ANTMINER S2 Discussion and Support Thread

[MrTeal]

Well, mine won't connect at 192.168.1.99, and according to an IP scan there's nothing from 1.1 to 2.254
My laptop detects the board and the lights on the BBB work, so it doesn't appear to be a hardware error.
Did anyone else get a unit with a different IP?

"My laptop detects the board"

What do you mean by that?

Check all internal connections.  I had loose boards, the cable to the display was disconnected, and the internal ethernet cable was not completely seated to the panel jack.

I mean that instead of getting a cable unplugged message like you do when the unit is on, it shows it's connected to an unknown network.
Strangely, sometimes if I power cycle the unit it will connect, and sometimes no lights on the BBB's MagJack come on and nothing is detected.

[ak49er]

Dumb UPS. Last night routed my package from Alaska to Kentucky and I am in California. Delayed 1 extra day. Was ready to un-box today.

I live in Alaska and they arrive in Anchorage from China in the afternoon, then get sent to California, and back to Anchorage overnight, to be delivered to me the next day.  I just received notification that there's been a flight delay and my latest S1 is in Anchorage and will be delayed.  And I'm thinking why not just send it down here then and have it arrive tomorrow as expected, can't you read the address?

[cxboyminer]

They...look...awesome!!!

[MrTeal]

I think overclocking is going to be extremely limited, with four TPS53355 per board, or one for 16 chips. On the S1, it's one for 8 chips.

Thats the detail I've been waiting for   *really hoping for a PCB photo still*

The S1 could consume around 45W per group of 8 chips overclocked (overclocking pushed the regulator right to its limits, but the real limiting factor was the voltage being set to 1.1V - some people have apparently pencil-modded the S1 to almost 215GH/~470W, or about 60w/8 chips)  
45W/1.1V=40.9A on a 30A-rated component. (bitfury h-boards faced very similar amperage limits when overvolted around 0.95V)


Its likely these chips run at around 0.8V and 1000W/640chips is about 25W per cluster of 16.  That leaves a LOT of headroom on the regulators, but some of this is lost to the difference in voltage (more amperage).  25W/0.8V= 31.25A stock.

 Presumably these should be able to overclock without any issues from the regulator. I imagine all other components are beefy enough also. I imagine you will hit hardware errors at 15-20% overclock as with the S1. These may even be capable of up to 1.5TH is voltage was increased to ~0.9V

0.75V unloaded, so a little higher than that on the current. Still some room and I'd expect people should be able to get 20-40% overclocks, but not the 100+% ones we might have been able to do with 8 chips per vrm.

[1l1l11ll1l]

Aaaaaaaaand here you go.


The Power supply keeps cutting out in 1 machine, I didn't have a spare PSU, so I removed 2 of the asic blades to bring down the power draw, we'll see if it stays alive tonight.

[Biodom]

Wouldn't you be limited by the internal 1000W PSU? Replacing internal PSU with 1250w or 1350W might work for over-clock, but at $300-350 expense and loss of warranty, i suppose.

I wonder, why no more batches for both S1 and S2?

[cxboyminer]

Just saying to all the batch 2 customers...

SE Asia is having some weird weather, such as hailing and heavy thunderstorm for the last couple of days continuously. Your orders could be delayed.

[jjiimm_64]

Wouldn't you be limited by the internal 1000W PSU? Replacing internal PSU with 1250w or 1350W might work for over-clock, but at $300-350 expense and loss of warranty, i suppose.

I wonder, why no more batches for both S1 and S2?

CRAP  i wanted to by a shitload of S1's if the price came down just a little.

[thomas_s]

Aaaaaaaaand here you go.


The Power supply keeps cutting out in 1 machine, I didn't have a spare PSU, so I removed 2 of the asic blades to bring down the power draw, we'll see if it stays alive tonight.

Open the S2 and replace the PSU jumper, by the way minersource.net sells jumpers

http://minersource.net/products/psu-jump-starter

[jamesg]

I wonder, why no more batches for both S1 and S2?

Crazy manic miners bought them all... 

[cxboyminer]

Aaaaaaaaand here you go.


The Power supply keeps cutting out in 1 machine, I didn't have a spare PSU, so I removed 2 of the asic blades to bring down the power draw, we'll see if it stays alive tonight.

Open the S2 and replace the PSU jumper, by the way minersource.net sells jumpers

http://minersource.net/products/psu-jump-starter

You can use a paperclip for that purpose too.

[klondike_bar]

I think overclocking is going to be extremely limited, with four TPS53355 per board, or one for 16 chips. On the S1, it's one for 8 chips.

Thats the detail I've been waiting for   *really hoping for a PCB photo still*

The S1 could consume around 45W per group of 8 chips overclocked (overclocking pushed the regulator right to its limits, but the real limiting factor was the voltage being set to 1.1V - some people have apparently pencil-modded the S1 to almost 215GH/~470W, or about 60w/8 chips)  
45W/1.1V=40.9A on a 30A-rated component. (bitfury h-boards faced very similar amperage limits when overvolted around 0.95V)


Its likely these chips run at around 0.8V and 1000W/640chips is about 25W per cluster of 16.  That leaves a LOT of headroom on the regulators, but some of this is lost to the difference in voltage (more amperage).  25W/0.8V= 31.25A stock.

 Presumably these should be able to overclock without any issues from the regulator. I imagine all other components are beefy enough also. I imagine you will hit hardware errors at 15-20% overclock as with the S1. These may even be capable of up to 1.5TH is voltage was increased to ~0.9V

0.75V unloaded, so a little higher than that on the current. Still some room and I'd expect people should be able to get 20-40% overclocks, but not the 100+% ones we might have been able to do with 8 chips per vrm.

thanks for the actual value    Still leaves a lot of hardware overhead, espescially looking at the PCB. improved cooling an an extra 600W PSU could allow some nice overclocking results. (how are the PCIe plugs wired - is it seperate 12V for each board or are they all sharing a common voltage?

[thomas_s]

Aaaaaaaaand here you go.


The Power supply keeps cutting out in 1 machine, I didn't have a spare PSU, so I removed 2 of the asic blades to bring down the power draw, we'll see if it stays alive tonight.

Open the S2 and replace the PSU jumper, by the way minersource.net sells jumpers

http://minersource.net/products/psu-jump-starter

You can use a paperclip for that purpose too.

You could, the one that was in the unit from Bitmain was using the paperclip trick...except it was bad when it was replaced with the actual jumper the machine powered on without dropping out or failing to turn on.

It really doesn't matter where you get it from, but I'd suggest replacing the one that came in it.

[ak49er]

Aaaaaaaaand here you go.


The Power supply keeps cutting out in 1 machine, I didn't have a spare PSU, so I removed 2 of the asic blades to bring down the power draw, we'll see if it stays alive tonight.

Open the S2 and replace the PSU jumper, by the way minersource.net sells jumpers

http://minersource.net/products/psu-jump-starter

I wouldn't have thought that it needs a jumper.  If I were building the thing I'd have that plug doing something useful like say powering things using 5V or 3.3V.  You know, like what it's intended for.

[cyberspaced]

I think overclocking is going to be extremely limited, with four TPS53355 per board, or one for 16 chips. On the S1, it's one for 8 chips.

Thats the detail I've been waiting for   *really hoping for a PCB photo still*

The S1 could consume around 45W per group of 8 chips overclocked (overclocking pushed the regulator right to its limits, but the real limiting factor was the voltage being set to 1.1V - some people have apparently pencil-modded the S1 to almost 215GH/~470W, or about 60w/8 chips)  
45W/1.1V=40.9A on a 30A-rated component. (bitfury h-boards faced very similar amperage limits when overvolted around 0.95V)


Its likely these chips run at around 0.8V and 1000W/640chips is about 25W per cluster of 16.  That leaves a LOT of headroom on the regulators, but some of this is lost to the difference in voltage (more amperage).  25W/0.8V= 31.25A stock.

 Presumably these should be able to overclock without any issues from the regulator. I imagine all other components are beefy enough also. I imagine you will hit hardware errors at 15-20% overclock as with the S1. These may even be capable of up to 1.5TH is voltage was increased to ~0.9V

0.75V unloaded, so a little higher than that on the current. Still some room and I'd expect people should be able to get 20-40% overclocks, but not the 100+% ones we might have been able to do with 8 chips per vrm.

I'm quite sure those puny heatsinks will not be enough when overclocked, 200Mhz+ will fry the chips unless some hefty cooling is applied!
If you want to overclock the BM chip, the S1 was made just for that!!

[klondike_bar]

I think overclocking is going to be extremely limited, with four TPS53355 per board, or one for 16 chips. On the S1, it's one for 8 chips.

Thats the detail I've been waiting for   *really hoping for a PCB photo still*

The S1 could consume around 45W per group of 8 chips overclocked (overclocking pushed the regulator right to its limits, but the real limiting factor was the voltage being set to 1.1V - some people have apparently pencil-modded the S1 to almost 215GH/~470W, or about 60w/8 chips)  
45W/1.1V=40.9A on a 30A-rated component. (bitfury h-boards faced very similar amperage limits when overvolted around 0.95V)


Its likely these chips run at around 0.8V and 1000W/640chips is about 25W per cluster of 16.  That leaves a LOT of headroom on the regulators, but some of this is lost to the difference in voltage (more amperage).  25W/0.8V= 31.25A stock.

 Presumably these should be able to overclock without any issues from the regulator. I imagine all other components are beefy enough also. I imagine you will hit hardware errors at 15-20% overclock as with the S1. These may even be capable of up to 1.5TH is voltage was increased to ~0.9V

0.75V unloaded, so a little higher than that on the current. Still some room and I'd expect people should be able to get 20-40% overclocks, but not the 100+% ones we might have been able to do with 8 chips per vrm.

I'm quite sure those puny heatsinks will not be enough when overclocked, 200Mhz+ will fry the chips unless some hefty cooling is applied!
If you want to overclock the BM chip, the S1 was made just for that!!

thats a losing attitude - anything can overclock!

[MrTeal]

Well, mine won't connect at 192.168.1.99, and according to an IP scan there's nothing from 1.1 to 2.254
My laptop detects the board and the lights on the BBB work, so it doesn't appear to be a hardware error.
Did anyone else get a unit with a different IP?

If you have a linux device on the network when it boots:
sudo tcpdump -n -nn -l -i any arp

I don't have a portable one, but I do have wireshark.

All I get is a bunch of requests from my IP for the gateway.

[cyberspaced]

I think overclocking is going to be extremely limited, with four TPS53355 per board, or one for 16 chips. On the S1, it's one for 8 chips.

Thats the detail I've been waiting for   *really hoping for a PCB photo still*

The S1 could consume around 45W per group of 8 chips overclocked (overclocking pushed the regulator right to its limits, but the real limiting factor was the voltage being set to 1.1V - some people have apparently pencil-modded the S1 to almost 215GH/~470W, or about 60w/8 chips)  
45W/1.1V=40.9A on a 30A-rated component. (bitfury h-boards faced very similar amperage limits when overvolted around 0.95V)


Its likely these chips run at around 0.8V and 1000W/640chips is about 25W per cluster of 16.  That leaves a LOT of headroom on the regulators, but some of this is lost to the difference in voltage (more amperage).  25W/0.8V= 31.25A stock.

 Presumably these should be able to overclock without any issues from the regulator. I imagine all other components are beefy enough also. I imagine you will hit hardware errors at 15-20% overclock as with the S1. These may even be capable of up to 1.5TH is voltage was increased to ~0.9V

0.75V unloaded, so a little higher than that on the current. Still some room and I'd expect people should be able to get 20-40% overclocks, but not the 100+% ones we might have been able to do with 8 chips per vrm.

I'm quite sure those puny heatsinks will not be enough when overclocked, 200Mhz+ will fry the chips unless some hefty cooling is applied!
If you want to overclock the BM chip, the S1 was made just for that!!

thats a losing attitude - anything can overclock!

If it's running at 57 degrees @193MHz there's not much headroom!!

[buckrogers]

what did you guys do who received the wrong plug? just use a different standard one in its place?

The one I got is for the UK i think and I a in the US.

thanks!

[palawan]

Well, mine won't connect at 192.168.1.99, and according to an IP scan there's nothing from 1.1 to 2.254
My laptop detects the board and the lights on the BBB work, so it doesn't appear to be a hardware error.
Did anyone else get a unit with a different IP?

If you have a linux device on the network when it boots:
sudo tcpdump -n -nn -l -i any arp

I don't have a portable one, but I do have wireshark.

All I get is a bunch of requests from my IP for the gateway.

Is there a "reset to factory" button like the one from S1?  If it's a raspberry pi, I'd take out the SD card and try to mount it on my laptop, copy some files from it to backup.  I'd also want to back it up (as it is from factory) to another SD card.  I don't know what options are available...

Sounds like a much more flexible device than the S1...