New Official AMT Thread

[opieum2]

This is Josh cell phone number:
<snip>

I know AMT has pissed alot of people off, but posting his cell is pretty poor form. Also opens you up to him suing you for cyberbullying and all that stuff. Just sayin. Better to not open a nasty door into this. I am sure there are some people with his cell number, but it also does not hurt to at least maintain SOME bit of civility. You only damage everyone's chances at getting anything by doing this. Not like his number was publicly listed for everyone to see somewhere else. How did you get it in the first place? Did you get his permission to post it? Just consider the consequences to yourself with that action.

[ISAWHIM]

Need help... Running into an annoying issue...

I have just a single test-module running...

However, the "Power" and "Reset" switches are not hooked-up to anything. Thus, I have to keep turning the PSU power on and off. This results in a hit-or-miss on the booting process. It likes to hang on the "Monitor" stage, after boot. Other-times, it just runs normally and connects, booting with the IP display.

Seems to hang on the IP portion... keeps getting reassigned a new IP each time. Doesn't like to keep using the old one over and over.

I assume the unit can be turned on and off, and actually reset, with the buttons. No?

Power-cycling the PSU like this is not healthy for continual use. It surges the hell out of the connected devices and the PSU itself.

So... Anyone have a photo of where the power and/or reset buttons should be plugged-in?

There are four unoccupied connections, 2x 4-pin (?power/reset?) and 2x 3-pin (Fans?). (On the daughter-board, attached to the Ras-Pi.)

P.S. This one board, runs cool at 218-220GHs... Started initially on AMT's test-pool. xD

P.P.S. NOTE TO AMT... Indicate the correct way to connect the backplane, on the backplane. (Realized I forgot to take note of that, prior to taking it apart. It fits both ways... One way is wrong. xD The correct way is with the ribbon-cable-link hanging over the heat-sink side, not the PCB side. Not sure what might have happened if I had it the wrong way, and powered it up. Also note, the floppy-power connection on the backplane has pins that stick-out dangerously close to the heat-sink. Should grind those down and place nylon protective tape over them.)

[ISAWHIM]

Ok, side note...

When solo-mining... if the CPU shows as 100% and the logs show "getworks"... then you ARE hashing away. However, since there are no "shares to submit", the program has nothing to calculate hash-rate with. Well, it does, but the program is not setup to display "work hash rate".

Looks like this, when solomining...
Go to the (LOGS) then (CGMINER) on the miners monitor page.

Code:

Difficulty Accepted=>0.0
Pool Rejected%=>0.0
Found Blocks=>0
Difficulty Rejected=>0.0
Device Rejected%=>0.0
Pool Stale%=>0.0
Work Utility=>0.0
Rejected=>0
Elapsed=>1532
Hardware Errors=>0
Accepted=>0
Network Blocks=>27
Local Work=>0
Get Failures=>0
Difficulty Stale=>0.0
Total MH=>0.0
Device Hardware%=>0.0
Discarded=>2027
Stale=>0
MHS av=>0.0
Getworks=>2097
MHS 5s=>0.0
Best Share=>0
Last getwork=>1393144660
Remote Failures=>0
Utility=>0.0

Sooo... If you solomine to your won wallet, it will show 0.0GHs for the speed. You just have to use the logs to see that it is actually hashing-away.

Still looking for a way to reboot and power-off, without having to flip the switch on my PSU.

[marto74]

Need help... Running into an annoying issue...

I have just a single test-module running...

However, the "Power" and "Reset" switches are not hooked-up to anything. Thus, I have to keep turning the PSU power on and off. This results in a hit-or-miss on the booting process. It likes to hang on the "Monitor" stage, after boot. Other-times, it just runs normally and connects, booting with the IP display.

Seems to hang on the IP portion... keeps getting reassigned a new IP each time. Doesn't like to keep using the old one over and over.

I assume the unit can be turned on and off, and actually reset, with the buttons. No?

Power-cycling the PSU like this is not healthy for continual use. It surges the hell out of the connected devices and the PSU itself.

So... Anyone have a photo of where the power and/or reset buttons should be plugged-in?

There are four unoccupied connections, 2x 4-pin (?power/reset?) and 2x 3-pin (Fans?). (On the daughter-board, attached to the Ras-Pi.)

P.S. This one board, runs cool at 218-220GHs... Started initially on AMT's test-pool. xD

P.P.S. NOTE TO AMT... Indicate the correct way to connect the backplane, on the backplane. (Realized I forgot to take note of that, prior to taking it apart. It fits both ways... One way is wrong. xD The correct way is with the ribbon-cable-link hanging over the heat-sink side, not the PCB side. Not sure what might have happened if I had it the wrong way, and powered it up. Also note, the floppy-power connection on the backplane has pins that stick-out dangerously close to the heat-sink. Should grind those down and place nylon protective tape over them.)

This is with RPi i think ?
@ least this way it was when i look over bitmine.ch design.
The problem with Rpi is that it shares common break pint in the CPU between USB and Ethernet
actually the Ethernet is emulated over usb port of the Pi.
That is the main reason we are working since the beginning with support for TP link routers.
Our way have it's downside too, lack of memory for fancy graphics and stuff

[ISAWHIM]

Need help... Running into an annoying issue...

I have just a single test-module running...

However, the "Power" and "Reset" switches are not hooked-up to anything. Thus, I have to keep turning the PSU power on and off. This results in a hit-or-miss on the booting process. It likes to hang on the "Monitor" stage, after boot. Other-times, it just runs normally and connects, booting with the IP display.

Seems to hang on the IP portion... keeps getting reassigned a new IP each time. Doesn't like to keep using the old one over and over.

This is with RPi i think ?
@ least this way it was when i look over bitmine.ch design.
The problem with Rpi is that it shares common break pint in the CPU between USB and Ethernet
actually the Ethernet is emulated over usb port of the Pi.
That is the main reason we are working since the beginning with support for TP link routers.
Our way have it's downside too, lack of memory for fancy graphics and stuff

Yes, with the Ras-Pi...

I will assume, due to the issue, that it is a generic issue with the Ras-Pi.

Might need the re-initialization process altered to restore a previous session (IP), which seems to cause the hang-up.

In any event, it seems that the PSU is not setup to be turned on/off through the daughter-board. (No watchdog monitor?) Just takes a high/low signal to tell the PSU to go into "Sleep/Hybernate" or "Off/On" or "Reset mode". I forget the pinouts and the actual command, but it should be addressable in the hardware/software. If not, a jumper connection to change power-states of the PSU might be in order. (Has to do with the pin you jumper, and the "Ready" pin.)

P.S. Have three modules running now. Reaching the limit of my feeble 1020watt PSU. Have room to spare, for the spikes...

Operating 1 card with 2 fans and a 70c target temp... (256 share diff)
218GHs 308w (1c)&(2f)&(70t) * Peak
203GHs 253w (1c)&(2f)&(70t) * Settled-in
198GHs 244w (1c)&(2f)&(70t)
180GHs 223w (1c)&(2f)&(70t)
175GHs 220w (1c)&(2f)&(70t)
106GHs 210w (1c)&(2f)&(70t)
76GHs 167w (1c)&(2f)&(70t) * Oddity, prewarm stage. (Settled down here for a few minutes, before going back to 203GHs)

Operating 3 cards with 6 fans and a 50c target temp... (512 share diff)
555GHs 770w (3c)&(6f)&(50t) * Peak
545GHs 760w (3c)&(6f)&(50t)
535GHs 730w (3c)&(6f)&(50t) * Settled-in
525GHs 731w (3c)&(6f)&(50t) * Settled-in
515GHs 725w (3c)&(6f)&(50t)
500GHs 721w (3c)&(6f)&(50t)

Operating 3 cards with 6 fans and a 60c target temp... (512 share diff)
585GHs 835w (3c)&(6f)&(60t) * Peak
575GHs 806w (3c)&(6f)&(60t)
565GHs 802w (3c)&(6f)&(60t) * Settled-in
555GHs 770w (3c)&(6f)&(60t) * Settled-in
545GHs 760w (3c)&(6f)&(60t)
535GHs 730w (3c)&(6f)&(60t)

Conclusion... With my setup... "At the wall"... for 120v/110v PSU operation.
Each card pulls between 200-250w (Leaning to 250w when settled.) * 5 = 1250w (* 6 = 1500w)
Fans pull about 10w each, 90w with all 8 fans, + 1 if you have the big one running too.
Total wattage 5x = 1340w at the wall (1400w PSU ideal.)
Total wattage 6x = 1590w at the wall (1500w PSU adequate, but borderline.)

Running 6x cards, it would help to use the 120v/110v fans, which would spare you 90w of headroom in the PSU.

Running two units of 3x cards, you can easily supply all power with a single 900w-1000w PSU. (850w may not have enough 12v rail power for peaks.)

So-far, the large heat-sinks seem to be overkill, and the mini-heat-sinks on the chips, seem to be too small to capture enough air-movement for cooling. They get blisteringly hot to the touch, and max-out on my thermal imager. Working on an after-market solution to appease my comfort-level for those under-sized units. (There is too much resistance through those fins, causing the air to just blow right around them. Short-term fix is to use a ducted system to deliver direct air evenly through each individual heatsink. Ultimately, I will slice-up some nice wide heat-sinks to replace the miniature blocks.)

I have completely wrapped the cards in a full shroud. All air is currently being ducted evenly through the cards. There is no loss of pressure through the sides or venting out the structure. Essentially, it is wrapped in a cardboard blanket. xD Looks ugly, but that will be replaced with plexiglass, and the whole unit flipped up, so the smaller heat-sinks have no chance to fall-off. While also being fastened down with a tension-bar, for better pressure contact. I am not a fan of casual contact on hot surfaces.

Otherwise, it is running great. Persistent operation and hash-rate. Just under what was expected, due to my 60c target temperature. Playing it safe until I have my wishlist complete.

[opieum2]

Now the million dollar question is how to bring the "dead" boards to life....Having 6 boards that are totally dead (one board preventing the system from even powering on). Any thoughts? Thanks again for all your help and contributions.

I am currently working on a modified linux version using the latest cgminer and pidora as the base with screen (as the web interface for me has no use others thought it might be better) As I use other tools to monitor my mining hardware I prefer to have all the needed pieces to set up agents for real datacenter management and monitoring. I hope to have a beta sometime this weekend. I will post it if anyone is interested. It should work on most hardware (I am compiling cgminer support into all the various chips out there using 4.3.3).

Oh and I got my Antminer S2 Batch 4 yesterday....Minor shipping issues with the cards being unseated partially, but overall came in great shape. I seated them all back in to the slot set up the networking on it and was mining away happily. It was overall a pleasant experience. It took me all of 20 minutes to get all setup after I reseated the cards. That was a good experience.

[NotFuzzyWarm]

Hmm, there is a new PnP display/control interface for the RasPi. Called PiFace http://www.edn.com/electronics-products/raspberry-pi/4423743/Raspberry-Pi-control-and-display-interface-is-plug-and-play Wonder if any use here.

Think a faster/more memory Pi clone called BananaPi might help things? http://www.bananapi.org/

[opieum2]

Hmm, there is a new PnP display/control interface for the RasPi. Called PiFace http://www.edn.com/electronics-products/raspberry-pi/4423743/Raspberry-Pi-control-and-display-interface-is-plug-and-play Wonder if any use here.

Think a faster/more memory Pi clone called BananaPi might help things? http://www.bananapi.org/

The existing PI is just fine for mining. The LCD screen tho is a nice to have. At the end of the day though for me its just about link lights and power on the machine....in the case of monitoring I do a build a NOC style setup using various monitoring tools. I consider all this ASIC hardware like I would an enterprise appliance. So I will monitor it in the same manner. The only aspect is to monitor hashrates. I am looking at cgwatcher BUT that requires proxying traffic through it for monitoring which means making a single worker using BFGminer.so that kinda sucks for me. I am working on my own scripts that polls cgminer's average hashrates every 2-5 minutes. Thats something on my roadmap to work on as I build out. Right now not an urgent priority for me personally. There are no real monitoring tools that cater to larger scale environments for measuring hashrates really.

[FrictionlessCoin]

Now the million dollar question is how to bring the "dead" boards to life....Having 6 boards that are totally dead (one board preventing the system from even powering on). Any thoughts? Thanks again for all your help and contributions.

Would really like to know this... I got 3 boards that are definitely dead and 2 boards that are physically damaged from shipment.

I also have an non-existent support staff for my boards and no means of how to RMA them.

[plp]

they have time to change there site ugh ... but no time to reply to any one ..

if any has a way to revive these boards ...... let us know ..

[opieum2]

I was wondering if part of the problem was the way they put on the thermal adhesive, perhaps it needed to be a 2 part epoxy, but only used one part?

I wonder if something like this would work better, 2 part adhesive:

http://www.ebay.com/itm/Arctic-Silver-AATA-5G-Alumina-Adhesive-Thermal-Epoxy-5g-/170708546413?pt=US_Thermal_Compounds_Supplies&hash=item27bf05b36d

Even still, not sure if that would be better than thermal grease:

Arctic Alumina Adhesive is a pure electrical insulator, neither electrically conductive nor capacitive.

Maybe cut something like this up?

http://www.ebay.com/itm/150x60x25mm-High-Quality-Aluminum-Heat-Sink-Cooling-for-LED-Power-Memory-Chip-IC-/360744193465?pt=LH_DefaultDomain_0&hash=item53fe078db9

The radio shack variety silicone based stuff works just fine. In the end though it likely is not that at all. There are clear electrical issues with the hardware. I am taking a look at the broken cards again and checking to see if there is any soldering issues or anything out of place that could have caused a short or a failure. I haven't had time to look at them. I am essentially sitting on two bricks I might not get a replacement for. I reached out to AMT to just replace these. Still have not heard back. Altho they could just show up on my door step at some point too unannounced. I guess we will see. The lawsuit probably has them on full gag so they may not even be able to talk to individual customers either. Least thats my thought.

[mrpark]

So you mean, we don't have to modify the heat sinks? The way they are is fine?

[NotFuzzyWarm]

<snip>
 The lawsuit probably has them on full gag so they may not even be able to talk to individual customers either. Least that's my thought.

Possibly correct but pretty damn sure that they would at least also be forced to give a vetted canned response to acknowledge us... I have never ever heard of a company essentially being gagged from ALL customer contact during a lawsuit.

[NotFuzzyWarm]

So you mean, we don't have to modify the heat sinks? The way they are is fine?

The big sinks are fine. Wouldn't say that about the small chipsinks... Set the rig so the boards are horizontal and chip side up so they cannot slide/peel off. If they are chip side down, um... Still better than sideways.

Yes the thermal epoxies are great but permanent. I would not use that until a better chip-side heat sink solution is used.

[opieum2]

So you mean, we don't have to modify the heat sinks? The way they are is fine?

The only recommendation I have is make sure there is some kind of airflow on the last card that has the little heatsinks....I used another fan to blow air on that because I was concerned that would blow out. But as ISAWHIM noted they get really hot. So it might not hurt to get some better heatsinks.

I think these would be ideal:
http://www.frozencpu.com/products/18468/vid-224/Akust_Copper_Ramsink_Chip_Heatsink_-_22mm_x_8mm_x_5mm_-_8_Pack_RS00-0604-AKS.html?tl=g40c16
They have a larger surface area and not so large as to mess with other components. And its copper so heat transfer will be optimal. Plus due to the how they are shaped for airflow purposes will allow heat transfer right off the fins in a nice neat way. I am going to go with these myself. Price is right for the amount I get. Plus they are oriented perfectly for the cards.

Be sure before applying that the gunk from the previous ones is rubbed off with rubbing alcohol. But the overall surface area will be great for these. This could reduce some overheat issues (if thats the problem). AMT probably went with the ones they did for cost purposes. And they do work. But once you overclock its when there is a worry. Of course there could be a bunch of other reasons too. At any rate for your working hardware, just apply these. I think they are a safer bet. If anyone has better recommendations, well there probably are better options. I just could not find better for the price.

[FrictionlessCoin]

So you mean, we don't have to modify the heat sinks? The way they are is fine?

The only recommendation I have is make sure there is some kind of airflow on the last card that has the little heatsinks....I used another fan to blow air on that because I was concerned that would blow out. But as ISAWHIM noted they get really hot. So it might not hurt to get some better heatsinks. I am looking into that and will make a recommendation once I find something that works.


That was the case on one of my miners which fried. That is the card that wont allow the system to power up. I have 5 others that can be plugged in but do not show up. I am hoping vainly that its a software issue but my guess is its not. This is one reason I am building the linux distro I am for these things. It could in all likelihood be a driver issue as the driver has been updated previously.

With my system,  I was able to see it hash for about a minute or so then it went dead.   That's 3 boards.  It went dead precisely when I had changed the pool locations.     When I run cgminer, it does not see any A1 chips even though the amt-setup command does show the presence of the boards.

So I am really unsure if it is also a software problem.  

[opieum2]

So you mean, we don't have to modify the heat sinks? The way they are is fine?

The only recommendation I have is make sure there is some kind of airflow on the last card that has the little heatsinks....I used another fan to blow air on that because I was concerned that would blow out. But as ISAWHIM noted they get really hot. So it might not hurt to get some better heatsinks. I am looking into that and will make a recommendation once I find something that works.


That was the case on one of my miners which fried. That is the card that wont allow the system to power up. I have 5 others that can be plugged in but do not show up. I am hoping vainly that its a software issue but my guess is its not. This is one reason I am building the linux distro I am for these things. It could in all likelihood be a driver issue as the driver has been updated previously.

With my system,  I was able to see it hash for about a minute or so then it went dead.   That's 3 boards.  It went dead precisely when I had changed the pool locations.     When I run cgminer, it does not see any A1 chips even though the amt-setup command does show the presence of the boards.

So I am really unsure if it is also a software problem.  

That is an interesting bit of info, which supports my hypothesis. You switched pools THEN it went dead. I had a similar problem. Since the PI, backplane and daughtherboard still work, I am going to compile A1 support tonight for cgminer and see what happens. Right now I just have the pi plugged in by itself for purposes of getting the distro up and cgminer installed. After that I will plug all the hardware in (tonight if possible) and post results. Knock on wood and hope this was a software problem with SOME of the cards (I am only sure one card is dead). We are running a pretty old version of cgminer on the AMT/Bitmine systems. Updating to 4.3.3 MIGHT address some of this as the driver code is newer from what I saw on github.

Zombie miner project is a go...

Oh and if it works, I will post the image up for use. I am going just make it a general image for any hardware. Compiling support for various products. I would rather contribute to the community in that sense. While vendors are far more conservative with their firmware, its not always feasiable. But be aware this is at your own risk! I will test it best I can, make sure its stable. But any mods beyond my changes are on the user. I will post it in the software section later and link it here. Depending on how well things go tonight I may or may not get it up tonight. Either way tho it will happen this week. At the very least the distro will be up even if it does not fix our problem.

[mrpark]

I may try my soon, are there any basic instructions for changing the pool? I was able to log onto the http panel before, maybe I can just change it in there... I have my boards in a case, so I would imagine I have to take the cage out and mount it the other way.

[opieum2]

I may try my soon, are there any basic instructions for changing the pool? I was able to log onto the http panel before, maybe I can just change it in there... I have my boards in a case, so I would imagine I have to take the cage out and mount it the other way.

Yea that would be best. You could do it via the web console. Just be aware of the risk with that. Frictionless and I lost boards changing pools. Which is what makes me wonder if it is actually a software problem that can be fixed. If I fix it using software I will post detailed instructions on how to apply the firmware. I am sure those of us with giant paperweights could do with some good news in this. I know I can lol.

[mrpark]

The problem is I never changed anything, so it's on the default pool. Can I mine with it on the defaults? I would obviously have to make a worker somewhere...