ANTMINER S7 is available at bitmaintech.com with 4.86TH/s, 0.25J/GH

[dmwardjr]

Jury is still out on exactly what they have done, however I think it is highly unlikely that they would use multiple board types. These were my thoughts on the 135 chip Batch 6.

https://bitcointalk.org/index.php?topic=1219211.msg12902245#msg12902245
https://bitcointalk.org/index.php?topic=1219211.msg12911092#msg12911092

However for me the big outstanding issue is if this is a 15 Node chain with a resultant 0.8V Core Voltage, what has really happened to the Power Consumption & J/GH?


Rich

That's my main question as well.  More chips equals less voltage dropped across each while achieving more hashing power from more chips.  Less chips equals more voltage drop across each chip but with less hashing power because of fewer chips.  One can't help but wonder if this rig is going to be 0.8V Core Voltage with 12 Volts Source Voltage.

It will be interesting to see the performance level from those who receive them who will share in the forum.  I almost bought a couple of them but held off.  Mainly because I want to see the actual performance before taking the plunge.  I believe I'm still diving in regardless.  I just want to see what I'm actually getting before I purchase.

[RichBC]

That's my main question as well.  More chips equals less voltage dropped across each while achieving more hashing power from more chips.  Less chips equals more voltage drop across each chip but with less hashing power because of fewer chips.  One can't help but wonder if this rig is going to be 0.8V Core Voltage with 12 Volts Source Voltage.

It will be interesting to see the performance level from those who receive them who will share in the forum.  I almost bought a couple of them but held off.  Mainly because I want to see the actual performance before taking the plunge.  I believe I'm still diving in regardless.  I just want to see what I'm actually getting before I purchase.

I wonder how long we will have to wait before we see a picture or for someone not worried about Warranty to take one apart? In prospect I prefer the spec of the 135 Chip system as I think it will be less on edge and will provide more scope for Under & Over Clocking and Volting. In the meantime we will continue to speculate, which is of course a major facet of this game. 

Rich

[dmwardjr]

I wonder how long we will have to wait before we see a picture or for someone not worried about Warranty to take one apart?

Probably at least 90 days after warranty.  I'll open mine up after 90 days for sure to have a look.

In prospect I prefer the spec of the 135 Chip system as I think it will be less on edge and will provide more scope for Under & Over Clocking and Volting.

Agreed.  I can see these remaining in the game a long time.  IMHO.

In the meantime we will continue to speculate, which is of course a major facet of this game. 

Rich

Agreed   

[mikefallen]

i got one shipped for 4.15......my BTC from coinbase didn't come in quick enough to get another one tho

[philipma1957]

coins moved up to 335 on coinbase today.

Batch 7 should show soon.

[mikefallen]

my poor poor f5 button

[Finksy]

my poor poor f5 button

LOL right!?!?!?

[vortexz]

do we know for sure there will be an batch 7 ?

[Biodom]

do we know for sure there will be an batch 7 ?

why not? they seem to be selling.
maybe they are waiting to see the other company miner before committing to prices & specs.

[dmwardjr]

On the Chinese page, I Bing translated the following on S7 for sell there:

Ant mine S7 third instalment sale, shipment date is December 02-December 12, the price is 2500 per t, order order, may be shipped in advance. Ant mine S7 using the 28 nanometer full custom IC BM1385. Count 4.46T, 1160 w power consumption, dual fan, separate heat sink.

Translated specs:

1. is rated: 4.05 TH/s
2. power on the wall: 1042 w +10% (bit-city, APW3-1600-Watt power supplies, AC/DC 93% efficiency, 25 ° c ambient temperature)
3. power: 0.25 j/GH (wall, AC/DC 93% efficiency, 25 ° c ambient temperature)
4. rated voltage: 12V ± 5%, should not be less than 12.0V
5. chip size: 135 BM1385
6. dimensions: 301 mm (l) *123 mm (w) *155 mm (h)
7. cooling: 2x12038 fan
8. working temperature: 0 ℃ to 40 ℃
9. network connection: Ethernet

I just think it's odd it says, 4.6 TH/s @ "1,160" watts on the Product page.  Then you click the product (S7) and the specifications say, 4.05 TH/s @ "1042" watts

[bbOOmm]

Well .... I think I used up my luck for the year ....

I found 4 blocks in 3 days ... with an S5...


Too bad I was not solo mining ....

[Biodom]

On the Chinese page, I Bing translated the following on S7 for sell there:

Ant mine S7 third instalment sale, shipment date is December 02-December 12, the price is 2500 per t, order order, may be shipped in advance. Ant mine S7 using the 28 nanometer full custom IC BM1385. Count 4.46T, 1160 w power consumption, dual fan, separate heat sink.

Translated specs:

1. is rated: 4.05 TH/s
2. power on the wall: 1042 w +10% (bit-city, APW3-1600-Watt power supplies, AC/DC 93% efficiency, 25 ° c ambient temperature)
3. power: 0.25 j/GH (wall, AC/DC 93% efficiency, 25 ° c ambient temperature)
4. rated voltage: 12V ± 5%, should not be less than 12.0V
5. chip size: 135 BM1385
6. dimensions: 301 mm (l) *123 mm (w) *155 mm (h)
7. cooling: 2x12038 fan
8. working temperature: 0 ℃ to 40 ℃
9. network connection: Ethernet

I just think it's odd it says, 4.6 TH/s @ "1,160" watts on the Product page.  Then you click the product (S7) and the specifications say, 4.05 TH/s @ "1042" watts

2500/t is ~$1589 (if 4.05, not 4.66), assuming 2500 are RMB/per th
maybe it is that internal (chinese only) consumption unit that has 550 mhz

[hodedowe]

do we know for sure there will be an batch 7 ?

why not? they seem to be selling.

They just recently changed chip size. There'll be a batch 7 unless chip availability becomes an issue.


On another note, I've been reading S7 support threads lately and I'm a noticing a theme around heat related issues with the earlier batches. I wonder if they pushed the S7 family out the door a bit soon and B6 was nothing more than a drastic correction to avoid burnouts. (house fires?)

[kaltar]

Well .... I think I used up my luck for the year ....

I found 4 blocks in 3 days ... with an S5...


Too bad I was not solo mining ....

Those are not bitcoin blocks, probable a new low diff coin.

[dmwardjr]

Those are not bitcoin blocks, probable a new low diff coin.

I was wondering the same thing after seeing the "best shares."

[dmwardjr]

On another note, I've been reading S7 support threads lately and I'm a noticing a theme around heat related issues with the earlier batches. I wonder if they pushed the S7 family out the door a bit soon and B6 was nothing more than a drastic correction to avoid burnouts. (house fires?)

I think it's more to do with many using PSU's with less than 12.0 volts DC.  Hence, the new design that's able to run stable with low NWE's at 11.6 Volts DC.  Many may have created tickets because of PSU's lacking 12.0 Volts minimum.


Could be what you said though.  However, I hope possible fire is not the case.

[bbOOmm]

Well .... I think I used up my luck for the year ....

I found 4 blocks in 3 days ... with an S5...


Too bad I was not solo mining ....

Those are not bitcoin blocks, probable a new low diff coin.

Oh, damn. I didn't think of that.

Nicehash has merged mining?

[64dimensions]

One other concern I have about the S7 are the heat sinks.

Earlier Bitmain models used a mechanically fastened heat sinks. All the PC motherboards I have ever seen do the same thing. In the mechanical setup, thermally conductive paste transfers the heat from the chip to the finned Al radiator.

The S7 appears to have individual heat sinks that are glued to chip. Hopefully BITMAIN is not trying to cheap out here. The thermal grease just has to be an excellent heat conductor, but a thermal glue has to be an excellent conductor AND good adhesive at temperature AND withstand temperature cycling.

3M has a special heat conductive tape family: http://www.adafruit.com/datasheets/3m8810.PDF

From this data sheet:

Per UL-746C testing, the maximum temperature rating for the 8805, 8810 or 8815 is 100°C when tested on glass epoxy or an
anodized aluminum substrate. The products meet the UL-746C test requirement of maintaining at least 50% of their initially
tested Overlap Shear strength after heat aging for 1000 hours at 150°C. (See UL file #MH17478 for details):
Additional testing completed and passed per UL-746C test methods include:
– Effect of Humidity: 7 days @ 95% Relative Humidity (RH) @ 60°C.
– Effect of Environmental Cycling (3 cycles): 1 cycle = 24h immersed in 25°C water / 24h @ 100°C / 96h @ 35°C @
90% RH / 8 h @ -35°C.
Note: 3M™ Thermally Conductive Adhesive Transfer Tape 8820 has not been tested per UL-746C or UL-746A test
procedures.

If heat sinks are falling off S7's there could be several things, an inferior product or the chip or Al surfaces are not clean or at final test they are running the chips too hot and damaging the bond to the heat sink.

[dmwardjr]

Oh, damn. I didn't think of that.


Nicehash has merged mining?

I know they mine several different coins.  I would not be surprised if there is merged mining there.

[bbOOmm]

On another note, I've been reading S7 support threads lately and I'm a noticing a theme around heat related issues with the earlier batches. I wonder if they pushed the S7 family out the door a bit soon and B6 was nothing more than a drastic correction to avoid burnouts. (house fires?)

I think it's more to do with many using PSU's with less than 12.0 volts DC.  Hence, the new design that's able to run stable with low NWE's at 11.6 Volts DC.  Many may have created tickets because of PSU's lacking 12.0 Volts minimum.


Could be what you said though.  However, I hope possible fire is not the case.

I cracked open my power supplies and tweaked them to be putting out 12.25V +/- 0.005V at a 50% load. It made a WORLD of difference with the S7's. My power supplies were ranging from 11.6V to 11.8V under load, I was getting some hardware errors but it wasn't so bad. I was trying to see if the voltage was the reason why the fifo buffers keep tossing the error that they are empty. It didn't fix that issue, but the HW errors went to 0.0000x ( beyond the readings ability )

For information, I'm using Delta DPS-1200FB 1200W power supplies, 2 hashboards per power supply, and used an S5 pulling right at 600W as the load when calibrating all the power supplies.