Topic: DragonMint T1 16TH/S halongmining.com - page 50. (Read 87857 times)

Just a side note....does that mean these miners basically will only work on Slush? or maybe a couple others? and/or they have to have an ASIC BOOST
capable pool JUST to work at all? and as such...would these be mainly bitcoin cash pools?

just trying to catch up ..this seems more than a bit confusing..

*just curious* *don't have a dragonmint on order*

brad

Is that because the pool will just reject the shares with modified version? If I understand correctly, there is nothing in the hardware implementation for AB that would make it incompatible with non-AB pools.

Are u sure?? If so, how did they pull off their demo video off https://www.youtube.com/watch?v=SRCsQUyR7_I
It showed 16.26 TH at 1470 watts? Recheck your readings perhaps. That video was using the standard PSU.

This video was in nov 2017, about 4-5 months back.

Your room temperature is 21 degrees celcius, converted from farenheit. At wat temps are your miners running?
S9 is around 13.8TH for 1420w in hagsfin review - https://bitcointalksearch.org/topic/reviewguide-antminer-s9-135-ths-1320-w-bitcoin-sha-256-asic-miner-2676763  This S9 is on 220-240v wall socket, same as the T1 numbers here.

S9 - 1420w/13.8th =   102.9 watts per TH
T1 - 1489w/15.13th =  98.41 watts per TH
T1 - 1521w/15.86th =  95.90 watts per TH
T1 - 1656w/15.38th =  107.67 watts per TH
T1 - 1688w/15.42th =  109.46 watts per TH

It is at around 5-7% more efficient as S9 with those numbers u mention. If the 1600+ watt range were used, it can be less efficient than an S9.
Can u double check because it makes no sense how it is that far apart. 1688w and 1489 watts is a 199 watts difference. It does not seem right.

https://bitcointalksearch.org/topic/m.33615552




this interests me

lets say  the numbers give at the wall are correct

15.13 th and 1489 watts

the 14 th s-9  does

13.9 to 14  at  1500 watts on a bitmain psu

and 13.9 to 14 at 1450 watts on a delta 2400 watts psu


so  1489/15130 =  0.0984 at the wall  T1

vs 1500/14000 =  0.10714 at the wall  s-9


that is 8.88%  better with the T1


unfortunately   Halong has yet to send me a unit  so I can do proper comaprasion of power savings.

the ones above  are  only partly mine the s-9 numbers are accurate.

the t1 may be accurate but it is with a different psu.

I can do side by side tests with 5 different quality psu's which will let us see if 8.88%  is your savings or improvement.

I reached out to Halong  today via text and got a fast reply  they apologized for not shipping gear to me.

they said "soon and sorry for delay"



Maybe Monday or Tues   is my hope

Not without a heck of a lot of work.

Can someone who got their Dragon take high res picture of chips without heatsink ?

Is this account yours?
https://twitter.com/cobrabitcoin?lang=en

It's a prototype Innosilicon BTC miner. How Halong has obtained these is anyone's guess? Even the Web UI is just re-skinned/re-shuffled from an existing Inno one: https://i.imgur.com/KzC3RoP.png. If Innosilicon wanted to compete with BITMAIN through Halong as their proxy, surely they would make better effort to hide the association?

Do you guys think there is even any ASICBOOST in the chips? We are not seeing the kind of performance and efficiency gains one would expect. We need some experienced people to reverse engineer this hardware and see what is happening. Something is not adding up here.

I assumed an average of 1550w based on 1480+/-8% and experience with S9s.  1400w could be reached with a few percent, but I'm also guessing that the yield won't be nearly as consistent as what you could expect from Canaan.

I'd be interested in knowing if the connectors on the power supplies and receptacles on the dragonmint are plated with anything.  I've been using a thermal camera images of my Canaan and Bitmain miners for preventative measures.

PCB designer here...

When I look at the side by side boards I see something that was shared.  No one would spend so much time to accidentally make a board look that similar.  It would take much more effort than just designing from scratch.

When we see similar boards we count the things that are the same in terms such as "a few", not the ones that are different.    If they were designed independently someone would have had to go to a lot of trouble to make them so similar.

I've had PCBs copied and reproduced.  This doesn't look like that.  This looks like someone had access to inno designs and made slight modifications, and routed the modifications.  This is far different than the car examples, or both boards are green memes that people have been posting on twitter.  The examples of both intel based motherboards having the same chip-sets just don't apply.

The level that boards are identical is just embarrassing.

For a couple grand they could have reverse engineered the PCB, turned it into a schematic, made their own layout, and at least made it look like they were trying not to copy someone.  There are lots of contractors who perform this service in China.  Of course the US has them too, but they are more expensive and typically prefer to do it on designs a company owns but lost to obsolesce. 

Occam's razor guys...

yes I have

 the ibm 2980
the ibm  2880
the delta 2400 --------best server psu ever of the 15 different ones I used
the hp   2000
the hp   1800

the corsair  axi 1500

the bitmain 1600

my guess is  the 2000 watt or more are better servers

such as the ibm 2980

and the delta 2400  will be under or close to  1400 watts.

if these do 15.5th at 1400 watts

vs the s-9  doing  13.5  at 1400 watts   they are a big improvement.

Still waiting for the demo  to show the side by side comparisons but no one got back to me with a tracking number.

It's fair to say that the code to tune for optimal hashrate at minimum power draw is still in development, and these are the first public runs of silicon so I expect hashrates and/or efficiency to improve successively with later ones, and firmware updates to improve things slightly for existing ones. Small nm ASICs are extraordinarily fussy and tricky to tune.

good thought, do you have the IBM 2980 as well to test with?

If your theory is correct, would be worth swapping these out

If I get some spare time, will try Bitman 1600w psu and check for any improvements in power consumption

yeah  which rates  at 1600 watts.

and while it may not burn up at 90% of its max  it loses efficient factor.  It may only be silver at high power use.

I would be willing to say the delta 2400 watt or the ibm 2980 would do a lot better
 maybe  1400 watts vs your lowest at 1489 watts

---

Moderator's note: This post was edited by frodocooper to remove multiple nested quotes.

other four rigs are using MyRig/Halong PSU

okay  the 110v evga 1600 g2 is lacking so it is worse at 1688 watts -  I suggest getting something else for that or using a 700 watt on 1 board.


but all of them are higher watts  then I thought they would be.

so what psu's ?

do you have a delta 2400 watt psu on any of them?

or all antminer ap3++

I would love to test my demo when I get it to see if a 2400 watt delta is better

Decent I suppose at 1489w and 15.13TH/s, but not really all that impressive. Especially with all the fuss about the asicboast and FlushPool.

My five units are running at various speeds and power draw: 1489 watts, 1521watts, 1656 watts on 220V single phase and one at 1688 watts 110v using an EVGA 1600 G2. (5th unit I have not checked wattage but pulling 7 amps on 220V). room temperature is 70 degrees

Hashrates showing on Dragonmint miner status page: 15.13TH/s, 15.86TH/s, 15.38TH/s, 15.42TH/s, & 15.40TH/s  (slushpool)

Inno - who knows.
Halong - given their very odd ideas of how publicity works and their hubris, seems perfect for their playbook.
Wonder how they are going to approach the Patent application from Intel filed today for a BTC hardware accelerator? Even though it makes my head hurt reading patent applications, it sure looks and feels like another variant of AB to me...

Just one thing that confuses me in this, is that why both parties (Halong and Inno) do not want to confirm anything about this OEM or other partnership deal?

Revealing the fact that the control board design is very much the same got me kicked out from the Dragonmint telegram chat room.

What are their reasons on this?