Bitmain AntMiner U1 Tips & Tricks

[klondike_bar]

Mine are running on this:
http://www.ebay.co.uk/itm/10-Port-USB-2-0-Hi-Speed-Multi-Hub-Expansion-with-Power-Adaptor-for-PC-Laptop-/390551625126

With this 5V8A power brick:
http://www.ebay.co.uk/itm/5V-5A-6A-7A-8A-AC-DC-Switching-Adapter-Desktop-Power-Supply-5-5mmx2-5mm-5-5x2-5-/310714289368?ssPageName=ADME:X:RTQ:GB:1123

And they look like this:


It is quite stable!

I have the same hub - 3.5A runs 5 antminers at 1.8GH each.
just added a 6th and am testing with 2.0GH - seems like it wont work though as only 3 detect and the rest throw port detection errors

[philipma1957]

Has anyone tried uprating the PSU for the Anker 9 port hub?  I was thinking of trying this one:
http://www.ebay.com/itm/AC-DC-Adapter-For-Model-JY071260U-Desktop-12V-10A-10000mA-Charger-Power-Supply-/400581749376?_trksid=p2054897.l4275

This way I could 9 U1's and still have the 'power' port available for a fan. Do you think the internal wiring of the hub will be Ok with 2x the draw (the stock PSU is 5000mA)

WOuld make for a nice tight rig, and would allow me to place this by an A/C that is always on and possibly over clock.

 too much if you have a 12 volt 4 amp or a 12 volt 5 amp power brick just go up to a 12 volt 6 amp hub.

This one is okay:

http://www.amazon.com/Adapter-Power-Supply-LCD-Monitor/dp/B003TUMDWG/ref=sr_1_1?ie=UTF8&qid=1390447824&sr=8-1&keywords=12+volt++6+amp

I have up graded a few 10 port hubs using it.  please be sure the oem hub is a 12 volt not a 5 volt.  plugging a 12 volt into a 5 volt is death to the hub.

[Walking Glitch]

I just looked at the user manual again. For anyone curious, you should be able to bring these past 1.2V, however, I would not recommend it without removing the solder mask and pasting these to a good heatsink with some good thermal paste. Once you close the R2 circuit with a 10K resister, the voltage goes up by .1V for every 1.25 KOhms of resistance you add to R1. 1.3V should definitely be doable with a 6.25K resistor on R1, which is as far as I would be willing to take mine.


EDIT: If you just put a 5k resistor in R2, that will give you 1.2V according to this:


EDIT EDIT: I found my magnifying glass. These resistors are encoded according to the 1% tolerance types on this page: http://www.marsport.org.uk/smd/res.htm (which makes sense... most experienced people I would expect to know this, but not hobbyists or tinkerers.)

[miniasic]

Hi,

I found a simple method to increase the voltage without soldering. On the image you can see several measurement points for different voltages on the antminer board. The USB voltage can be found between the outer pins of the the connector (red = 5V and black = Ground). Then we have two points near the USB - Serial converter which give us about 3.3V. Next you can find 1.1V close to U5 what seems to be the communication voltage of the ASIC. The relevant voltage for operation is the green point near U2. There are 0.8V (or in my case 0.86V in factory mode, what coud explain the differences in overclockability of the devices). Now to the interesting part. There is no need to replace both R1 and R2 just to follow the table in the U1 manual. The table there is more or less the same as in the AOZ1021AI datasheet. What I did was to use a medium hard pencil and do 3-4 strokes between the R2 connectors, plugged the device in an measured between black and green and voila I got 1.06V. You could easily do just one stroke for the start and see what happens with operation voltage. lesser strokes should give a higher resistance what would lower the voltage, so no risk in trying this. Unfotunately I "ruined" my device before this test (what made me bold enough to try it this way) so I can't tell you what hash rate I coud make after this modification. My antminer doesn't reach more than 450MHs as I can't clock it higher than 57.81MHz (0x1207). Somehow I messed up the PLL in earlier tests with small steps in clock rate and never made the device recover 

https://i.imgur.com/iFizDR3.png

Greetings,

              miniasic.

[Walking Glitch]

Hi,

I found a simple method to increase the voltage without soldering. On the image you can see several measurement points for different voltages on the antminer board. The USB voltage can be found between the outer pins of the the connector (red = 5V and black = Ground). Then we have two points near the USB - Serial converter which give us about 3.3V. Next you can find 1.1V close to U5 what seems to be the communication voltage of the ASIC. The relevant voltage for operation is the green point near U2. There are 0.8V (or in my case 0.86V in factory mode, what coud explain the differences in overclockability of the devices). Now to the interesting part. There is no need to replace both R1 and R2 just to follow the table in the U1 manual. The table there is more or less the same as in the AOZ1021AI datasheet. What I did was to use a medium hard pencil and do 3-4 strokes between the R2 connectors, plugged the device in an measured between black and green and voila I got 1.06V. You could easily do just one stroke for the start and see what happens with operation voltage. lesser strokes should give a higher resistance what would lower the voltage, so no risk in trying this. Unfotunately I "ruined" my device before this test (what made me bold enough to try it this way) so I can't tell you what hash rate I coud make after this modification. My antminer doesn't reach more than 450MHs as I can't clock it higher than 57.81MHz (0x1207). Somehow I messed up the PLL in earlier tests with small steps in clock rate and never made the device recover  



Greetings,

              miniasic.

Makes sense, as you are closing the circuit. The only problem I forsee is that without knowing the resistance of the pencil mark, you will not know the voltage. In your case, it worked out to be around 10.3KOhms.


EDIT: Well this is fabulous. Battery died on my multimeter right as I went to test how much resistance a pencil stroke has. None of my three alarm clocks have a 9V battery in them, and I can't find a 9V DC power supply. I'm not gonna wire up 6 AAs in series either...

EDIT EDIT: Now the doubt sets in that it was really the battery, because this meter has been wonky sometimes. Off to look for another multimeter to test the battery with...

Meh, not walking up to the shop when it's 10 below at 4 am, be back later today...

[chromosoma]

Hi,
By  drawing   graphite line in R2  you  just shorted the R1 to ground. (graphite has pretty low resistance)

See datasheet here:
http://www.aosmd.com/pdfs/datasheet/aoz1021ai.pdf
The AOZ1021A can provide huge  amount of current up to 3A, and still maintain 0.8V Output.
So by shorting R2  huge current  flowes through  R1 (1K on default).
Most probably you fried R1 (because i doubt it could  survive high current) and FeedBack input is open, thus frequency adjustment doesn't work.

By measuring the resistance with you multimeter  you again connected   output of DC/DC converter to ground, and  high current was forced into your multimeter. If it was cheap one, it may killed something in it, or maybe it was only  a fuse:)

Try to replace R1 and remove  graphite magic:)
Maybe this few cents for new resistor will save your Antminer.
If it helps, don't forget to donate;)

[kabopar]

Hi,
By  drawing   graphite line in R2  you  just shorted the R1 to ground. (graphite has pretty low resistance)

See datasheet here:
http://www.aosmd.com/pdfs/datasheet/aoz1021ai.pdf
The AOZ1021A can provide huge  amount of current up to 3A, and still maintain 0.8V Output.
So by shorting R2  huge current  flowes through  R1 (1K on default).
Most probably you fried R1 (because i doubt it could  survive high current) and FeedBack input is open, thus frequency adjustment doesn't work.

By measuring the resistance with you multimeter  you again connected   output of DC/DC converter to ground, and  high current was forced into your multimeter. If it was cheap one, it may killed something in it, or maybe it was only  a fuse:)

Try to replace R1 and remove  graphite magic:)
Maybe this few cents for new resistor will save your Antminer.
If it helps, don't forget to donate;)

If R2 has some lines of graphite drawn in parallel, this is unlikely to totally 'short' it, but just to reduce its resistance.  As Vo=0.8*(1+R1/R2) and if R2 is reduced, then Vo will increase.  Ofcourse one should not connect an ohmmeter into a live circuit, but measuring the voltage should typically cause no harm, assuming that the voltmeter has sufficiently high input impedance to avoid loading the circuit under test.  I have not tried this, but it reminds me of the 'pencil mod' on the Blue/Red Fury devices.

Cheers

[chromosoma]

Hi,
By  drawing   graphite line in R2  you  just shorted the R1 to ground. (graphite has pretty low resistance)

See datasheet here:
http://www.aosmd.com/pdfs/datasheet/aoz1021ai.pdf
The AOZ1021A can provide huge  amount of current up to 3A, and still maintain 0.8V Output.
So by shorting R2  huge current  flowes through  R1 (1K on default).
Most probably you fried R1 (because i doubt it could  survive high current) and FeedBack input is open, thus frequency adjustment doesn't work.

By measuring the resistance with you multimeter  you again connected   output of DC/DC converter to ground, and  high current was forced into your multimeter. If it was cheap one, it may killed something in it, or maybe it was only  a fuse:)

Try to replace R1 and remove  graphite magic:)
Maybe this few cents for new resistor will save your Antminer.
If it helps, don't forget to donate;)

If R2 has some lines of graphite drawn in parallel, this is unlikely to totally 'short' it, but just to reduce its resistance.  As Vo=0.8*(1+R1/R2) and if R2 is reduced, then Vo will increase.  Ofcourse one should not connect an ohmmeter into a live circuit, but measuring the voltage should typically cause no harm, assuming that the voltmeter has sufficiently high input impedance to avoid loading the circuit under test.  I have not tried this, but it reminds me of the 'pencil mod' on the Blue/Red Fury devices.

Cheers

On page 7 in the datasheet it says you should choose high resistors to avoid current loss through the devider. I Think graphite gived less then kohm of resistance,which leads to massive current from the output.

[miniasic]

Hehe,

good luck with your battery. Due to my understanding of the VR I got about 3kOhm for R2 with my pencile strokes.

0.8 * ( 1 + 1/3 ) = 1.066V

Doing lesser strokes should result in a higher resistance what gives us a lower voltage. Measuring the resistor in circuit doesn't result in the correct value due to other components (L/C/AOZ) also being connected during measurement. If I measure R1 I get about 0.82k and 0.87k for my pencile strokes. Tested with my reference pencil on a paper I get about 10kOhm for one stroke of 2mm decreasing by about 1kOhm per additional stroke. Graphite has about 8 Ohm per mm²/m. Compared to copper this is rather bad.

The resistor measuring was done while the device was off/not connected to power. And my antminer was slow before I tried the pencil magic  It's not totally wrecked and is after the magic still as slow as before. But yes, your warning is correct. Shortening R1 would increase the Voltage and current flow throuh voltage divisor R1/R2.

Greetings,

              miniasic.

[Jay_Pal]

Just one (perhaps "stupid") question, but...
If R2=5k, Vo = 0.8(*1+1/5) = 0.96V.
Would it be "safe" to put this resistor?
Or is there some strong reason to use the 10k and change R1 to 1.24k?

[miniasic]

Hi,

you would send 160mA through the voltage divider -> 0.15W of heat dissipation are produced at the resistors.
A SMD 603 is rated at about 0.1W. So the 5kOhm are out of specs. Using 10 and 1.24 gives you only 0.085A and 0.07W in the 10k resistor what is inside specs. So my pencil stroke makes the 1k resistor to be massive out of specs 

Greetings,

              miniasic.

[Jay_Pal]

So my pencil stroke makes the 1k resistor to be massive out of specs 

Hi!!!

Glad I've asked, then!

[miniasic]

Hi,

just in case someone needs a finer speed setting:

Frequency    | Hashrate  | value | timeout
125.00 MHz | 1000 MHs | 0481 | 042 ms
137.50 MHz | 1100 MHs | 0501 | 039 ms
150.00 MHz | 1200 MHs | 0581 | 035 ms
162.50 MHz | 1300 MHs | 0601 | 033 ms
175.00 MHz | 1400 MHs | 0681 | 030 ms
187.50 MHz | 1500 MHs | 0701 | 028 ms
200.00 MHz | 1600 MHs | 0781 | 026 ms
212.50 MHz | 1700 MHs | 0801 | 025 ms
225.00 MHz | 1800 MHs | 0881 | 023 ms
237.50 MHz | 1900 MHs | 0901 | 022 ms
250.00 MHz | 2000 MHs | 0981 | 021 ms
262.50 MHz | 2100 MHs | 0A01 | 020 ms
275.00 MHz | 2200 MHs | 0A81 | 019 ms

or even finer for tweaking your individual miner:
237.50 MHz | 1900 MHs | 1285 | 022 ms
243.75 MHz | 1950 MHs | 1305 | 022 ms
250.00 MHz | 2000 MHs | 1385 | 021 ms
256.25 MHz | 2050 MHs | 1405 | 020 ms
262.50 MHz | 2100 MHs | 1485 | 020 ms
268.75 MHz | 2150 MHs | 1505 | 019 ms
275.00 MHz | 2200 MHs | 1585 | 019 ms
281.25 MHz | 2250 MHs | 1605 | 019 ms
287.50 MHz | 2300 MHs | 1685 | 018 ms
293.75 MHz | 2350 MHs | 1705 | 018 ms
300.00 MHz | 2400 MHs | 1785 | 017 ms

But beware. While playing around with this values or better some values not respecting the forumular given in the U1 manual I somehow wrecked my antminers clock generation.

Greetings,

                miniasic.

[miniasic]

Hi again,

I now removed graphite magic again. R1 now gives me 1k and the non connected R2 is at 1.02k if measured in circuit. Voltage dropped to 0.83V again. Miner is still hashing to slow but hashing! Did anyone figure out what all the components on the board are?

I have:

Y1 25 MHz Oszillator (thats the 25 in the freqency formular)
U1 CP2102 USB to UART Bridge
U2 BM1380 ASIC
U3 503B or 5038 (hard to read) WTF?
U4 AOZ1021A Voltage Regulator
U5 5B402? WTF?

I'm interested in U3 and U5 because I still don't understand where the device stores it's calibration data but there seems to be something like that...and the serial protocoll the BM1380 speaks would be interesting, too.

Greetings,

              miniasic.

[eroxors]

Just one (perhaps "stupid") question, but...
If R2=5k, Vo = 0.8(*1+1/5) = 0.96V.
Would it be "safe" to put this resistor?
Or is there some strong reason to use the 10k and change R1 to 1.24k?

I painstakingly soldered 2.5k to all of my R2s which should result in 1.12, but in reality, they behave exactly the same as before. I'm going to have to put the 10k on all of the R1s now.

[chromosoma]

To the guy with slow miners:
Did you try to start the miner with the lowest possible freq, while powered alone in hub/ separate usb port, and then increase freq slowly, like 10 min for each value?

[miniasic]

Hi,

yes, tried this. At the moment I'm on 550MHs again but due to freq. it should be 600MHs. I had to do this game several times but after the last "failure" I couldn't recover the miner. It was hashing at 2GHs and started to decrease hashrate over the day. At around 1.4GHs I restarted it and since then it never got over 600MHs again. I have to do very small frequency steps and around 100 Accepts before increasing frequency otherwise bfgminder or the bitmaint cgminer report 0 or something starting with c5xxxxxx after the getfreq command. If this happens the miner is f**** up again and I have to start with a lower freq until it recovers.

Thanks for your reply,

                 miniasic.

[Janezki]

Hi,

just in case someone needs a finer speed setting:

Frequency    | Hashrate  | value | timeout
125.00 MHz | 1000 MHs | 0481 | 042 ms
137.50 MHz | 1100 MHs | 0501 | 039 ms
150.00 MHz | 1200 MHs | 0581 | 035 ms
162.50 MHz | 1300 MHs | 0601 | 033 ms
175.00 MHz | 1400 MHs | 0681 | 030 ms
187.50 MHz | 1500 MHs | 0701 | 028 ms
200.00 MHz | 1600 MHs | 0781 | 026 ms
212.50 MHz | 1700 MHs | 0801 | 025 ms
225.00 MHz | 1800 MHs | 0881 | 023 ms
237.50 MHz | 1900 MHs | 0901 | 022 ms
250.00 MHz | 2000 MHs | 0981 | 021 ms
262.50 MHz | 2100 MHs | 0A01 | 020 ms
275.00 MHz | 2200 MHs | 0A81 | 019 ms

or even finer for tweaking your individual miner:
237.50 MHz | 1900 MHs | 1285 | 022 ms
243.75 MHz | 1950 MHs | 1305 | 022 ms
250.00 MHz | 2000 MHs | 1385 | 021 ms
256.25 MHz | 2050 MHs | 1405 | 020 ms
262.50 MHz | 2100 MHs | 1485 | 020 ms
268.75 MHz | 2150 MHs | 1505 | 019 ms
275.00 MHz | 2200 MHs | 1585 | 019 ms
281.25 MHz | 2250 MHs | 1605 | 019 ms
287.50 MHz | 2300 MHs | 1685 | 018 ms
293.75 MHz | 2350 MHs | 1705 | 018 ms
300.00 MHz | 2400 MHs | 1785 | 017 ms

But beware. While playing around with this values or better some values not respecting the forumular given in the U1 manual I somehow wrecked my antminers clock generation.

Greetings,

                miniasic.

That's interesting - what does the timeout mean?
Is there any difference for example using 0981 or 1385 for 2GHz?
Can you give values for higher hashrates? I'm currently running mine steady @ 2.6 (4c81) with HW=0, just wanting to see if I could push it just a bit more. 
(2.8 (4d81) is too much, the miner doesn't start)

[BRADLEYPLOOF]

Can you give values for higher hashrates? I'm currently running mine steady @ 2.6 (4c81) with HW=0, just wanting to see if I could push it just a bit more. 
(2.8 (4d81) is too much, the miner doesn't start)

What sort of black magic did you cast on these things to get it up that high?  I can barely run one of them at x0981 (2 GHS) with 0 HW errors and the other at x0881 because if I bump the lower one up, I get mad HW errors, and if I bump the other one up to x0A81, I get about a shit ton of hardware errors... 

[Jay_Pal]

Can you give values for higher hashrates? I'm currently running mine steady @ 2.6 (4c81) with HW=0, just wanting to see if I could push it just a bit more. 
(2.8 (4d81) is too much, the miner doesn't start)

Wow!!! 4c81? Totally curious!!!
Anything soldered?
Where did you found those values?

What sort of black magic did you cast on these things to get it up that high?  I can barely run one of them at x0981 (2 GHS) with 0 HW errors and the other at x0881 because if I bump the lower one up, I get mad HW errors, and if I bump the other one up to x0A81, I get about a shit ton of hardware errors... 

I am using 0A01 and this is the result I get:

Code:

 cgminer version 3.9.0 - Started: [2014-01-23 13:06:54]
--------------------------------------------------------------------------------
 (5s):10.47G (avg):9.593Gh/s | A:31970  R:10  HW:211  WU:133.1/m
 ST: 2  SS: 0  NB: 28  LW: 44331  GF: 0  RF: 0
 Connected to mmpool.bitparking.com diff 10 with stratum as user 
 Block: 1ef58c7f...  Diff:1.79G  Started: [16:36:40]  Best share: 12.1K
--------------------------------------------------------------------------------
 [P]ool management [S]ettings [D]isplay options [Q]uit
 AMU 0:                | 447.6M/447.4Mh/s | A: 1430 R: 0 HW:52 WU: 6.3/m
 AMU 1:                | 456.4M/459.0Mh/s | A: 1470 R: 0 HW:66 WU: 5.9/m
 BAJ 0:  max 33C 0.99V | 4.628G/4.685Gh/s | A:16160 R:10 HW: 1 WU:65.4/m
 ANT 0:                | 3.054G/2.012Gh/s | A: 6600 R: 0 HW:40 WU:27.9/m
 ANT 1:                | 2.778G/1.993Gh/s | A: 6310 R: 0 HW:52 WU:27.6/m
--------------------------------------------------------------------------------

I'd guess this is somehow, stable, althought it isn't the anounced value!!
And I get somewhat the same value with 0981.