Next step would be to remove the 4RL inductor and connect a DC-DC converter to the low voltage input of the miner chip.
In addition, you could change the oscillator to about 1.5 times current oscillator.
This would offload the heat generated by the power supply (no need for huge heatsink) and also disengage the usb hub (usb hub used only for data, not power)
I've done this to the block erupter and got it to go 55% faster, with no HW errors.
Full documentation:
http%3A%2F%2Fwww.idobartana.com%2Ffiles%2Fbitcoin%2520miners%2520overclocking%2520tutorial.pdf
Topic: Experiments in overclocking an Antminer U1 (Read 4648 times)
Interesting, I wonder if a S1 could be modified in a similar fashion?
8 )
8 )
Try to put one more fan on up side just to see if result will change,for me it works to 3,0 ghz
already on a powered hub with 5 other u1 ants, supplied off a 8A power supply 
all other ants are ok, its only 1 that has this issue, no matter if its the only ant in hub, it always starts ok ..... then goes slow (almost stopping/not hashing etc)
all other ants are ok, its only 1 that has this issue, no matter if its the only ant in hub, it always starts ok ..... then goes slow (almost stopping/not hashing etc)
You need a powered USB hub that will supply more than the standard 5V/500mA. Some hubs are powered but will limit the amount of power that will go to the device.
Thanks for the info.
i am having an issue with one of my U1's, it will not hash over the stock frequency (starts at the higher frequency, then drops down and stops hashing all together)
, so i am planning to do a resister upgrade with it
i am having an issue with one of my U1's, it will not hash over the stock frequency (starts at the higher frequency, then drops down and stops hashing all together)
, so i am planning to do a resister upgrade with it
I posted this originally on reddit.com, but I figured some people here would find it of use.
Part One:
I ordered 2 of the Antminer U1's to experiment with overclocking them after reading the user manual and datasheet which details the information needed to do so.
The stock hash rate is 1.6Gh/s and according to the manual, with the correct replacement of resistors that can be pushed to 2.8Gh/s.
Replacing the 1kOhm R1 resistor with a 3.74kOhm resistor, and populating R2 with a 10kOhm resistor configures the DC/DC converter IC to output 1.1V instead of the stock 0.8V.
The two Antminer U1's arrived and were slotted into a 7 port powered (5V/2A) hub for 24 hours to test. I ended up using bfgminer 3.10.0 on the Linux computer I've got setup to do scrypt mining. I just had to compile bfgminer, setup the pool login information, and then using the device manager specified 'antminer:/dev/ttyUSB0' to detect the first Antminer device, adding the second device was just a matter of specifying 'antminer:/dev/ttyUSB1' instead.
After the 24 hours were up, and the magic smoke hadn't escaped from the ASIC chips, I started to mess with the software overclocking command line options. To do this with bfgminer use the option "--set-device antminer:clock=x0981" and refer to the table in the manual for the various timing options. In this case the value x0981 sets the clock frequency to 250MHz which results in an ASIC hash rate of around 2.0Gh/s. After running with this clock speed for around 4 hours resulted in a 15% HW error rate on the second Antminer U1, but not the first. Bumping the clock speed up to 275MHz resulted in around a 30% HW error rate on both. Any option above 275MHz doesn't appear to change the hash rate, only increases the HW error rate.
Now it was time to strip off the heatsink from one of the Antminer's. I chose the unit that had been throwing HW errors at 250MHz and stripped off the heatsink.
Antminer U1 component side:
https://i.imgur.com/Mq54uj6.jpg
Antminer U1 heatsink side:
https://i.imgur.com/2hEd8l8.jpg
The gold square area on the back board side is where the ASIC gets rid of most of it's heat, so using a heatsink that was around I secured the Antminer using zip ties and the piece of thermal tape that was on the original heatsink.
Antminer U1 rigged to a CPU heatsink:
https://i.imgur.com/JBJ7QnM.jpg
The second time running the Antminer's at 250MHz for 4 hours resulted in just a single hardware error between both devices. As of writing this there're 3 hardware errors across both devices with each device operating at an average 2.06Gh/s. When I tried bumping the clock frequency up to 275MHz again there were still a significant number of HW errors, so I'm thinking maybe the ASIC isn't getting enough power. The next step will be to go about modifying the hardware, I've ordered the resistors from Mouser and they should be arriving soon. Part two coming soon.
Part Two:
Continuing in my attempt to overclock an Antminer U1 to 2.8Gh/s I acquired the replacement resistors, new heatsink, thermal compound, and some Chip Quik to make removing the old SMD resistors a simple task.
Heatsink: http://www.enzotechnology.com/cnb_s1.htm
Thermal Compound: http://tuniq.com.tw/accessories/tx2.html
3.74Kohm resistor: http://www.mouser.com/ProductDetail/Panasonic/ERA-3AEB3741V/?qs=%2fha2pyFaduiamcDwRD0o%252bWBgLh2ctiEOqyDEftyAmQ%252b3OHaB8qtsrQ%3d%3d
10Kohm resistor: http://www.mouser.com/ProductDetail/Panasonic/ERJ-3EKF1002V/?qs=%2fha2pyFadujWqMGonbcVXVyzBa1QZFMkth5YtDEtRaJfSdml3v9%2fbQ%3d%3d
The first step was to remove the 1Kohm resistor in the R1 location. After wiping the board down with an alcohol pad, adding flux, and a little dab of Chip Quik on one side of the resistor kept the solder molten long enough to heat up the other side and slide it off the board.
Antminer U1 with R1 removed:
https://i.imgur.com/TMswPdF.jpg
Next step was to clean off the old flux with another alcohol wipe, inspect the board to make sure I hadn't lifted the solder pad, apply more flux and solder in the 10Kohm resistor, followed by the 3.74Kohm resistor. I ordered a bunch of the 10Kohm resistor as they were really cheap, but the tolerance is 1%, so I went through a few resistors on the multimeter before finding a 10.01Kohm resistor. As there was already solder on the pads for R2 it was just a matter of using tweezers to hold the resistor in place while applying heat to one pad. After that it was trivial to just apply heat to the other end and get a good connection. The 3.74Kohm resistor is 0.1%, so I ended up settling on a 3.72Kohm out of the few I ordered. Adding the R1 resistor was as simple as the R2, though I had to add a bit more solder on the side that had been Chip Quik'd. Another clean and inspection resulted in the following.
Antminer U1 with replaced R1 and R2:
https://i.imgur.com/nws3EPW.jpg
The final step was to chop off the side mounts on the heatsink and drill two holes to mount the Antminer to it. A quick bit of metal cutting with a Dremel got rid of those, and also smoothed off the rough edges. I ended up having to position the Antminer at an angle on the heatsink to get it to fit nicely. A 2mm bit drilled the correct size holes to accommodate the two screws that held the old Antminer heatsink in place. I didn't have to tap threads into the drilled holes as the screws were strong enough to cut into the copper on their own. A small application of the thermal compound to the gold plated area on the back of the Antminer (refer to Part 1 images) and I attached the Antminer securely to the heatsink.
Top view of Antminer U1 attached to heatsink:
https://i.imgur.com/b87BWDy.jpg
Bottom view of Antminer U1 attached to heatsink:
https://i.imgur.com/UDztoHK.jpg
Bottom view of other side:
https://i.imgur.com/KXvt5EC.jpg
I apologize for the awful camera potato pictures, I'll attempt to get a proper camera and take nicer ones at some point in the near future.
Now for the moment of truth, after unplugging all the other USB miners I slotted the modified Antminer and started up bfgminer running the ASIC at stock frequency. After letting it run for about an hour without anything catching on fire or exploding it was time to up the frequency. I slowly upped the frequency to 350MHz to try reach 2.8Gh/s, though at 350MHz I ran into what I'm assuming are power supply issues and the hash rate dropped to 0.8Gh/s, but didn't throw any HW errors. At 300MHz it hashes away happily at 2.58Gh/s with no HW errors, so I'll leave it at that speed for 24 hours and see what happens. Thanks for reading, hope this helps other people overclock their Antminers.
Small update:
I did the same resistor replacement on an Antminer U1 with the stock heatsink. It hashed for about 40 seconds before getting too hot and throwing a lot of HW errors. Checking the voltage going to the ASIC is pretty simple, at the bottom edge of the ASIC right near the silkscreen marking "U2" there is a test point. This test point will indicate how much voltage is going to the ASIC, in my case it's 1.094V. Which would probably explain why I can't get to 2.8Gh/s, it would appear that you really need to hit the 10Kohm and 3.74Kohm resistor values without much deviation to get 1.1V.
If you find this guide useful and would like to send some BTC my way, here's my address: 1CT8yiqCRSKk4XHy7PvgYLTbejzyYagYZ7
Part One:
I ordered 2 of the Antminer U1's to experiment with overclocking them after reading the user manual and datasheet which details the information needed to do so.
The stock hash rate is 1.6Gh/s and according to the manual, with the correct replacement of resistors that can be pushed to 2.8Gh/s.
Replacing the 1kOhm R1 resistor with a 3.74kOhm resistor, and populating R2 with a 10kOhm resistor configures the DC/DC converter IC to output 1.1V instead of the stock 0.8V.
The two Antminer U1's arrived and were slotted into a 7 port powered (5V/2A) hub for 24 hours to test. I ended up using bfgminer 3.10.0 on the Linux computer I've got setup to do scrypt mining. I just had to compile bfgminer, setup the pool login information, and then using the device manager specified 'antminer:/dev/ttyUSB0' to detect the first Antminer device, adding the second device was just a matter of specifying 'antminer:/dev/ttyUSB1' instead.
After the 24 hours were up, and the magic smoke hadn't escaped from the ASIC chips, I started to mess with the software overclocking command line options. To do this with bfgminer use the option "--set-device antminer:clock=x0981" and refer to the table in the manual for the various timing options. In this case the value x0981 sets the clock frequency to 250MHz which results in an ASIC hash rate of around 2.0Gh/s. After running with this clock speed for around 4 hours resulted in a 15% HW error rate on the second Antminer U1, but not the first. Bumping the clock speed up to 275MHz resulted in around a 30% HW error rate on both. Any option above 275MHz doesn't appear to change the hash rate, only increases the HW error rate.
Now it was time to strip off the heatsink from one of the Antminer's. I chose the unit that had been throwing HW errors at 250MHz and stripped off the heatsink.
Antminer U1 component side:
https://i.imgur.com/Mq54uj6.jpg
Antminer U1 heatsink side:
https://i.imgur.com/2hEd8l8.jpg
The gold square area on the back board side is where the ASIC gets rid of most of it's heat, so using a heatsink that was around I secured the Antminer using zip ties and the piece of thermal tape that was on the original heatsink.
Antminer U1 rigged to a CPU heatsink:
https://i.imgur.com/JBJ7QnM.jpg
The second time running the Antminer's at 250MHz for 4 hours resulted in just a single hardware error between both devices. As of writing this there're 3 hardware errors across both devices with each device operating at an average 2.06Gh/s. When I tried bumping the clock frequency up to 275MHz again there were still a significant number of HW errors, so I'm thinking maybe the ASIC isn't getting enough power. The next step will be to go about modifying the hardware, I've ordered the resistors from Mouser and they should be arriving soon. Part two coming soon.
Part Two:
Continuing in my attempt to overclock an Antminer U1 to 2.8Gh/s I acquired the replacement resistors, new heatsink, thermal compound, and some Chip Quik to make removing the old SMD resistors a simple task.
Heatsink: http://www.enzotechnology.com/cnb_s1.htm
Thermal Compound: http://tuniq.com.tw/accessories/tx2.html
3.74Kohm resistor: http://www.mouser.com/ProductDetail/Panasonic/ERA-3AEB3741V/?qs=%2fha2pyFaduiamcDwRD0o%252bWBgLh2ctiEOqyDEftyAmQ%252b3OHaB8qtsrQ%3d%3d
10Kohm resistor: http://www.mouser.com/ProductDetail/Panasonic/ERJ-3EKF1002V/?qs=%2fha2pyFadujWqMGonbcVXVyzBa1QZFMkth5YtDEtRaJfSdml3v9%2fbQ%3d%3d
The first step was to remove the 1Kohm resistor in the R1 location. After wiping the board down with an alcohol pad, adding flux, and a little dab of Chip Quik on one side of the resistor kept the solder molten long enough to heat up the other side and slide it off the board.
Antminer U1 with R1 removed:
https://i.imgur.com/TMswPdF.jpg
Next step was to clean off the old flux with another alcohol wipe, inspect the board to make sure I hadn't lifted the solder pad, apply more flux and solder in the 10Kohm resistor, followed by the 3.74Kohm resistor. I ordered a bunch of the 10Kohm resistor as they were really cheap, but the tolerance is 1%, so I went through a few resistors on the multimeter before finding a 10.01Kohm resistor. As there was already solder on the pads for R2 it was just a matter of using tweezers to hold the resistor in place while applying heat to one pad. After that it was trivial to just apply heat to the other end and get a good connection. The 3.74Kohm resistor is 0.1%, so I ended up settling on a 3.72Kohm out of the few I ordered. Adding the R1 resistor was as simple as the R2, though I had to add a bit more solder on the side that had been Chip Quik'd. Another clean and inspection resulted in the following.
Antminer U1 with replaced R1 and R2:
https://i.imgur.com/nws3EPW.jpg
The final step was to chop off the side mounts on the heatsink and drill two holes to mount the Antminer to it. A quick bit of metal cutting with a Dremel got rid of those, and also smoothed off the rough edges. I ended up having to position the Antminer at an angle on the heatsink to get it to fit nicely. A 2mm bit drilled the correct size holes to accommodate the two screws that held the old Antminer heatsink in place. I didn't have to tap threads into the drilled holes as the screws were strong enough to cut into the copper on their own. A small application of the thermal compound to the gold plated area on the back of the Antminer (refer to Part 1 images) and I attached the Antminer securely to the heatsink.
Top view of Antminer U1 attached to heatsink:
https://i.imgur.com/b87BWDy.jpg
Bottom view of Antminer U1 attached to heatsink:
https://i.imgur.com/UDztoHK.jpg
Bottom view of other side:
https://i.imgur.com/KXvt5EC.jpg
I apologize for the awful camera potato pictures, I'll attempt to get a proper camera and take nicer ones at some point in the near future.
Now for the moment of truth, after unplugging all the other USB miners I slotted the modified Antminer and started up bfgminer running the ASIC at stock frequency. After letting it run for about an hour without anything catching on fire or exploding it was time to up the frequency. I slowly upped the frequency to 350MHz to try reach 2.8Gh/s, though at 350MHz I ran into what I'm assuming are power supply issues and the hash rate dropped to 0.8Gh/s, but didn't throw any HW errors. At 300MHz it hashes away happily at 2.58Gh/s with no HW errors, so I'll leave it at that speed for 24 hours and see what happens. Thanks for reading, hope this helps other people overclock their Antminers.
Small update:
I did the same resistor replacement on an Antminer U1 with the stock heatsink. It hashed for about 40 seconds before getting too hot and throwing a lot of HW errors. Checking the voltage going to the ASIC is pretty simple, at the bottom edge of the ASIC right near the silkscreen marking "U2" there is a test point. This test point will indicate how much voltage is going to the ASIC, in my case it's 1.094V. Which would probably explain why I can't get to 2.8Gh/s, it would appear that you really need to hit the 10Kohm and 3.74Kohm resistor values without much deviation to get 1.1V.
If you find this guide useful and would like to send some BTC my way, here's my address: 1CT8yiqCRSKk4XHy7PvgYLTbejzyYagYZ7
Jump to: