New Official AMT Thread

[NotFuzzyWarm]

Probably the thermal tape was poorly applied, creating an air pocket that led to that. Usually the case from what I have seen. Considering the assembly quality in general and overall leakage of the thermal grease this is not a shock. It looked like a rushed job.

This is indeed problematic for a couple of reasons:

(1) There is no screw that ensures that the heat sink has some pressure against the chip.
(2) The orientation of the boards are such that the boards are vertical and therefore gravity isn't even going to work to keep the heat sinks in place.  

Also, I've been saying for a while now.  These heatsinks are too small.

Look at the designs of both Technobit and Dragon.  They've got massive heat sinks on the chip side and it's okay even if the chips are all bunched together.   Even the AMT system has a much larger heat sink than what we see here at AMT.

I can't understand the design.  These are 28nm chips running on full throttle, AMT however decides to put the smallest heat sink one can imagine!  

Yes, in my opinion very botched layouts. However - remember that the chips were first tested what, late Dec/early Jan? BMch and AMT had no idea of the power loads the chips would produce and had planned on it being a lot less. Technobit being a bit hobbiest-orientated had the time to deal with the real-world specs and before selling their systems had the advantage of seeing the trouble BMch ran into. Also helps that they started with 2 and 4 chip single boards to cut their teeth on.

Le sigh... To answer some question points posted, one more time thermal/power design 101:
Thermal flow of the A1 chip. Look at the A1 data sheet on Github.
70% goes out the bottom, through thermal vias on the board and then to the main heat sink on the backside. That leaves 30% out the top/sides. Obviously no one ran the math to know what that means for power in/sink capacity (inc airflow used)/chip temp on the top side much less what the die temp is based on the heatsinks used top & bottom along with their required airflows. I'd guess that like many non-design folks do they just thought 'big sink and a few smaller ones on top will be fine'...

Physical layout - "okay even if the chips are all bunched together".
No 'if'. That is how it should have been done to begin with.
It makes thermal control much easier to deal with - including overtemp sensing... (is there any?)

Same with clustering the Vcore buck inductors much closer to the chips not to mention more point-of-load filtering surrounding the chip cluster as Technobit uses. From the power standpoint, given the low voltages used distance between the Vcore supplies and the chips must be as minimal as possible to keep losses on the power planes/traces to an absolute minimum. The A1 spec says max core current is 20amps. It does *not* make any mention of spike loads - all processors have them and typically are over 2x the average running value. Spread out as everything is on the Bitmine-AMT boards I'd expect to see serious fluctations if anyone bothered to probe the chips power pins...

Use of thermal tape - really only advised if there are no shear loads put on the heatsink. If there are (gravity, g-forces) maximum permissible temps and load values must be within the mfrg recommendations or other means to secure are needed.

[mrpark]

Probably the thermal tape was poorly applied, creating an air pocket that led to that. Usually the case from what I have seen. Considering the assembly quality in general and overall leakage of the thermal grease this is not a shock. It looked like a rushed job.

This is indeed problematic for a couple of reasons:

(1) There is no screw that ensures that the heat sink has some pressure against the chip.
(2) The orientation of the boards are such that the boards are vertical and therefore gravity isn't even going to work to keep the heat sinks in place.  

Also, I've been saying for a while now.  These heatsinks are too small.

Look at the designs of both Technobit and Dragon.  They've got massive heat sinks on the chip side and it's okay even if the chips are all bunched together.   Even the AMT system has a much larger heat sink than what we see here at AMT.

I can't understand the design.  These are 28nm chips running on full throttle, AMT however decides to put the smallest heat sink one can imagine!  

Yes, in my opinion very botched layouts. However - remember that the chips were first tested what, late Dec/early Jan? BMch and AMT had no idea of the power loads the chips would produce and had planned on it being a lot less. Technobit, Dragon and others had the time to deal with the real-world specs and before selling their systems had the advantage of seeing the trouble BMch ran into.

Le sigh... To answer some question points posted, one more time thermal/power design 101:
Thermal flow of the A1 chip. Look at the A1 data sheet on Github.
70% goes out the bottom, through thermal vias on the board and then to the main heat sink on the backside. That leaves 30% out the top/sides. Obviously no one ran the math to know what that means for power in/sink capacity (inc airflow used)/chip temp on the top side much less what the die temp is based on the heatsinks used top & bottom along with their required airflows. I'd guess that like many non-design folks do they just thought 'big sink and a few smaller ones on top will be fine'...

Physical layout - "okay even if the chips are all bunched together".
No 'if'. That is how it should have been done to begin with.
It makes thermal control much easier to deal with - including overtemp sensing... (is there any?)

Same with clustering the Vcore buck inductors much closer to the chips not to mention more point-of-load filtering surrounding the chip cluster as Technobit uses. From the power standpoint, given the low voltages used distance between the Vcore supplies and the chips must be as minimal as possible to keep losses on the power planes/traces to an absolute minimum. The A1 spec says max core current is 20amps. It does *not* make any mention of spike loads - all processors have them and typically are over 2x the average running value. Spread out as everything is on the Bitmine-AMT boards I'd expect to see serious fluctations if anyone bothered to probe the chips power pins...

Use of thermal tape - really only advised if there are no shear loads put on the heatsink. If there are (gravity, g-forces) maximum permissible temps and load values must be within the mfrg recommendations or other means to secure are needed.

Clearly the design not tested or calculated, thermal tape is used so you can put the tape on the chips instead of applying the silicon on from a tube. (you see it on boxed Intel processors) It is not meant to glue things on a circuit board. There should be a thin metal bar holding down the heat sink. Not only didn't it work standing still, you have to account for things moving around as well. Fail, fail, fail...

[FrictionlessCoin]

Clearly the design not tested or calculated, thermal tape is used so you can put the tape on the chips instead of applying the silicon on from a tube. (you see it on boxed Intel processors) It is not meant to glue things on a circuit board. There should be a thin metal bar holding down the heat sink. Not only didn't it work standing still, you have to account for things moving around as well. Fail, fail, fail...

I'm not too familiar with the characteristics of thermal tape, however I do know for thermal paste, a lot of them start flowing when heated up.  So if thermal tape is like this, then these chips would just fall off when given enough heat.

[rik_khaos]

Probably the thermal tape was poorly applied, creating an air pocket that led to that. Usually the case from what I have seen. Considering the assembly quality in general and overall leakage of the thermal grease this is not a shock. It looked like a rushed job.

This is indeed problematic for a couple of reasons:

(1) There is no screw that ensures that the heat sink has some pressure against the chip.
(2) The orientation of the boards are such that the boards are vertical and therefore gravity isn't even going to work to keep the heat sinks in place.  

Also, I've been saying for a while now.  These heatsinks are too small.

Look at the designs of both Technobit and Dragon.  They've got massive heat sinks on the chip side and it's okay even if the chips are all bunched together.   Even the AMT system has a much larger heat sink than what we see here at AMT.

I can't understand the design.  These are 28nm chips running on full throttle, AMT however decides to put the smallest heat sink one can imagine!  

Yes, in my opinion very botched layouts. However - remember that the chips were first tested what, late Dec/early Jan? BMch and AMT had no idea of the power loads the chips would produce and had planned on it being a lot less. Technobit being a bit hobbiest-orientated had the time to deal with the real-world specs and before selling their systems had the advantage of seeing the trouble BMch ran into. Also helps that they started with 2 and 4 chip single boards to cut their teeth on.

Le sigh... To answer some question points posted, one more time thermal/power design 101:
Thermal flow of the A1 chip. Look at the A1 data sheet on Github.
70% goes out the bottom, through thermal vias on the board and then to the main heat sink on the backside. That leaves 30% out the top/sides. Obviously no one ran the math to know what that means for power in/sink capacity (inc airflow used)/chip temp on the top side much less what the die temp is based on the heatsinks used top & bottom along with their required airflows. I'd guess that like many non-design folks do they just thought 'big sink and a few smaller ones on top will be fine'...

Physical layout - "okay even if the chips are all bunched together".
No 'if'. That is how it should have been done to begin with.
It makes thermal control much easier to deal with - including overtemp sensing... (is there any?)

Same with clustering the Vcore buck inductors much closer to the chips not to mention more point-of-load filtering surrounding the chip cluster as Technobit uses. From the power standpoint, given the low voltages used distance between the Vcore supplies and the chips must be as minimal as possible to keep losses on the power planes/traces to an absolute minimum. The A1 spec says max core current is 20amps. It does *not* make any mention of spike loads - all processors have them and typically are over 2x the average running value. Spread out as everything is on the Bitmine-AMT boards I'd expect to see serious fluctations if anyone bothered to probe the chips power pins...

Use of thermal tape - really only advised if there are no shear loads put on the heatsink. If there are (gravity, g-forces) maximum permissible temps and load values must be within the mfrg recommendations or other means to secure are needed.

Which leads us to the most important question of all which of this is the official AMT theme song?

http://www.youtube.com/watch?v=xNnAvTTaJjM&feature=kp

or

http://www.youtube.com/watch?v=jkDvZ2GlwsM

[NotFuzzyWarm]

anyone have an idea of an alternative heat sink that can be attached to the AMT board?

If you mean top side, just follow BMch's example and use a long one. Just be sure to find a way to bolt it down so it doesn't fall off.

[mrpark]

[sirminesalot]

Retarded.

You can blame the companies AMT hired to do work on these all you want, but eventually the blame falls squarely on AMT for their incompetence. They took everyones money and picked the manufacturers to build these machines. They are 100% responsible.

Anyone that is hardware savvy knows that heatsinks need to be clamped down tightly. The fact that they thought they could use thermal tape to stick the heatsinks on these chips shows complete ignorance of technical skill.

I'm sick of reading all these excuses both Opieum and Fuzzy want to make for AMT. They are responsible for this mess, regardless of who dropped the ball on the design and/or execution.

[FrictionlessCoin]

anyone have an idea of an alternative heat sink that can be attached to the AMT board?

If you mean top side, just follow BMch's example and use a long one. Just be sure to find a way to bolt it down so it doesn't fall off.

Can anyone point to one that will work?

I guess a this time it is not advisable to run AMT design without having a sufficiently large heat sink to transfer the heat.

[NotFuzzyWarm]

At least would stay on but needs at least 3x dissipation area than that support chip sink has. In free air I'd start with a guess of 5 sq. inches of surface PER-CHIP.

[mrpark]

At least would stay on but needs at least 3x dissipation area than that support chip sink has. In free air I'd start with a guess of 5 sq. inches of surface PER-CHIP.

Yea, not that size, but it should have a bar like that or screws on the corners to stay secure. You would never use JUST thermal tape on something that would smoke and burn if it fell off. You might use a tiny heat sink on something like a Pi board that just needed a little help, but could run fine without the heat sink.

[FrictionlessCoin]

At least would stay on but needs at least 3x dissipation area than that support chip sink has. In free air I'd start with a guess of 5 sq. inches of surface PER-CHIP.

5 square inches?  The problem is that there are capacitors in the way of the chip, so you can't mount a big one.

Here is photo:



See the four orange capacitors that's in the way of mounting a sufficiently large heat sink?  Anyway around this?

[mrpark]

At least would stay on but needs at least 3x dissipation area than that support chip sink has. In free air I'd start with a guess of 5 sq. inches of surface PER-CHIP.

5 square inches?  The problem is that there are capacitors in the way of the chip, so you can't mount a big one.

Here is photo:



See the four orange capacitors that's in the way of mounting a sufficiently large heat sink?  Anyway around this?

I guess one could customize it, cut the corners off... (not suggesting you per say).

[mrpark]

Come on guys, get serious. This is the heatsink for my first $300 10Mh/s Bitcoin board. What do you think you should use on a 1.2 Th/s system? Don't go cheap on a $6,000 miner.

[freddyfarnsworth]

At least would stay on but needs at least 3x dissipation area than that support chip sink has. In free air I'd start with a guess of 5 sq. inches of surface PER-CHIP.

5 square inches?  The problem is that there are capacitors in the way of the chip, so you can't mount a big one.

Here is photo:



See the four orange capacitors that's in the way of mounting a sufficiently large heat sink?  Anyway around this?

Solid copper spacers to raise ASIC level above other components.
However the small SMT components may then overheat due to lack of airflow.
Immersion is the only cureall for this design of boards top level.

Holddowns for the small sinks as suggested world take a board revision to accomplish.
I think since that chip is isolated (other chips show no signs of overheating), was a short from dislocated backplane or other damage occurred during ship.

[opieum2]

The most relevant question now that we know what we will be getting for option 2, is how long? This situation is just a mess and we are being forced to wait even longer. So a clear cut commitment from AMT on how long this will now take (and something we can hold to account). Now seeing as this is being sourced to a 3rd party vendor Maybe they can chime in and provide some input on how long this will take. Being that its Technobit they can give some timeline.

 Honestly I would rather just have the hardware shipped directly from them to me than go from technobit to AMT to me. Seems like it just adds more unnecessary time to the whole process. If Technobit just ships us the miners directly it will save insane amounts of time. Like I said before this way the support matter now gets passed to technobit to address should there be any issues. Considering the shipping issues people have experienced it seems like a more viable option for us to just get it directly. At this point it would be a paperwork issue with AMT and Technobit confirming orders fulfilled and checking with customers. Frees AMT up from alot of added hassles.

[mrpark]

The most relevant question now that we know what we will be getting for option 2, is how long? This situation is just a mess and we are being forced to wait even longer. So a clear cut commitment from AMT on how long this will now take (and something we can hold to account). Now seeing as this is being sourced to a 3rd party vendor Maybe they can chime in and provide some input on how long this will take. Being that its Technobit they can give some timeline.

 Honestly I would rather just have the hardware shipped directly from them to me than go from technobit to AMT to me. Seems like it just adds more unnecessary time to the whole process. If Technobit just ships us the miners directly it will save insane amounts of time. Like I said before this way the support matter now gets passed to technobit to address should there be any issues. Considering the shipping issues people have experienced it seems like a more viable option for us to just get it directly. At this point it would be a paperwork issue with AMT and Technobit confirming orders fulfilled and checking with customers. Frees AMT up from alot of added hassles.

Technobit might actually spend money on a real box too.

[opieum2]

Come on guys, get serious. This is the heatsink for my first $300 10Mh/s Bitcoin board. What do you think you should use on a 1 Th/s system? Don't go cheap on a $6,000 miner.

At this point its a safe bet they are not using that old design any more. Seems like its more trouble than its worth. Seems like everyone with a miner has problems of some kind AMT already owned that. On the off chance there might be one lucky person with a working miner but likley harder to find than bigfoot. But at this point noone we know of has 100 percent working hardware. Might be easier to just focus now on getting the technobit solutions (or whatever option you go with). I am reading the forum on this https://bitcointalk.org/index.php?topic=376351.0  And it seems like a solid solution. If we are getting the stock design then great....

They were supposed to send out an email to us or some option on their site to make this all official today........AMT any updates on the site change or email?
Also Timelines and process for the delivery of new hardware? Thanks

[mrpark]

Come on guys, get serious. This is the heatsink for my first $300 10Mh/s Bitcoin board. What do you think you should use on a 1 Th/s system? Don't go cheap on a $6,000 miner.

At this point its a safe bet they are not using that old design any more. Seems like its more trouble than its worth. Seems like everyone with a miner has problems of some kind AMT already owned that. On the off chance there might be one lucky person with a working miner but likley harder to find than bigfoot. But at this point noone we know of has 100 percent working hardware. Might be easier to just focus now on getting the technobit solutions (or whatever option you go with). I am reading the forum on this https://bitcointalk.org/index.php?topic=376351.0  And it seems like a solid solution. If we are getting the stock design then great....

They were supposed to send out an email to us or some option on their site to make this all official today........AMT any updates on the site change or email?
Also Timelines and process for the delivery of new hardware? Thanks

Also interested in the timelines, and what to do with our RMA's, please AMT include that in your posting as well.

[eightcylinders]

This would work to cool the chip.  Will it fit when you include fittings (alas, I don't even have non-working parts to look at)?

http://www.frozencpu.com/products/13206/ex-blc-916/Alphacool_HF_14_Smart_Motion_Universal_Chipset_GPU_Block_-_Copper_Edition.html?tl=g30c89&id=wo6LsxuJ

The adjustable mounting bracket should make it possible to create some kind of jerry rigged mounting solution, though right now I can't visualize it without seeing the board and the existing backplane.  Also, even without the mounting brackets, you can use thermal GLUE.  I did this with my BFL 80 and just using thermal glue I got the chip temps down from 80c to 45c.  Of course that is not a reversible solution lol.

[marto74]

HI,all,

Actually AMT purchase a  non-exclusive licence from us about 2 weeks ago.
They did business with us in the past, so having troubles with their current design/manufacturing process , they decided to get an alternate design.
In any case all their boards will be produced in USA as  far as I know.
We will not be involved in the production and assembly except with know-how help if needed from distance.


Best Martin