|
I rather wonder how the 1st 1/2 of the year sales of some component makers are going to do from the overall miner mess. Safe bet that a lot of critical parts are pretty tapped out like Vcore buck inductors, the A1 uses 6 per board other chip designs use similar number per-watt needed, the multitude of power chip caps, etc. Multiply that by 5 boards for 1TH system and then by how many hundreds (thousands?) of orders from AMT, Bitmine, Innosilicon/Dragon, etc. Not to mention again the other ASIC rig makers using other chips.
With all of them hitting the market at the same time - eep. The supply chain doesn't have that much of a surge buffer for power components if everyone waits until the last minute to order. Worse, as AMT found out a component from one maker is not necessarily the same as one from a different maker. Ya gotta look at *all* the data in the spec sheets not just a 7 column summary of family min/max values... Capacity from trusted and design-proven component makers is there but takes time to catch up.
|
|
|
|
Hi AMT. Can you just like buy me a miner from another company? Then we can call it even and you get to keep the majority of my $6000?
Been trying that for ages with them. Heck, they sending me one Antminer s2 would let them keep almost 1/2 as profit with the $4,089 I paid for a lowly 520... Even if they are under contract with Bitmine.ch to exclusively deal with A1 hardware, pretty sure that given the state of thing all restrictions like that are or should be off. |
|
|
|
Can someone who has received one of the 1.2 Th/s miners confirm that the heatsink on the A1 chip is only on the top (e.g., there is no heatsink on the backside of the PCB)? I ask because it is clear looking at Innosilicon specs that a majority of the heat is designed to go to the PCB side (back) and heatsinks on both back and front are *required* by spec... that could explain the meltdowns. Sorry if that was covered before.
The chips appear to be assembled properly. BUT one problem is the amount of thermal grease too much and too little and general build quality of the hardware. It seems like there is no uniformity in the build batches. one could have one voltage and another board another....so if you run cgminer on one voltage you will have a board or 4 burn out on you. So any uniform or standard setting (which in actuality was the default bitmine settings that were much higher and would cause the boards to fry almost as soon as they came online) could cause the boards to fry. I ran into this with one miner that had 3 boards pretty much die on me in 5 minutes and then a 4th shortly and I was able to salvage a 5th onto my working miner that had one dead board....that 4th board died a couple of days later while leaving me now with 4 out of 10 working boards..at this point i am expecting them to die at some point. But knock on wood they are still working. So the design is a working design. But the build quality and general QA testing was pretty much non-existent hence the problems. The erratic grease coverage doesn't surprise me. Bet the boards were not tightned down to the main sinks by cross-torquing the screws (and not all the way at once) to spread the grease evenly outward as best one can. Any other way and you get bubbles & dry(ish) spots. Dinna help that AMT said they were getting folks from the office to help build them. Isn't that the blind leading the blind?... Realistically, the only areas on the main sinks that need thermal compound/grease is for the Vcore FET's and where the A1's thermal vias are. With the boards being FR4 there is minimal lateral heat flow aside from what the ground plane cladding provides. Now if Tlam or Tclad boards were used... I've been meaning ask about the core voltages and how they are set. I take it with the current (original) design all boards are set to the same Vcore level? Or can each boards be tweaked on it's own? With the technobit solution and seeing the sshot of BitMine's 2.4 TH it looks like each um, module, is pretty much an independent miner which hopefully means they can be tweaked one at a time. Does cgminer allow for individual module addressing for speed/voltages? |
|
|
|
Can someone who has received one of the 1.2 Th/s miners confirm that the heatsink on the A1 chip is only on the top (e.g., there is no heatsink on the backside of the PCB)? I ask because it is clear looking at Innosilicon specs that a majority of the heat is designed to go to the PCB side (back) and heatsinks on both back and front are *required* by spec... that could explain the meltdowns. Sorry if that was covered before.
If you look back at pics of some the the miner footballs you will see very large backside heatsinks. In fact the boards are bolted to them. edit: better yet, Opiums system https://bitcointalk.org/index.php?topic=569769.msg6467633#msg6467633 |
|
|
|
Beats the 3.6kw my 9 s1's take for total of 1.7TH... The name of the game now is energy efficiency and performance. Honestly the S2s for the price dont do too bad. NOt great...but not bad. THe spondoolies would be even better.....but if you are a home miner expect a hair dryer like sound coming from that thing....Those were made solidly for datacenters. But S1s as nice and reliable as they are no longer AS efficient powerwise. Bitmines blog pic is a bit misleading. We dont have any information on how long its been running and such. There are key points they left out of the cgminer window.....when it was started and the other important data....like the hashrate of each card and errors and such. They could technically just edit out things like the pool info or other info that would be identifiable. Considering their track record also I am a bit doubtful until they post a full screenshot of that. That is the first time I have seen a screenshot cropped like that for a cgminer stat. Ja. My S1 pharm is still pulling in $36 per block at todays rate - after electric cost. Today our pool popped 4 blocks (good Luck on all 4 usually do ~2 a day) and with DGM, very happy here  So still well worth the load. I'll soon move it to our plant where the electric is 1/2 my home cost of 15 cents/kWh. Oh, and of course the S1's have a eco-modes of 160, 140, & 120GH/s. Running @120 the power draw is 1/3 of the load when running @ 190GH so not a bad trade off for later. Love that BM's 2.4TH has a real server PSU with very nice looking buss-bar system. Structurally, I wonder how the cards are held in place. The heatsinks bolted to the bottom of the case would be great. But - for power eff gotta say BFL's new chip is amazing @0.55w/GH when running at 1TH. That's 550w for the math impared... https://forums.butterflylabs.com/blogs/bfl_josh/386-monarch-update-30-april-2014.html Now they just gotta get them out the door... |
|
|
|
Beats the 3.6kw the 9 s1's in my Ant pharm take for total of 1.7TH... |
|
|
|
Burn it tests?  Are you freakin encouraging AMT to mine with our boards? HELL YES! Do you know of another way they can be tested under load to make sure the djinn doesn't get angry and release its majik smoke? Dinna think so... They could either contact us for a pool address & worker name/pw or just use testcoins. |
|
|
|
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. This might do it. http://www.digikey.com/product-detail/en/ATS-52230P-C2-R0/ATS1134-ND/1284848 Certainly more surface area that the ones BM/AMT used, just be sure to check its base size to get one that covers the entire top of the chip. Again, due to gravity highly suggest having the boards horizontal. |
|
|
|
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. |
|
|
|
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. |
|
|
|
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. |
|
|
|
@AMT
I understand that you were throwing out some ideas to get a feel for what people might find acceptable, but now the professional approach would be to choose which are most viable, then get out your list and your email machine, and starting at the top (lowest unfulfilled order # ) send out a n email with a detailed description of options and what each one will entail so you can get back on track and get this show on the road in the fairest possible manner.
I don't think a public forum is the place to book peoples preferences as I know of at least one customer in the upper 600's who doesn't fallow this forum and has not received any kind of response to his 3 non-harassing phone messages, left over the last several months.
if you don't plan on compensating people who have waited the longest with extra hash power at least make them your top #1 priority.
Thanks
This is a fair point. But again the simplest thing would be to just get third party miners as suggested. FYI good job for all keeping things civil. There are certain people who I will name publicly (clenell) who hate that we are having a civil discussion. He keeps pming me and apparently others about how we are "sucking AMT's dick" Mostly referring to me really. I do happen to think having a civil discussion is necessary to get results. And AMT has been more forthcoming because of it. Hopefully now tho we get actual tangible results from it. Ditto all around. As for clenell, considering he is lawyered up already he should ask them about guidance regarding the laws against cyber harassment... A 'difference of opinion' and 'Freedom of Speech' is one thing - continued targeted harassment of people over it is a whole 'nother matter and in most locations is a civil offense. Report him to the Forum Admin at least. |
|
|
|
Hello everyone,
I just noticed something very very interesting. On the AMT website, there are new, more in detail pictures of the lower power miners.
1. How could they possibly sell any of these for such a high price?
2. Why would they want to try to sell more of them if no one will buy them???
3. Does this mean that they're trying to make themselves look more professional?
4. Coincraft chip page is also updated and moved to the front of the list. Maybe they want to sell out all of the chips?
What do you guys think?
Also to AMT-
If you hadn't received my last message I will go with option 1. Please confirm. Thanks. (Order #1275)
That is the same page they have had up for months. here is a wayback from a week ago http://web.archive.org/web/20140501001729/https://advancedminers.com/ , was the same when I ordered in late Feb as well. Yes the prices on it are insane but have not changed on the smaller ones in months. |
|
|
|
|
Snippet from SirMines everyones elses business, "THEY ARENT EVEN REGISTERED AS A U.S BUSINESS! They've been using Advanced Mining Technology's name, which is an actual mining company (other business) founded in the 80's. "
They are registered in Delaware: File Number: 5341525 Incorporation Date / Formation Date: 05/29/2013
There are several companies using the name Advanced Mining Technologies shown on a quick search, one in Australia, one in India and one in Canada as well as the one in PA. So what? As long as the 1st entity to claim the name doesn't bitch is all perfectly legal.
|
|
|
|
Is there any evidence of a 1.2THs working that is stable?
I know only of a 520 GHs system working reliably.
Bitmine will only sell systems that is below 800 GHs.
There is no evidence at this time of an AMT system that runs reliably at 1.2THs.
I am guessing that the A1 chip for now at least needs to be ran at lower speed for best stability/life. Ergo BMch's only selling 800's. Also note that the Dragon is also sold as 1-1.2TH implying they prefer the slow rate. Well, then maybe they need to bump up the # of chips per board if possible or add another blade. Say maybe 10 per-board vs 8. Hmm, wonder how that balances the GH/W figure? Maybe you can't build such a system without requiring a power draw the exceeds most residential wiring. Well, yes we *are* talking about the max per standard 110v plug/circuit... They need to emphasize that in a home setting these are what would be classified as a large appliance as far as power draw goes. Preferably use a 220v circuit like one has to if they say, buy a welder. Just how it goes. |
|
|
|
Is there any evidence of a 1.2THs working that is stable?
I know only of a 520 GHs system working reliably.
Bitmine will only sell systems that is below 800 GHs.
There is no evidence at this time of an AMT system that runs reliably at 1.2THs.
I am guessing that the A1 chip for now at least needs to be ran at lower speed for best stability/life. Ergo BMch's only selling 800's. Also note that the Dragon is also sold as 1-1.2TH implying they prefer the slow rate. Well, then maybe they need to bump up the # of chips per board if possible or add another blade. Say maybe 10 per-board vs 8. Hmm, wonder how that balances the GH/W figure? |
|
|
|
|
The pool selection point would be my concern as well. I very very much like the one I use and throwing extra horsepower at it be it hosted or my own would be marvelous...
Still prefer a miner of my own though.
|
|
|
|
greases fine for the back side only. Top side should have a gap filler eg Fujipoly's sarcon line or again Berquist Blue to actually embed the chip so ya pull heat from every sq micron of area ya can.
Actuallly, that dynatron setup looks pretty good, only concern would be the fan life, Tip: what kind of bearing used in the fan? Sleeve bearings do not like vertical applications and give like 1/2 or less of 'rated life' when ran vertical. Look for rifled or ball bearing.
That's a ball bearing blower i believe. But the setup is a pricy, like $50 per, and we dont know about the lifespan yet either, we've had it running for a day so we want be a bit more sure than that before we decide to move on 2k dynatron sinks. One thing we've learned from Martins stuff is that if you use it a delivered you can usually get 3-6 months out of it without altering it. Also the board/program are always set to overclock, so under-clocking out of safety would be needed as well. Hmm. the only thing I don't see at the dynatron site eg http://www.dynatron-corp.com/en/product_detail_1.aspx?cv=1-4-216&id=94&in=0 is the thermal ratings... As in thermal resistance at least so you can figure out the temps vs power in. Look at Wakefield for example http://www.wakefield-vette.com/products/forced-convection.aspx complete specs so you can get a handle on chip temps vs sink size vs cfm of air needed instead of just testing what looks good.... Um poking around that Wakefield site, at these power levels you really should look at their liquid cooling ideas http://www.wakefield-vette.com/products/liquid-cooling/liquid-cold-plates.aspx That would allow a very very compact THS-dense processor package and allow moving the radiator/fans(s) to wherever the PSU bank is... Look Koolance or frozen cpu for dripless fluid connectors to use between the boards and cooling loop Just a thought considering multi-TH miners are multi-kw... |
|
|
|
|
greases fine for the back side only. Top side should have a gap filler eg Fujipoly's sarcon line or again Berquist Blue to actually embed the chip so ya pull heat from every sq micron of area ya can.
Actuallly, that dynatron setup looks pretty good, only concern would be the fan life, Tip: what kind of bearing used in the fan? Sleeve bearings do not like vertical applications and give like 1/2 or less of 'rated life' when ran vertical. Look for rifled or ball bearings because the last thing folks running farms are going to want to do is replace a heap'o fans every couple k hours (a few months running 24x7).
|
|
|
|
|
Show Posts
