New Official AMT Thread

[NotFuzzyWarm]

Hmm I just got an email from AMT....I got an invoice...balance paid in full and all that. Looks like they might be restructuring their records keeping process.....least to cover things. Wonder if the company has changed hands or something.

I would hope the next step is to get us all the working hardware we ordered for the invoiced amounts. OR if they are being forced into honoring the MPP that would be good too. At this point tho I am doubtful of any refund option or RMA. I just want working hardware I can hash with. I am stuck with what is a 500Ghs miner when I should have been hashing at 2.4Ths with my two miners. Quite a gap. Well hopefully the next email is that I am getting a product shipped for me if I got an invoice. (even though I got one already from them once before)

Since I had never gotten one I had requested my Invoice from Alyssa a few weeks ago when AMT was replying. Got it the same day. Dates are correct as is amount they received from me, payment method was wrong. Says was wire xfr but was personal check.

[1713553906]

1713553906

[FrictionlessCoin]

http://www.aavid.com/

Pricy, but fast.
You should have two types of boards.

1st with the 4 caps surrounding the chips.
2nd with the only 1 cap out of the 4 in that location.

- First board- this was the original design, provided by bitmine, which also called for a 12x 1500 cap where you'll see the 12x 560 cap there. The manufacturer ordered this component switch which differed from the bom.

- beneath the 560 caps you'll see another location 170 cap on the first board and 330 on the 2nd version with (1500 caps).

The ability to put a railed heatsink exists on the second version but could not exist on the first.
you - and anyone else working on this - list your question in a simple form which allows for an easy to follow response.


ie.  Which value should this component be - snap a pic or something. End post.

So what are you saying?  We can remove any number of caps leaving at least one and it should work the same.  That way we can add a adequate heat sink.  Not the time piece that it currently has.   

[FrictionlessCoin]

Got my order on the way for my plexi-bending stuff... should be here by friday. I will get it setup, and make two cases. One for the full-size 5-card standard design, and one with the 3-card mini design.

Then I will throw everything back into the tower, and show the heat-shroud thing. (Just can't show it running, since I only have a PSU large enough to run 3 cards, and I am down to 2 running cards at the moment.)

Now I have to go to work... I will look for more sources for low-profile, multi-fin heat-sinks when I get back. (I am trying to stay away from the thick heat-sinks, as those just retain heat. You need lots of thin fins, for the greater surface area, to reflect the heat off the aluminum, and into the humidity in the air.)

Also looking for some sheets of thermal-transfer pads, that I can cut-up to match the plates on the PCB's. As opposed to solid shims.

We'll mail you out what we've been using on the first 300 boards before the manufacturer changed the board's mask on the bottom. Basically produced 300 boards with exposed via's in all areas and assumed a thermal transfer pad would be enough to protect against potential shorts.

Pic coming shortly.

What are you saying here?   The manufacturer had exposed vias that assumes the use of a thermal transfer pad.... was a thermal transfer pad used?   

[NotFuzzyWarm]

<snip>
Yea, those seem to be the largest source of heat, once the cards stop functioning. May be the coils insulation failing. Those too, it seems, are not real flush-mounted to the board. There are exposed heat-pads under them, on the heat-sink side. However, the varnish actually raises the heat-sink from direct contact to the metal at all. (Looks like the heat-sinks should be machined to match the pads, but they are not. They are essentially using the heat-sink compound to "fill the gap" between the exposed heat-pad and the heat-sink, which is not the function of thermal grease. Thermal grease is designed to fill the microscopic gaps, not gaps of zero-contact. Which is where heat-pads are designed to function.)
<snip>
Did notice another thing that bugs me a little, about the cards... The PCB sticks-out beyond the heat-sinks, by a fraction of a MM... This bugs me because the frame mounts firmly to the cards by the heat-sink, which has a metal-edge pressing hard against the PCB itself. This too stresses the PCB, and I am not 100% sure that the inner layers may or may not have any stray "copper traces" near the edges, which are in direct contact with the metal frame. Mostly, my concern is the compression of the PCB, as the aluminum expands and contracts from the heat. (This also being a concern due to the solid screw mounting which is firmly holding the non-expanding fiberglass PCB to the expanding alu<snip>

Ja the buck inductors are supposed to be in contact with *something* to suck out some of the waste heat. Along those lines, got a part number on them? With all that is turning up would be a good idea to make sure they are running under their max dc-current rating. If over it then the cores will saturate and say goodbye to regulation.

On traces near the edges:  If whoever laid out the boards knows anything about EMC basics then all traces will be at least 2x the board thickness away from the edges to close near-field loops. Any closer and emissions goes through the roof from the loops becoming true radiated EMI.

Um, that rather raises a point - have these been EMC tested at all? Not sure if miners fall under FCC much less EU exempted equipment...

[NotFuzzyWarm]

What are you saying here?   The manufacturer had exposed vias that assumes the use of a thermal transfer pad.... was a thermal transfer pad used?  

No I take it as the signal and power vias were left uncovered leading to them shorting. Only the thermal xfr vias should be bare. And - due to the thickness of the green/red screen layer that is supposed to seal the boards one damn well better make sure that something bridges the resulting gap between the sinks and the thermal vias.

[FrictionlessCoin]

Got my order on the way for my plexi-bending stuff... should be here by friday. I will get it setup, and make two cases. One for the full-size 5-card standard design, and one with the 3-card mini design.

Then I will throw everything back into the tower, and show the heat-shroud thing. (Just can't show it running, since I only have a PSU large enough to run 3 cards, and I am down to 2 running cards at the moment.)

Now I have to go to work... I will look for more sources for low-profile, multi-fin heat-sinks when I get back. (I am trying to stay away from the thick heat-sinks, as those just retain heat. You need lots of thin fins, for the greater surface area, to reflect the heat off the aluminum, and into the humidity in the air.)

Also looking for some sheets of thermal-transfer pads, that I can cut-up to match the plates on the PCB's. As opposed to solid shims.

We'll mail you out what we've been using on the first 300 boards before the manufacturer changed the board's mask on the bottom. Basically produced 300 boards with exposed via's in all areas and assumed a thermal transfer pad would be enough to protect against potential shorts.

Pic coming shortly.

ETA on the Technobit boards?  Do you even plan on compensating customers with the Technobit board?

The only reason we are trying to debug these boards is because we are trying to salvage what little we got from you!

You sent us junk and we are trying to debug the junk.  

[FrictionlessCoin]

[quote author=FrictionlessCoin link=topic=569769.msg6823734#msg6823734
What are you saying here?   The manufacturer had exposed vias that assumes the use of a thermal transfer pad.... was a thermal transfer pad used?   

No I take it as the signal and power vias were left uncovered leading to them shorting. Only the thermal xfr vias should be bare. And - due to the thickness of the green/red screen layer that is supposed to seal the boards one damn well better make sure that something bridges the resulting gap between the sinks and the thermal vias.
[/quote]

Well if they were indeed shorting...  that problem would have been identified with initial testing.

Don't tell me that these systems where shipped without testing??!

If the layer where improperly exposed, then one could have easily made temporary fix by taping the exposed parts.

[FrictionlessCoin]

Here is the back side.... doesn't appear to have any layer improperly exposed.

[NotFuzzyWarm]

Well if they were indeed shorting...  that problem would have been identified with initial testing.

Don't tell me that these systems where shipped without testing??!

If the layer where improperly exposed, then one could have easily made temporary fix by taping the exposed parts.

Depends. Initially the thermal pad may have kept may/most from shorting out so any initial test would have passed. Unless there were pointy bits that is. After that things moving from thermal cycling and just time will be enough for it pierce the pads. It may well explain the delayed-death syndrome we have been seeing...

Yes Mylar tape would fix that. And again care must be taken to allow for that added gap produced. So far I would not count on that as it strikes me that whoever designed/reviewed the design of all this have zero-experience at it.

[NotFuzzyWarm]

Here is the back side.... doesn't appear to have any layer improperly exposed.

At least that bit of boards looks covered. Remember they dis say that many - not all boards had the uncovered via fault.

Center left is a major no-no that thu-hole for screws is way way too close to those covered vias either side of it. Pressure IS going to cause faults in the screen layer.

[FrictionlessCoin]

Here is the back side.... doesn't appear to have any layer improperly exposed.

At least that bit of boards looks covered. Remember they dis say that many - not all boards had the uncovered via fault.

Center left is a major no-no that thu-hole for screws is way way too close to those covered vias either side of it. Pressure IS going to cause faults in the screen layer.

Yes very close.   Is thermal paste conductive?

[AMT_miners]

Here is the back side.... doesn't appear to have any layer improperly exposed.

First 300 boards. If there are no exposed via's on your board then I guess you're board isn't one of the first 300 and was tested and worker when shipped.

[NotFuzzyWarm]

Yes very close.   Is thermal paste conductive?

For the vast majority on the market, not at the voltages used on these. Possibly a bit capacitive but that is more an issue if it is smeared across bare connections.
Possibly corrosive to aluminum though if containing silver.

[AMT_miners]

<snip>
Yea, those seem to be the largest source of heat, once the cards stop functioning. May be the coils insulation failing. Those too, it seems, are not real flush-mounted to the board. There are exposed heat-pads under them, on the heat-sink side. However, the varnish actually raises the heat-sink from direct contact to the metal at all. (Looks like the heat-sinks should be machined to match the pads, but they are not. They are essentially using the heat-sink compound to "fill the gap" between the exposed heat-pad and the heat-sink, which is not the function of thermal grease. Thermal grease is designed to fill the microscopic gaps, not gaps of zero-contact. Which is where heat-pads are designed to function.)
<snip>
Did notice another thing that bugs me a little, about the cards... The PCB sticks-out beyond the heat-sinks, by a fraction of a MM... This bugs me because the frame mounts firmly to the cards by the heat-sink, which has a metal-edge pressing hard against the PCB itself. This too stresses the PCB, and I am not 100% sure that the inner layers may or may not have any stray "copper traces" near the edges, which are in direct contact with the metal frame. Mostly, my concern is the compression of the PCB, as the aluminum expands and contracts from the heat. (This also being a concern due to the solid screw mounting which is firmly holding the non-expanding fiberglass PCB to the expanding alu<snip>

Ja the buck inductors are supposed to be in contact with *something* to suck out some of the waste heat. Along those lines, got a part number on them? With all that is turning up would be a good idea to make sure they are running under their max dc-current rating. If over it then the cores will saturate and say goodbye to regulation.

On traces near the edges:  If whoever laid out the boards knows anything about EMC basics then all traces will be at least 2x the board thickness away from the edges to close near-field loops. Any closer and emissions goes through the roof from the loops becoming true radiated EMI.

Um, that rather raises a point - have these been EMC tested at all? Not sure if miners fall under FCC much less EU exempted equipment...

Yes you are correct regarding the potential shorts. Traces are close to the edges yes.

[opieum2]

@AMT I sent you guys an email regarding some info regarding the drivers, that is a critical point in making this new firmware work. I have already added some datacenter friendly tools to make this large scale monitor friendly. Ideally I would port your webUI over if you have documentation on that, otherwise I will just try to get phpminer working. A more experienced dev than I already volunteered to assist me with that front.

[NotFuzzyWarm]

AMT_Miners: If you want a board physical design review pm me. I've been involved in power systems not to mention a helluva lot of other circuit design for almost 40 years...

[AMT_miners]

Yes very close.   Is thermal paste conductive?

For the vast majority on the market, not at the voltages used on these. Possibly a bit capcitive but that is more an issue if it is smeared across bare connections.
Possibly corrosive to aluminum though if containing silver.

That would take a while to get the point of corrosion, some of the pastes did contain silver, like the Arctic Silver 5 thermal paste. In general,  despite it having silver we have not found that as the main reason for the problems. The main problem lies in the amount of copper in the board itself, which we had not discovered until recently.

[NotFuzzyWarm]

Yes very close.   Is thermal paste conductive?

For the vast majority on the market, not at the voltages used on these. Possibly a bit capcitive but that is more an issue if it is smeared across bare connections.
Possibly corrosive to aluminum though if containing silver.

That would take a while to get the point of corrosion, some of the pastes did contain silver, like the Arctic Silver 5 thermal paste. In general,  despite it having silver we have not found that as the main reason for the problems. The main problem lies in the amount of copper in the board itself, which we had not discovered until recently.

Amount of copper? As in the cladding being to thin for the currents/thermal loads involved perhaps?

[AMT_miners]

http://www.aavid.com/

Pricy, but fast.
You should have two types of boards.

1st with the 4 caps surrounding the chips.
2nd with the only 1 cap out of the 4 in that location.

- First board- this was the original design, provided by bitmine, which also called for a 12x 1500 cap where you'll see the 12x 560 cap there. The manufacturer ordered this component switch which differed from the bom.

- beneath the 560 caps you'll see another location 170 cap on the first board and 330 on the 2nd version with (1500 caps).

The ability to put a railed heatsink exists on the second version but could not exist on the first.
you - and anyone else working on this - list your question in a simple form which allows for an easy to follow response.


ie.  Which value should this component be - snap a pic or something. End post.

So what are you saying?  We can remove any number of caps leaving at least one and it should work the same.  That way we can add a adequate heat sink.  Not the time piece that it currently has.   

No, you can't do anything and we don't advise that you try anything either.  If your board has 4 caps, you have a 560 cap on the psu.  If you  have 1 cap, you have 1500 caps on the psu.  The difference is 20,000mmf.

[AMT_miners]

Yes very close.   Is thermal paste conductive?

For the vast majority on the market, not at the voltages used on these. Possibly a bit capcitive but that is more an issue if it is smeared across bare connections.
Possibly corrosive to aluminum though if containing silver.

That would take a while to get the point of corrosion, some of the pastes did contain silver, like the Arctic Silver 5 thermal paste. In general,  despite it having silver we have not found that as the main reason for the problems. The main problem lies in the amount of copper in the board itself, which we had not discovered until recently.

Amount of copper? As in the cladding being to thin for the currents/thermal loads involved perhaps?

perhaps.