage 0805  --bmsc-freq 1306
Hashrate: 53-60Gh/s
  

If you are still using --bmsc in your .bat file then you are using the old Bitmain cgminer fork.  You should update to the latest 4.9.2 release, it had some changes made just for the U3.  You can d/l it from https://github.com/ckolivas/cgminer, check readme.asic for update U3 commands, --au3-volt and --au3-freq

The only settings that my U3 will run with <5% HW error is 785volt and 237.5freq, averages like 42GH/s.

[notlist3d]

I tried countless combinations and found this to be the best:
--bmsc-voltage 0805  --bmsc-freq 1306
Hashrate: 53-60Gh/s
  

If you are still using --bmsc in your .bat file then you are using the old Bitmain cgminer fork.  You should update to the latest 4.9.2 release, it had some changes made just for the U3.  You can d/l it from https://github.com/ckolivas/cgminer, check readme.asic for update U3 commands, --au3-volt and --au3-freq

The only settings that my U3 will run with <5% HW error is 785volt and 237.5freq, averages like 42GH/s.

I'm still using a R1 with crazy guy's custom firmware.   It really is like night/day compared to when I was running it on window's dealing with zombies.

I go day's first time was 5 day's without any zombies, and before I had to fix multiple times a day.  Now I reset after that 5 day's and are seeing how long I go again.  It's a lot nicer not waking up to fixing zombies and fixing before bed every night was my ritual with 3 U3's.