ANN: BITMAIN has Tested Its 28nm Bitcoin Mining Chip BM1382

[kabopar]

Dont know, mine was batch 1

What was your order time?

[1714656046]

1714656046

[1714656046]

1714656046

[deadreply]

me too frome batch 1 and nothing yet

[shamacala]

Ordered 30th June/ despatched 15th July/ delivered 18th July and im happy to say all 5 are up and running fingers crossed ok

[psjw4450]

Did anyone who receive their units get a email confirmation from BITMAIN with the tracking number?

[TracerX]

Did anyone who receive their units get a email confirmation from BITMAIN with the tracking number?

I did, yes.  That said, I got four in a row for the same order, so the email system they are using may be flakey.

[ChuckBuck]

Bitmain Antminer S3 unboxing on youtube.

The unit itself looks really sturdy, and packaging looks just as good as the S1's package.

Can't wait to see farms of this little beast.  

[Mudbankkeith]

setup is the same as S1

Super Quiet    "NO NEED TO SHOUT"

cables are a bit warm, on S1 they were air cooled by the main fan.
on S3 they only have room air.
The plugs are cold,(small air flow from the hole in the case)

[sjc1490]

Did anyone who receive their units get a email confirmation from BITMAIN with the tracking number?

Yes, but by the time I got the email UPS had been showing them in their system for a day using the track by reference method.

[TracerX]

Bitmain just update their site to include a "News" section, although it's so far an empty page.  They may be preparing to shoot us an update.

[fragout]

silly question probably, but does it matter which 2 connectors you use out of the 4 available or is it better to connect up the 4 if you have them available? Will this option draw extra power?

[sjc1490]

silly question probably, but does it matter which 2 connectors you use out of the 4 available or is it better to connect up the 4 if you have them available? Will this option draw extra power?

Just use the front two or back two (1 on each side) no need to connect 4 unless you plan to OC and really need a 800w or higher psu to do that.

[sjc1490]

silly question probably, but does it matter which 2 connectors you use out of the 4 available or is it better to connect up the 4 if you have them available? Will this option draw extra power?

Just use the front two or back two (1 on each side) no need to connect 4 unless you plan to OC and really need a 800w or higher psu to do that.

To clarify for 2 units.

[aztecminer]

nice .. buy some geld extreme thermal paste from newegg and replace the thermal paste could see drop in temps and higher efficiency .

http://www.newegg.com/Product/Product.aspx?Item=N82E16835426035

[aztecminer]

[Can you supply the part number for the buck regulator component? It's not the big grey square, that's a coil - instead it'll be a small IC right beside and will have a code (or codes) written in ridiculously small letters which indicate the part and manufacturer. Someone in the S3 is going to research to see if we can replace?

http://www.ti.com/product/tps53355

same type as the S1/S2 and bitfury units both used. Its a good chip that can handle up to 35-40A if cooled and from a good batch.

MATH TIME

original spec: 32 chips running at 0.75V producing 15.75GH each = 504GH at 390W
you have 2 chips per TPS53355 regulator, 16 regulators total. 480A available at 30A draw

480*0.75 = 360W (assumes other 30W is the pair of 15W fans and the controller)  This was what Bitmain assumed would work

my guess is that either the chips are slightly less efficient than thought, perhaps drawing closer to 420W at 504GH. (420W total, or 560A, or 35A which is a point where not all regulators will be stable). Alternatively, the chip voltage is less than 0.75, thus drawing more amps. 35A draw would happen at 360W, 0.643V.  A combination of both factors is also possible.

gentlemen, start your multimeters!

put aluminum heat sinks on those ..

[jimrome]

MATH TIME

original spec: 32 chips running at 0.75V producing 15.75GH each = 504GH at 390W
you have 2 chips per TPS53355 regulator, 16 regulators total. 480A available at 30A draw

480*0.75 = 360W (assumes other 30W is the pair of 15W fans and the controller)  This was what Bitmain assumed would work

my guess is that either the chips are slightly less efficient than thought, perhaps drawing closer to 420W at 504GH. (420W total, or 560A, or 35A which is a point where not all regulators will be stable). Alternatively, the chip voltage is less than 0.75, thus drawing more amps. 35A draw would happen at 360W, 0.643V.  A combination of both factors is also possible.

gentlemen, start your multimeters!

Overvolting to 0.8v should be fine to have 10% extra hashing.
It will draw something like 450watt at the wall. 400watt before the PSU.
That makes slightly over 31A load on each regulator. With proper cooling, it should be OK.
You will be able to clock at 275 and 14ms delay.

0.85v will be too high load: 640A divided over the 16 regulators will be 40A per regulator.

Hopefully, my S3 will arrive next week so I'll be able to test.

For those who would want to reduce power draw in a few months, 0.72v will draw 270watt at wall while hashing at 430-440 GH/s

That buck regulator has been shown to be capable of >40A on a sustained basis (when fitted with heatsinks and forced air cooling). That said, batch to batch variation or any changes to the part internally could end up causing the part to meet only the rated specs. Additionally, the PCB design may not benefit from additional power (e.g. other bottlenecks)

We won't know until we try


I wonder if Bitmain was leaning on the capabilities of this part too much and ended up getting bit by parts not performing over-spec. I'd call that pretty stupid ballsy if the case

[R4v37]

MATH TIME

original spec: 32 chips running at 0.75V producing 15.75GH each = 504GH at 390W
you have 2 chips per TPS53355 regulator, 16 regulators total. 480A available at 30A draw

480*0.75 = 360W (assumes other 30W is the pair of 15W fans and the controller)  This was what Bitmain assumed would work

my guess is that either the chips are slightly less efficient than thought, perhaps drawing closer to 420W at 504GH. (420W total, or 560A, or 35A which is a point where not all regulators will be stable). Alternatively, the chip voltage is less than 0.75, thus drawing more amps. 35A draw would happen at 360W, 0.643V.  A combination of both factors is also possible.

gentlemen, start your multimeters!

Overvolting to 0.8v should be fine to have 10% extra hashing.
It will draw something like 450watt at the wall. 400watt before the PSU.
That makes slightly over 31A load on each regulator. With proper cooling, it should be OK.
You will be able to clock at 275 and 14ms delay.

0.85v will be too high load: 640A divided over the 16 regulators will be 40A per regulator.

Hopefully, my S3 will arrive next week so I'll be able to test.

For those who would want to reduce power draw in a few months, 0.72v will draw 270watt at wall while hashing at 430-440 GH/s

That buck regulator has been shown to be capable of >40A on a sustained basis (when fitted with heatsinks and forced air cooling). That said, batch to batch variation or any changes to the part internally could end up causing the part to meet only the rated specs. Additionally, the PCB design may not benefit from additional power (e.g. other bottlenecks)

We won't know until we try


I wonder if Bitmain was leaning on the capabilities of this part too much and ended up getting bit by parts not performing over-spec. I'd call that pretty stupid ballsy if the case

maybe they ordered, and it'll cost them again if they want to change the part
so they just stick to the 1st plan

if they just send samples to some users here, i mean the expert in hardware like this
let them test and change some parts in the sample to make the best and efficient miner
then they'll be making a really damn good miner

but... i dont know, lets hope they do it like this before they release or bulk processing a miners

[fragout]

Sigh . Ups arrived with a bill for 176 Euro for 2 S3s. Problem is there was only one one box. I didnt want to pay until I got the 2 boxes and have the hassle of tracking down the other one since it would be scanned as delivered, so Ill have to wait until monday

[jpchrist]

Sigh . Ups arrived with a bill for 176 Euro for 2 S3s. Problem is there was only one one box. I didnt want to pay until I got the 2 boxes and have the hassle of tracking down the other one since it would be scanned as delivered, so Ill have to wait until monday

How do these machines have any chance to ROI in Europe with the electricity prices and import taxes!?!?

[rograz]

How do these machines have any chance to ROI in Europe with the electricity prices and import taxes!?!?

magic

[dropt]

magic

That's pretty much the only explanation.